<%BANNER%>

Assessment of the Risk of Loss of Subsidized Multifamily Housing, a Simulation of Net Cash Flow

Permanent Link: http://ufdc.ufl.edu/UFE0024721/00001

Material Information

Title: Assessment of the Risk of Loss of Subsidized Multifamily Housing, a Simulation of Net Cash Flow
Physical Description: 1 online resource (215 p.)
Language: english
Creator: Roset-Zuppa, Patricia
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: 8, affordability, apartment, assistance, cash, florida, flow, housing, multifamily, opt, out, policy, prepayment, preservation, program, public, rental, restriction, risk, section, simulation, subsidy, termination, use
Design, Construction, and Planning -- Dissertations, Academic -- UF
Genre: Design, Construction, and Planning Doctorate thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The affordability gap between household income and the cost of rental housing has been an issue for many families, exacerbated by the loss of subsidized units due to deterioration or conversion to market-rate housing. Preventing the loss of affordable housing has been hindered by a lack of systematic risk analysis methods to identify and examine properties at risk of loss. The main research goal was to characterize the properties at heightened risk of loss. The purpose was twofold: To inform governments to allocate resources; and to provide a tool to target properties for preservation. The research methodology was simulation modeling of the net cash flow of properties that received HUD project-based rental assistance in Duval and Miami-Dade County in Florida. Simulation modeling was performed as a tool to incorporate uncertain variables. A net cash flow approach was used, because net operating income (NOI) provides an indication of a property s financial and physical health. A fail-out risk model was created to identify properties at higher risk of deterioration and default, as measured by a lower NOI, debt coverage ratio below 1.0, or a failing HUD physical inspection score. An opt-out risk model was designed to analyze properties at higher risk of conversion, as measured by for-profit ownership, an increase in NOI of at least 20% if market rents are charged, rental assistance contract expiration by 2014, original contract term, and low poverty area. Significance tests were performed to compare characteristics of higher and lower risk properties by risk model. Stepwise regression analysis was conducted to analyze the impact of variables on NOI. The conclusion was made that properties should be flagged at heightened fail-out risk according to three indicators: Project rent to Fair Market Rent (FMR) ratio (lower ratio, higher risk), year built (earlier year, higher risk) and HUD inspection score (below 60, higher risk). The study also concluded that the key indicator for opt-out risk was project rent to FMR ratio (lower ratio, higher risk). Once a shortlist of properties is created based on this indicator, the shortlist could be refined by applying other criteria such as subsidy expiration year.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Patricia Roset-Zuppa.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Stroh, Robert C.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024721:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024721/00001

Material Information

Title: Assessment of the Risk of Loss of Subsidized Multifamily Housing, a Simulation of Net Cash Flow
Physical Description: 1 online resource (215 p.)
Language: english
Creator: Roset-Zuppa, Patricia
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: 8, affordability, apartment, assistance, cash, florida, flow, housing, multifamily, opt, out, policy, prepayment, preservation, program, public, rental, restriction, risk, section, simulation, subsidy, termination, use
Design, Construction, and Planning -- Dissertations, Academic -- UF
Genre: Design, Construction, and Planning Doctorate thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The affordability gap between household income and the cost of rental housing has been an issue for many families, exacerbated by the loss of subsidized units due to deterioration or conversion to market-rate housing. Preventing the loss of affordable housing has been hindered by a lack of systematic risk analysis methods to identify and examine properties at risk of loss. The main research goal was to characterize the properties at heightened risk of loss. The purpose was twofold: To inform governments to allocate resources; and to provide a tool to target properties for preservation. The research methodology was simulation modeling of the net cash flow of properties that received HUD project-based rental assistance in Duval and Miami-Dade County in Florida. Simulation modeling was performed as a tool to incorporate uncertain variables. A net cash flow approach was used, because net operating income (NOI) provides an indication of a property s financial and physical health. A fail-out risk model was created to identify properties at higher risk of deterioration and default, as measured by a lower NOI, debt coverage ratio below 1.0, or a failing HUD physical inspection score. An opt-out risk model was designed to analyze properties at higher risk of conversion, as measured by for-profit ownership, an increase in NOI of at least 20% if market rents are charged, rental assistance contract expiration by 2014, original contract term, and low poverty area. Significance tests were performed to compare characteristics of higher and lower risk properties by risk model. Stepwise regression analysis was conducted to analyze the impact of variables on NOI. The conclusion was made that properties should be flagged at heightened fail-out risk according to three indicators: Project rent to Fair Market Rent (FMR) ratio (lower ratio, higher risk), year built (earlier year, higher risk) and HUD inspection score (below 60, higher risk). The study also concluded that the key indicator for opt-out risk was project rent to FMR ratio (lower ratio, higher risk). Once a shortlist of properties is created based on this indicator, the shortlist could be refined by applying other criteria such as subsidy expiration year.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Patricia Roset-Zuppa.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Stroh, Robert C.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024721:00001


This item has the following downloads:


Full Text

PAGE 1

1 ASSESSMENT OF THE RISK OF LOSS OF SUBSIDIZED MULTIFA MILY HOUSING, A SIMULATION OF NET CASH FLOW By PATRICIA E. ROSET-ZUPPA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009

PAGE 2

2 2009 Patricia E. Roset-Zuppa

PAGE 3

3 To my Sunshine

PAGE 4

4 ACKNOWLEDGMENTS Special thanks go to Dr. Robe rt Stro h, my Committee Chair a nd Director of the Shimberg Center for Housing Studies, for all his guidance a nd encouragement. I am also thankful for my committee members, Dr. Marc Smith, Dr. Jose li Macedo, Dr. Anne Williamson and Dr. Wayne Archer, for sharing their valuab le time and immense expertise. Bill ODell, Associate Director of the Shim berg Center for Housing Studies, deserves many thanks for providing me with the financia l assistance and the opport unity to contribute to the research of Flor idas housing issues. I am grateful for Dr. Paul Zwick, Associate Dean in the College of Design, Construction and Planning, who acted as my advisor in the first year and who gave me the opportunity to gain teaching experience. The College of Design, Construction and Planning is to thank for awarding me a Grinter Fellowship. I wish to acknowledge my parents, Piet and Je anne Roset, for never doubting my decisions and always providing me with unconditional su pport, no matter where in the world I live. It is impossible to express my gratitude for the two people that are the center of my universe, my husband Dino and daughter Julia. Th is journey became so much more enjoyable when Dino joined me on the Ph.D. path and when Julia entered our lives.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4 LIST OF TABLES ...........................................................................................................................8 LIST OF FIGURES .......................................................................................................................12 ABSTRACT ...................................................................................................................... .............13 CHAP TER 1 INTRODUCTION .................................................................................................................. 15 Statement of the Problem ...................................................................................................... ..15 Preservation Defined ..............................................................................................................17 Research Goal, Purpose and Questions ..................................................................................17 Research Methodology ...........................................................................................................18 Programmatic and Geographic Scope ....................................................................................19 Organization of the Study .......................................................................................................20 2 THE RATIONALE FOR PRESERVATI ON OF AFFORDABLE HOUSING ..................... 22 Demographic Shifts and Implications for Housing De mand .................................................. 22 Housing Supply Characteristics and Trends ........................................................................... 24 Affordability and Cost Burden Defined .................................................................................26 Affordability in Florida ...................................................................................................... .....27 Rationale for Preservation ......................................................................................................29 Obstacles to Preservation ..................................................................................................... ...30 3 FEDERAL GOVERNMENT PROGRAMS AND IMPLICATIONS FOR OWNER CHOICES ....................................................................................................................... ........33 First Government Involvement in Housing: Mortgage Insurance and Public Housing .........33 Interest Rate Subsidy Programs: Sect ion 221(d)(3) BMIR and Section 236 ......................... 35 Section 221(d)(3) Below Market Interest Rate ............................................................... 35 Section 236 ......................................................................................................................37 Rental Assistance Programs: Section 8 .................................................................................. 38 Housing Preservation Act .......................................................................................................41 Programmatic Reasons for Loss of Affordability and Owner Choices .................................. 43 4 RISK AND RISK AS SESSMENT M ETHODS.....................................................................48 Definition of Risk ...................................................................................................................48 The Role of Property Data ......................................................................................................49 Risk Assessment Method Type One: Target Inventory .......................................................... 52

PAGE 6

6 General Description and Purpose .................................................................................... 52 Methodology and Risk Indicators ...................................................................................52 Data Sources ....................................................................................................................56 Output Format and Updates .............................................................................................57 Risk Assessment Method Type One: Risk Rating ..................................................................57 General Description and Purpose .................................................................................... 57 Methodology and Risk Indicators ...................................................................................58 Data Sources ....................................................................................................................60 Output Format and Updates .............................................................................................60 Risk Assessment Method Type Two: Cross Tabulations ....................................................... 60 General Description and Purpose .................................................................................... 60 Methodology and Risk Indicators ...................................................................................61 Data Sources ....................................................................................................................62 Output Format and Updates .............................................................................................63 Risk Assessment Method Type Two: Regression Analysis ...................................................63 General Description and Purpose .................................................................................... 63 Methodology and Risk Indicators ...................................................................................64 Data Sources ....................................................................................................................66 Output Format and Updates .............................................................................................67 Risk Assessment Method Type Three: Simulation Modeling ................................................ 67 General Description and Purpose .................................................................................... 67 Methodology and Risk Indicators ...................................................................................68 Data Sources ....................................................................................................................70 Output Format and Updates .............................................................................................71 Other Risk Indicators ..............................................................................................................71 5 METHODOLOGY ................................................................................................................. 80 Research Overview ............................................................................................................. ....80 Net Cash Flow Approach to Fail-out Risk ............................................................................. 81 Net Cash Flow Approach to Opt-out Risk .............................................................................. 82 Simulation Modeling of Net Cash Flow ................................................................................. 83 Overview of Monte Carlo Simulation and Application to Real Estate ........................... 83 Input and Output variables .............................................................................................. 85 Probability Distributions ................................................................................................. 85 Rationale for Simulation Modeling .................................................................................87 Research Design .....................................................................................................................88 Unit of Analysis, Population and Sampling Frame ......................................................... 88 Data Collection and Database Design .............................................................................90 Fail-out Risk: Structure of the Mode l and Methodological Assum ptions ....................... 93 Opt-out Risk: Structure of the Mode l and Methodological Assum ptions ....................... 94 Input Variables and Prob ability Distributions .................................................................96 Output Variables ............................................................................................................103 Data Analysis and Analytical Tools .....................................................................................103 Descriptive Analysis and Significance Tests ................................................................ 103 Correlation and Multiple Regression Analysis .............................................................. 105 Scenario Analysis ..........................................................................................................106

PAGE 7

7 Analytical Tools ............................................................................................................106 Limitations of Net Cash Flow A pproach and Simulation Modeling ....................................107 6 DATA ANALYSIS .............................................................................................................. 112 Input Variables and Simulation Re sults: Descriptive Analysis ............................................112 NOI and DCR Output Analysis ............................................................................................114 Descriptive Analysis and Significance Tests ................................................................ 114 Correlation and Multiple Regression Analysis .............................................................. 117 Fail-out Model: Descri ptive Analysis and Significance Tests ............................................. 119 Opt-out Model ......................................................................................................................125 Descriptive Analysis and Significance Tests ................................................................ 125 Correlation and Scenario Analysis ................................................................................ 128 7 SUMMARY, CONCLUSIONS AND RECOMMENDATIONS ........................................166 Summary and Conclusion s: Fail-out Risk ............................................................................ 166 Summary and Conclusion s: Opt-out Risk ............................................................................ 168 Recommendations for Policy ................................................................................................171 Recommendations for Future Research ................................................................................172 APPENDIX A INPUT VARIABLES AND SIMULATION RESULTS ..................................................... 175 B OUTPUT RESULTS BY PROPERTY, OUTPUT VARIABLE AND MODEL ................180 REFERENCES .................................................................................................................... ........204 BIOGRAPHICAL SKETCH .......................................................................................................215

PAGE 8

8 LIST OF TABLES Table page 4-1 Example of Cross Tabulations for Tena nt Characteristics in H UD Properties ................. 74 4-2 Regression Model Variables for the L ogistic Regression Model of the Opt-out Decision .............................................................................................................................75 4-3 Coefficient Estimates of the Logistic Regression Model of the Opt-out D ecision ............ 77 5-1 Datasets Used to Create the Database .............................................................................. 110 5-2 Net Operating Income Benchmarks ................................................................................. 110 5-3 Number of Properties by Cond ition for Higher Fail-out Risk ......................................... 110 5-4 Number of Properties by C ondition for Higher Opt-out Risk ......................................... 111 5-5 Input Variables, Distribu tions and Param eters for the Fail-out and Opt-out Risk Models..............................................................................................................................111 6-1 Summary of Estimated Net Operating In com e Values per Unit per Month for All Properties .........................................................................................................................130 6-2 Comparison of the Simulated Mean a nd Median NOI Values to Reported NOI Values for Non-Market Rent Properties Per Unit per Month in Jacksonville MSA and Miam i MSA .............................................................................................................. 131 6-3 Number of Properties with an Estimated Net Operati ng Incom e Below $200 per Unit per Month by Probability for All Properties .................................................................... 132 6-4 Summary of Estimated Debt Coverage Ratio .................................................................. 132 6-5 Summary of Estimated Net Operating In com e Values per Unit per Month by Rent Scenario for Opt-out Risk Model .....................................................................................133 6-6 Significance Test of Mean NOI by Re nt Scenario for Opt-out Risk Model .................... 134 6-7 Percentage Change in Estimated NOI Value from Project R ents to Market Rents for the Opt-out Risk Model ................................................................................................... 134 6-8 Number of Properties with an Estimated Change of Net Operating Income of At Least 20% by Probability for the Opt-out Risk Model ....................................................135 6-9 Summary of Estimated Debt Coverage Ratio for the Opt-out Risk Model ..................... 136 6-10 Correlation Matrix ....................................................................................................... ....137

PAGE 9

9 6-11 Results of Stepwise Regression for All Prope rties with Mean Net Operating Income as the Dependent Variable ...............................................................................................138 6-12 Property Characteristics for the Fail-out Model ..............................................................140 6-13 Significance Test of Total Units per Property by Risk Group for the Fail-out Risk Model ......................................................................................................................... ......141 6-14 Significance Test of Average Unit Size by Risk Group for the Fail-out Risk Model ..... 142 6-15 Significance Test of Proportion of Elderly Properties by Risk Group for the Fail-out Risk Model .......................................................................................................................142 6-16 Significance Test of Proportion of Non-pr ofit O wners by Risk Group for the Fail-out Risk Model .......................................................................................................................143 6-17 Significance Test of Year Built by Ri sk Group for the Fail-out Risk Model ..................143 6-18 Significance Test of REAC Physical In spection Score by Risk Group for the Fail-out Risk Model .......................................................................................................................144 6-19 Financial Characteristics for the Fail-out Model .............................................................144 6-20 Subsidy Characteristics for the Fail-out Model ............................................................... 145 6-21 Significance Test of Rent to Fair Mark et Rent Ratio by Risk Group for the Fail-out Risk Model .......................................................................................................................147 6-22 Significance Test of Proportion of Properties with a LMSA contract by Risk Group for the Fail-out Risk Model ..............................................................................................147 6-23 Significance Test of Proportion of Renewe d Contracts by Risk Group for the Fail-out Risk Model .......................................................................................................................148 6-24 Significance Test of Contract Effectiv e Year by Risk Group for the Fail-out Risk Model ......................................................................................................................... ......148 6-25 Significance Test of Contract Expiration Year by Risk Group for the Fail-out Risk Model ......................................................................................................................... ......149 6-26 Significance Test of Number of Funding Layers by Risk Group for the Fail-out Risk Model ......................................................................................................................... ......149 6-27 Tenant Characteristics for the Fail-out Model .................................................................150 6-28 Significance Test of Percentage of Fe m ale Heads of Household with Children by Risk Group for the Fail-out Risk Model ..........................................................................152

PAGE 10

10 6-29 Significance Test of Percentage of Head s of Households at Age 62 or Older by Risk Group for the Fail-out Risk Model ..................................................................................152 6-30 Significance Test of Percentage of Ex trem ely Low Income Households by Risk Group for the Fail-out Risk Model ..................................................................................153 6-31 Significance Test of Household Income as a Percentage of th e Local Median Family Income by Risk Group for the Fail-out Risk Model ........................................................ 153 6-32 Significance Test of Annual Household Incom e by Risk Group for the Fail-out Risk Model ......................................................................................................................... ......154 6-33 Significance Test of Percentage of Mi nority Households by Risk Group for the Failout Risk Model .................................................................................................................154 6-34 Significance Test of Percentage of Ov erhoused H ouseholds by Risk Group for the Fail-out Risk Model .........................................................................................................155 6-35 Property Characteristic s for the Opt-out Model ............................................................... 156 6-36 Significance Test of Total Units per Property by Risk Group for the Opt-out Risk Model ......................................................................................................................... ......157 6-37 Significance Test of Proportion of Elderly Properties by Risk Group for the Opt-out Risk Model .......................................................................................................................158 6-38 Significance Test of Average Unit Size by Risk Group for the Opt-out Risk Model ..... 158 6-39 Significance Test of Year Built by Ri sk Group for the Opt-ou t Risk Model .................. 159 6-40 Significance Test of REAC Physical In spection Score by Risk Group for the Opt-out Risk Model .......................................................................................................................159 6-41 Financial Characteristics for the Opt-out Model .............................................................159 6-42 Significance Test of the Mean NOI Ch ange by Risk Group for the Opt-out Risk Model ......................................................................................................................... ......160 6-43 Subsidy Characteristics for the Opt-out Model ............................................................... 161 6-44 Significance Test of Rent to Fair Mark et Rent Ratio by Risk Group for the Opt-out Risk Model .......................................................................................................................162 6-45 Significance Test of Number of Funding Layers by Risk Group for the Opt-out Risk Model ......................................................................................................................... ......163 6-46 Tenant Characteristics for the Opt-out Model ................................................................. 164 6-47 Comparison of Rents by County and Bedroom Size ....................................................... 165

PAGE 11

11 A-1 Input Variables and Simulation Results ...........................................................................176 B-1 Output Results by Property, Output Variable and Model ................................................ 181

PAGE 12

12 LIST OF FIGURES Figure page 2-1 Population of the State of Florida, 1900-2060.. ................................................................. 31 2-2 Median Family Income, Median Sales Price and Num ber of Sales of Existing SingleFamily Homes in Florida, 1996-2008.. ..............................................................................31 2-3 Units Authorized by Buildi ng Perm its in Florida, 1996-2008.. ......................................... 32 4-1 Predicted Number of Older Assisted HUD Properties and Units by Property Status under Current HUD Obl igations for 1990-1994.. .............................................................. 78 4-2 Predicted Number of Older Assisted HUD Properties and Units by Property Status under Full F unding Scenario for 1990-1994. ..................................................................... 79 5-1 Probability Distributions. ................................................................................................ .109 6-1 Mean Net Operating Income ............................................................................................ 131 6-2 Mean Net Operating Income under Two Scenarios of Opt-out Model ...........................135

PAGE 13

13 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ASSESSMENT OF THE RISK OF LOSS OF SUBSIDIZED MULTIFA MILY HOUSING, A SIMULATION OF NET CASH FLOW By Patricia E. Roset-Zuppa August 2009 Chair: Robert C. Stroh Major: Design, Construction, and Planning The affordability gap between household income a nd the cost of rental housing has been an issue for many families, exacerbated by the loss of subsidized units due to deterioration or conversion to market-rate housing. Preventing the loss of affordable housing has been hindered by a lack of systematic risk analysis methods to identify and examine properties at risk of loss. The main research goal was to characterize the properties at heightened risk of loss. The purpose was twofold: To inform governments to allocate resources; and to provi de a tool to target properties for preservation. The research methodology was simulation modeling of the net cash flow of properties that received HUD project-based rental assistance in Duval and Miami-Dade County in Florida. Simulation modeling was performed as a tool to in corporate uncertain variab les. A net cash flow approach was used, because net operating income (NOI) provides an indication of a propertys financial and physical health. A fa il-out risk model was created to identify properties at higher risk of deterioration and defau lt, as measured by a lower NOI, debt coverage ratio below 1.0, or a failing HUD physical inspection score. An opt-out risk model was designed to analyze properties at higher risk of conversion, as measured by forprofit ownership, an increas e in NOI of at least

PAGE 14

14 20% if market rents are charged, rental assistance contract expi ration by 2014, orig inal contract term, and low poverty area. Significance tests were performed to compar e characteristics of higher and lower risk properties by risk model. Stepwise regression analysis was conducted to analyze the impact of variables on NOI. The conclusion was made that properties should be flagge d at heightened failout risk according to three indicato rs: Project rent to Fair Market Rent (FMR) ratio (lower ratio, higher risk), year built (earlier year, higher risk) and HUD insp ection score (below 60, higher risk). The study also concluded that th e key indicator for opt-out risk was project rent to FMR ratio (lower ratio, higher risk). Once a shortlist of pr operties is created based on this indicator, the shortlist could be refined by applying othe r criteria such as subs idy expiration year.

PAGE 15

15 CHAPTER 1 INTRODUCTION Statement of the Problem The affordability gap between household income and the cost of rental housing has been an issue for many families across the country. Th e U.S. Census Bureau (2007a) estimated that almost 46% of renter households nationwide were cost-burdened, spending at least 30% of their income on housing; almost 23% pay at least half their income on rent. The U.S. Department of Housing and Urban Development (HUD) Secretar y Shaun Donovan referred to the problem of affordability as a persistent crisis that far too many low-income Americans face (Rice 2009, 1). As reported by the Center on Budget and Policy Pr iorities, exacerbating the affordability issue has been the reduction in federal funding for re ntal assistance programs such as the Housing Choice Vouchers, and funding shortfalls for progra ms such as Section 8 Project-Based Rental Assistance (Rice 2009; Rice and Sard 2007). On a national level, new construction of re ntal housing has slowed down year over year since 2000, despite a surge in the completion of new units in re sponse to a stronger demand for rentals in various geographic poc kets (Joint Center for Housi ng Studies 2008). While about 2.3 million new units in multifamily structures were added nationwide between 1993 and 2003, many of these were not affordable to low-income families for two reasons (Joint Center for Housing Studies 2006). First, much of the new c onstruction of unassisted rental housing was in the form of high-cost luxury ap artments (Katz and Austin Turn er 2007). Second, the majority of new construction of subsidized rental housing was realized unde r the Low Income Housing Tax Credit program, which generally targets households at 60% of the area median income (AMI), not families in the lowest income brackets. Ne w supply of rental housing cannot catch up to the demand for assistance (Joint Cent er for Housing Studies 2007).

PAGE 16

16 Another challenge of the existing rental stoc k is that both unassisted and assisted (i.e., subsidized) units are lost each year. Between 1993 and 2003, more than 1.5 million units were removed from the affordable rental stock, e ither through abandonment, demolition or conversion (Joint Center for Housing Studies 2006). Subsidized rental properties have also experienced a loss of affordability due to the termination of subsidies. The Nationa l Housing Trust (2004) analyzed the loss of HUD-assisted proj ect-based multifamily units during 1995 to 2003 throughout the country. It estimated a net loss of 300,000 units as a resu lt of prepayment of subsidized mortgages and opt-out of rental assistance contract s. HUD has recognized the impact of the loss. The National Housing Law Project (1999, 7) reported th at for the first time, HUDs annual report on worst-case housing needs [i n 1997] specifically cited the loss of HUD multifamily housing stock as a contributing factor in the growing gap between the nations needs and available units. While the need for affordable housing is growing, the supply is dwindling and new construction cannot keep pace, nor reach the lowe st income households that have the greatest need. Therefore, the preservation of existing affo rdable rental properties is another strategy to address the need for housing. A si gnificant hurdle to preservation has been that there are no widely available standardized risk analysis t ools to assist states and local governments in identifying and examining properties that may be facing expiration and/or opt-out situations so that preservation strategies can be built around the specific needs of each property (Affordable Housing Study Commission 2005, 24). A related issue ha s been the lack of current and historical multifamily data that are publicly available and comprehensive. Data are necessary for facilitating analysis of the portfolio and identification of at -risk properties (Center for Housing Policy 2008). The risk analysis and data hurdles may explain w hy the Joint Center for Housing

PAGE 17

17 Studies reported that the country has seen only piecemeal preservation effort s (Joint Center for Housing Studies of Harvard University 2006, 29). However, in recent years, the research of the preservation issue as well as the funding of pres ervation projects have become a higher priority nationwide among state and local governments, foundations such as the MacArthur Foundation, research institutes such as the Center for Hous ing Policy and the Shimberg Center for Housing Studies, and advocacy bodies such as the National Low Income Housing Coalition. Preservation Defined Preservation of assisted multifamily housi ng means maintaining affordability for lowincome households for an extended period, a nd keeping the propertie s in good physical and financial condition (The John D. and Cather ine T. MacArthur Foundation 2007). Assisted properties have finite periods of affordability under the terms of their subsidies and use restrictions. Upon the prepayment or maturity of a subsidized mortgage, th e expiration of a use restriction or the opt-out of a rental assistance program, the housi ng will likely be lost to lowincome families unless other funding can be secured to keep it affordable. A property is also at risk of losing affordability if it faces large capital needs and repa irs, but lacks the reserves, cash flow and/or owner motivation to address these. This situation can lead to mortgage default, foreclosure and displ acement of tenants. Research Goal, Purpose and Questions The main goal of the dissertation research was to identify the types of properties at heightened risk of loss to the affordable assist ed rental housing stock. Types of properties were categorized according to property, financial, subsidy and tenant characteristics. The purpose of this research was twofold. The first purpose was to inform governments about the risk of displacement of low-income re sidents and the risk of community disinvestment in the case of deteriorating properties. This information can be used by policy-makers and

PAGE 18

18 planners to allocate resources. The second purpose was to provide a tool for housing advocates to identify properties at heightened risk of loss. Advocates can approach th ese properties to offer assistance in the form of legal aid, financial re sources, or financing and development expertise. Advocates can also target these properties for acquisition and preservation. The following two main questions drove the research: What are the property, financial, subsidy and tena nt characteristics of properties identified at fail-out risk, as measured by the financial or physical condition? What are the property, financial, subsidy and tena nt characteristics of properties identified at opt-out risk, as measured by the opportunity to increase project rents and improve cash flow? In order to address the ma in research questions, the li terature review focused on addressing the following subset of questions: What are the terms and conditions of the major federal subsidy programs? What options are available to property owners concerning termination and continuation of subsidies and use restrictions? Which variables are indicators that a property is at risk of deteri oration and default? Which variables are indicators that a property is at risk of conversion? Which risk assessment methods have been used to analyze the affordable housing stock? Research Methodology The research methodology that was applie d to address the rese arch questions was simulation modeling of the net cas h flow of properties that re ceived HUD project-based rental assistance in Duval and Miami-Dade County in Fl orida. A cash flow approach was used, because net operating income (NOI) provides an indication of the financial and ph ysical health of a property. The financial and physic al condition impacts the likeli hood that an owner will opt-out of the housing subsidy program and convert the property to market ra te housing. The condition also affects the likelihood of loss of the affo rdable housing due to phys ical deterioration and

PAGE 19

19 mortgage default. As explained by Wallace et al. (1993, 2-25), net cash flow is a key indicator of a propertys viability. Achtenberg (2002, 43) also stated that cash flow the bottom line is an important indicator of the overall health of the property. Simulation modeling was used, because many of the input variables to compose net cash flow statements were uncertain due to a lack of public information about the current and future financial and physical condition of subsidized properties. Missing property-level data included actual operating expenses, vacancy rates and capital needs. Simula tion modeling allowed for the estimation of uncertain values according to probability distributions. A database was composed of multifamily prope rties with project-based rental assistance contracts. From this database, development-level profor mas were created, using actual as well as simulated data for rental income, operating expe nses and debt service. Descriptive analysis, significance tests and regression an alysis were performed to analyze the simulated net operating income data and to characterize the properties that were identified at heightened risk of loss to the affordable housing stock. Programmatic and Geographic Scope The programmatic scope of the research wa s on properties that re ceived project-based rental assistance under the HUD Sec tion 8 program. The reason for this focus was that rental assistance is considered a deep subsidy that targ ets the lowest income families. Households that receive project-based rental assistance generally pay no more than 30% of the gross household income on rent plus utilities. The rental assi stance subsidy covers the difference between the collected rent and the ac tual cost of housing or the market re nt, within limits established by the federal government (Rice 2009). The two other ma jor federal rental assistance programs are Housing Choice Vouchers and Public Housing, both also providing deep subsidies. The Center on Budget and Policy Prioritie s reported that there is strong evidence that rental assistance:

PAGE 20

20 Alleviates poverty; frees up financial resources that poor families can use for other basic needs; and improves housing stability and reduces the risk of homelessness (Rice and Sard 2009). However, only one in four low-income house holds eligible for federal housing assistance receives it because of funding limitations (Rice and Sard 2009, 4). Many other housing programs at the federal, state and local levels on ly impose income restrictions and in some cases maximum rent levels. These so-called shallow subsidies cannot prevent cost burden among the lowest income households. For example, the Lo w Income Housing Tax Credit program generally targets households at 50 to 60% of area median income. While households below these income limits are eligible to live in a tax credit propert y, they would spend more than 30% of income on housing (unless they have a Housing Choice Voucher). Without deep subsidies families in the lowest income brackets will experience cost burden. It is therefore essential to preserve the properties that receive project -based rental assistance. The geographic focus was Miami-Dade a nd Duval County, because the number of properties and units with project-ba sed rental assistance in these c ounties trumps that of all other individual counties in Florida (HUD 2008a). More th an one third of properties and units covered under HUD rental assistance contra cts in Florida were located in Miami-Dade or Duval. Preventing the loss of properties with rental as sistance was considered important, because both counties housed renters that made less than 60% of the area median income and were paying more than 40% of income on housing; this grou p made up more than 30% of renter households in Miami-Dade and more than 21% in Duval (Shimberg Center for Housing Studies 2007). Organization of the Study The study first addresses the rationale for pr eservation in Chapter 1. This chapter includes a demand and supply analysis of affordable housing in Florida, and defines affordability and cost burden. Chapter 2 describes the major federa l housing programs that contributed to the

PAGE 21

21 construction and substantial rehabilitation of rental housing in the 1960s through 1980s. This chapter also outlines the first federal preservation legislati on, as well as the programmatic reasons for the loss of subsidized properties. Ch apter 3 addresses the definition of risk in the context of preservation, and review s risk assessment methods that ha ve utilized property data to estimate at-risk multifamily properties or to dete rmine key indicators of risk. Chapter 4 explains the research methodology of the net cash flow approach and si mulation modeling. The analysis results are presented in Chapter 5, followed by conclusions and recommendations in Chapter 6.

PAGE 22

22 CHAPTER 2 THE RATIONALE FOR PRESERVATI ON OF AFF ORDABLE HOUSING Demographic Shifts and Implications for Housing Demand The state of Florida has experienced a stea dy increase in population th at is expected to keep pace in the decades ahead. The incline started with the a doption of air conditioning in homes in the 1950s. Since that time, growth wa s impacted by a large natural increase in population during the baby boom in the 1950s and 1960s and the child-bearing years of the baby boomers in the 1970s and 1980s. But the most significant reason for Fl oridas growth in population has been migration from other states in the country, especially the northeastern states, and increasingly from foreign countries (Smith 2005). From a population of nearly five million in 1960, Florida increased to almost 16 million in 2000 and is expected to surpass 35 million residents by 2060, as graphed by Figure 2-1 (U.S Census Bureau 1990; Zwick and Carr 2006). The U.S. Census Bureau (2005) projected that Fl orida together with California and Texas will account for almost one-half of the growth in U.S. population between 2000 and 2030 with Florida becoming the third most pop ulous state in the country. Smith (2005, 6) predicted that, if current projections prove to be accurate, net migration will account for all of Floridas population growth within 20-25 years. The growth and composition of Floridas popul ation have directly impacted the housing market. The groups that especially shaped the demand for affordab le housing included the elderly, the workforce and forei gn immigrants. Each of these groups is briefly discussed next. Florida is home to many retirees who ar e settled in the state permanently, and to snowbirds who temporarily spend the winter m onths. Retirees and snowbirds typically include households at ages 55 and older with accumula ted wealth and home equity, although recent changes in the capital markets may have eroded their spending power. For years, these

PAGE 23

23 households were attracted to Flor ida for its warm climate and relatively affordable real estate, especially compared to the northeastern part of the United States. Their purchasing power contributed to a rise in the price and supply of homes in Florida until market conditions started to change in 2006. Florida is also home to elderly residents who live on limited fixed incomes and are facing difficulty in finding and keeping housing that is affordable. Sixty percent of renter households that have a head of household at age 65 or older were estimated to be cost burdened (U.S. Census Bureau 2007b). The population group of households at age 65 and older was projected to almost double by 2025 (Shimberg Center for Housing Studi es 2008a). The predicted growth of the elderly population as well as th eir relatively small household size will continue to spur demand for affordable housing units. Floridas population growth also stemmed fr om an increase in its youth and working age population (White et al. 2005). Smith (2005) reported that the majority of people that migrated to Florida were younger (under 35) rather than ol der (65 and over). They were in pursuit of employment opportunities, which they found in c onstruction, leisure, hospitality and retail (Nissen and Zhang 2006). These sectors offe red low-paying jobs, which made housing affordability an issue for these young people when Floridas housing market was booming. As illustrated by Figure 2-2, Floridas workforce experienced stagnating incomes and increasing home prices until 2006. Even essential services pe rsonnel such as teachers, nurses, firefighters and policemen were confronted with an inability to afford to live in the community that they served. Especially in southern Florida, compan ies were finding it increasi ngly difficult to attract and retain employees (Rawls 2006).

PAGE 24

24 While home prices have now stabilized or dropped in Floridas submarkets, this is not resolving the housing affordability challenge. Fam ilies are under financial pressure as they are faced with unemployment and foreclosure. The gap between house prices and flat wages continues to be large enough to make ho meownership unattainable for many households, especially those in lower income brackets, on fixed income or with damaged credit scores. Immigrants from other countries now repr esent one quarter of all people settling in Florida. In particular, Florida witnessed a rapid influx of Hispanics who include people from countries in Latin America and the Caribbean su ch as Cuba, Puerto Rico and Mexico. By the year 2000, Florida counted nearly 2.7 million Hisp anics who were mostly concentrated in the southeastern part of the stat e (Smith 2005). Miami-Dade Count y had the largest cluster of Hispanics or Latinos at 57% of the countys to tal population. This was pr ojected to increase to 70% by 2030 (Center for Urban and Environmenta l Solutions 2006). Foreign immigration has had an impact on housing demand. Foreign-born hous eholds have dominated the rental market, although they have also been fi nding opportunities to transiti on into homeownership. Between 1990 and 2000, foreign-born renter hous eholds made up over 60% of the increase in the total number of renter households in Florida. Fore ign-born homeowners comprised just under 28% of the total growth in homeowner households in Fl orida for that same period (Myers and Yang Liu 2005). Strong net migration continued in Flor ida during 2000-2007. The state experienced the highest domestic in-migration and the fourth largest number of international immigrants compared to the rest of the country (J oint Center for Hous ing Studies 2008). Housing Supply Characteristics and Trends Floridas housing stock is made up of singl e-family homes (as the predominant type), condominiums, multi-family rental developments and manufactured homes. The inventory is not

PAGE 25

25 uniformly distributed throughout the state. There is a real diversity between urban and rural areas as well as coastal and non-co astal regions. The counties th at are most urbanized and predominantly coastal are home to almost 94% of Floridas singly-family dwellings and more than 98% of its condominiums (White et al. 2009). Multi-family housing is also concentrated in urbanized areas. Floridas population of rural and interior counties is mostly housed in singlefamily homes, although the mobile home is also most common to the rural counties. The state of Florida experi enced a noticeable expansion and shift in housing supply since the late 1990s. It witnes sed a substantial surge in new constr uction, particularly in coastal and metropolitan areas and in southern Florida. From 2001 to 2005, the number of single family homes authorized by building permits increased by 73.1%, compared to an increase of 22.9% between 1996 and 2000, as graphed in Figure 2-3 (HUD 1996-2008). Florida also underwent a wave of conversions of privat ely-owned, subsidized and non-subs idized rental apartments to condominiums. Condominiums offered ownership oppor tunities for first-time homebuyers who were encouraged by relatively low interest ra tes and favorable financ ing terms. Among the purchasers of new homes and condos were also investors; about a quarter or more of all home purchase loans were made to investors in 2005 and 2006 (HUD 2009). In addition to the surge in new constructi on and the condo conversion craze, the volume of sales of existing single family homes almost doubled between 1996 and 2005, reaching a record level of close to 2 50,000 sales, as shown in Figure 2-2 (Florida Association of REALTORS 1996-2008). Floridas homeownership ra te reached 72.4% by 2005, compared to the national average of 68.9% (U.S. Census Bureau 2008). The year 2006 marked a shift in Florid as housing market conditions: A drop in home sales and prices (Figure 2-2), a drop in the num ber of units authorized by building permits

PAGE 26

26 (Figure 2-3), a rising unsold i nventory of single-family homes and new condominiums, and a reconversion of condominiums back to rental properties (Joint Center for Housing Studies 2008). The situation worsened during 2007 and 2008, as signaled by the steep increase in home mortgage foreclosures. Affordability and Cost Burden Defined From the first half of the nineteenth century until the 1970s, the American concern with housing was focused on the problem of poor physi cal conditions, overcrowding and inadequate supply of units (Vidal 1997) By the 1980s, government housing policy had shifted from a supply-side focus to a demand-side emphasis when housing affordability had become a much greater concern than substandard conditions and overcrowding (Schwartz 2006; Vidal 1997). This shift was evidenced by the Presidents Commissions declaration in 1982: Today the largest problem is not the quality of housing in which most people live but its affordability (Listokin 1991, 166). From a simplistic perspective, housing afford ability means that a household can afford to pay for the cost of housing from their income, wh ile still financially able to meet other basic needs such as food, clothing and health care. Ec onomists generally take another view. They believe that if a household is paying a given amount for housing, it implies that the household can afford to do so (Green and Malpezzi 2003). From a government policy and programmatic perspective, housing affordability usually means that a household pays no more than 30% of its annual gross income for housing. For renters, th e cost of housing is c onsidered the rental payments and utilities (excluding telephone); fo r homeowners, it includes principal, interest, property taxes and insurance. When a household sp ends more than 30% of its gross income on housing, it is regarded cost burdened; when spending more than 50%, a household is deemed severely cost burdened.

PAGE 27

27 Prior to 1981, the 30% standard was 25% of income, as established by the Brook Amendment in 1969 to limit the rent and cost bur den of tenants living in federally assisted housing (Eggers and Moumen 2008). Authors Eggers and Moumen (2008) as well as Belsky and Bogardus Drew (2006) concluded that the reason s for initially setting the standard at 25% rather than another percentage are unclear. They did explain that the standard was increased to 30% in order to reduce federal outlays for hous ing. Green and Malpezzi (2003, 137) argued that the 30% or any fixed benchmark is always deba table. According to Schwartz (2006, 23), the current 30% and 50% cost burden thresholds have no intrinsic meaning until the 1980s the maximum acceptable cost burden was typically set at 25%; nevertheless, they are widely used. Many housing programs impose income restrictio ns rather than rent restrictions. Under income restrictions a household wi ll only be eligible to reside in the housing development if its income is below a specified percen tage of the area median income. Affordability in Florida Floridas spur in new construction, conversi on and sales was driven by a combination of factors: A decrease in interest rates, an increase in mortga ge products with more flexible and favorable terms, a growth in population and numbe r of households through migration from other states and foreign countries, and the emergen ce of investors and speculators. The demand and supply dynamics drove up the cost of land, the pri ce of new and existing homes, property values and property taxes. While the cost of housing increased substantially, median income did not see much of a rise and the affordability gap wide ned, as was illustrated by Figure 2-2. Now that home prices have fallen, affordability remains an issue as a result of the economic downturn. Cost burden data presents insight into the issue of affordability in Florida. The proportion of total households that wa s cost burdened was estimated at 40.5% or more than 2.8 million households in 2007. Cost bur den data by tenure revealed that almost 52%

PAGE 28

28 of renters paid more than 30% of household in come on rent, which totaled more than 1 million households. This compared to almost 36% of homeowners that were cost burdened, which concerned more than 1.7 million households. Severe cost burden affected roughly a quarter of all renter households; more than half a million house holds were spending at least half their gross income on housing (U.S. Census Bureau 2007c). Cost burdened households had lower levels of income and were thereby more challenged to meet other basic needs. About 80% of renter households with an annual income of less than $20,000 were cost burdened. Data confirmed that co st burden diminishes as income increases. Less than 4% of renter households that enjoy an annual income of at least $75,000 experienced cost burden (U.S. Census Bureau 2007d). Cost burden affected people of all ages, but most noticeable was the proportional impact it had on the youngest and oldest generation of rent ers. Over 58% of renter households in the 15 to 24 age group were cost burdened (U.S. Census Bureau 2007b). This could be explained by their relatively low household in come that results from their entry positions and predominant one-person household size. Cost burden also affected more than 60% of the renter households in the 65 and older cohort (U.S. Census Bur eau 2007b). Many of these elderly live on fixed incomes and have a small household size. The number of households that paid more than 40% of income for rent and that had an income at or below 60% of area median were es timated by county in the latest Rental Market Study that was conducted by the Shimberg Center for Housing Studies (2007). The data revealed that the households that met th ese income criteria were not ge ographically distributed evenly throughout Florida. Rather, more than 60% of th e households are located in the states seven largest counties: Broward, Duval, Hillsborough, Miami-Dade, Orange, Palm Beach and Pinellas.

PAGE 29

29 These are urbanized counties that are home to many lower income households. These are also the counties that experienced a significant increase in the cost of housing. Miami-Dade and Broward housed 30% of Floridas cost burdened househol ds, as defined in the Rental Market Study. The number of cost burdened households doe s not equal demand for affordable housing units. But analysis of cost burden figures doe s provide insight into the severity of the affordability issue, as well as the characteristics of those households most impacted. Rationale for Preservation The rationale for preservation of subsidi zed housing is four-fold. First, preservation secures housing options for the lowest income households. More than 76% of the privatelyowned units assisted by HUD were estimated to serve extremely lo w-income families that have a gross income of less than 30% of the area me dian income (HUD 2006). These households are at a relatively high risk of displacem ent, since the HUD properties are at a relatively high risk of loss. Second, preservation protects assets that we re built with public funds. Preservation and the prevention of subsidized mortgage default and foreclosure also minimize loss claims on the Federal Housing Administration insuran ce fund (National Housing Trust 2006a). Third, preservation offers several advantages compared to new construction of affordable housing. While cost estimates vary, a general assump tion is that rehabilitation of affordable units costs 40% less than new construction (Khadduri and Wilkins 2006). Also, preservation is not hindered by NIMBY-ism and regulat ory barriers such as zoning. Fourth, preservation can c ontribute to community revita lization. Many older federallyassisted properties are de teriorating and require capital improvements (Wilkins 2002). Dilapidated properties negatively im pact the lives of residents and property values in the area. In a distressed neighborhood where co nversion to market-rate housi ng is not a feasible option,

PAGE 30

30 preservation can contribute to revitalization when mortgages are restructured and funds are made available for rehabilitation. Obstacles to Preservation While it is important to preser ve the stock of assisted rent al housing, preservation at any cost is not feasible. One obstacle to preservati on is that a high infusion of capital is required, since many subsidized housing properties have substantial deferred capita l needs or have to undergo major reconfiguration of the internal la yout (Wilkins 2002). Rehabilitation can be extra costly and risky due to unknown conditions inherent to an older building structure. Unforeseen costs and repair needs can negatively impact a projects bottom line a nd construction schedule. Due to limited resources, it is not feasible to preserve every property. Depending upon their overall quality, their locational desi rability to tenants relative to other housing options, and their current annual subsidy costs (if assisted), it may be more cost eff ective to retire some of these especially high backlog properties from the stock of HUD-insured housing rather than to repair them (Wallace et al. 1993, 3-32, 3-33). Preservation cannot be achieved without th e cooperation of the property owner. An owner could decide to terminate affordability an d participation in a subsidy program, even if existing or new subsidies are available to continue to operate the property with use restrictions. An owner may be driven by fina ncial motivations to convert a property to market-rate housing, or may seek relief from the administrative burd en related to complex reporting requirements, especially if a property has multiple funding layers (Governors Task Force for Housing Preservation 2004). An owner of an assisted proper ty could also decide to not make substantial capital improvements. For some distressed properties, lack of an owner willing to cooperate may make it impossible to undertake an effec tive program of physical improvements. HUDs

PAGE 31

31 ability to assist properties depends upon the pr esence of a cooperative owner (Wallace et al. 1993, 3-33). 5,000,000 10,000,000 15,000,000 20,000,000 25,000,000 30,000,000 35,000,000 40,000,000 19001910192019301940195019601970198019902000201020202030204020502060 Figure 2-1. Population of the State of Florida, 1900-2060. Source: Zwick and Carr (2006); U.S. Census Bureau (1990). $0 $50,000 $100,000 $150,000 $200,000 $250,000 $300,000 1996199719981999200020012002200320042005200620072008 0 50,000 100,000 150,000 200,000 250,000 300,000 Median Family Income ($) Median Sales Price Single-Family Existing Homes ($) Number of Sales of Single-Family Existing Homes Figure 2-2. Median Family Income, Median Sales Price and Number of Sales of Existing SingleFamily Homes in Florida, 1996-2008. Source: Florida Association of REALTORS (1996-2008); U.S. Department of Housi ng and Urban Development (1996-2008).

PAGE 32

32 0 50,000 100,000 150,000 200,000 250,000 300,000 1996199719981999200020012002200320042005200620072008Number of Units Single Family Multifamily Figure 2-3. Units Authorized by Building Pe rmits in Florida, 1996-2008. Source: U.S. Department of Housing and Urban Development (1996-2008).

PAGE 33

33 CHAPTER 3 FEDERAL GOVERNMENT PROGRAMS AND IMPLICATIONS FOR OWNE R CHOICES First Government Involvement in Housing: Mortgage Insurance and Public Hou sing Public concern with housing aros e in the first half of the nineteenth century in reaction to poor housing conditions. Industria lization and employment opport unities attracted impoverished rural residents and foreign immigrants to Ameri can cities. Urbanization went hand-in-hand with the rise of substandard living conditions charac terized by overcrowding, l ack of sanitation and shoddy construction, which were believed to harm the poor (von Hoffman 1998). Public concern led to the housing reform movement that focused on the physical conditi on of structures, housing codes and legislation at local and stat e levels (Friedman 1985; von Hoffman 1998). The U.S. government first became direct ly involved in housing during the Great Depression of the 1930s. The involvement was a reac tion to a decline in housing starts and real estate values; high unemployment, especially in the building trades; large need for housing among poor, unemployed and evicted households; a nd bank failures as a result of increased withdrawals of deposits and mortgage defaults (Mitchell 1985). The pos ition of the government was that these challenging conditions could be tackled by stimulating construction and thereby stimulating related industries, which would create employment, enlarge the housing stock and improve incomes. The government developed tw o major strategies to achieve this: Home mortgage insurance and public housing. The Housing Act of 1934 was passed to establish the Federal Housing Administration (FHA). FHA was created as an independent agency that introduced mortgage insurance for private loans on single-family homes as a 100% guarantee against default. It was intended to encourage homeownership, fuel the construction industry and stabilize the mortgage market (Vandell 1995). It reduced lending risk for fina ncial institutions and allowed them to replace

PAGE 34

34 short-term, high down payment balloon notes by long-term, low down payment, fully amortized and level-payment mortgages (Hays 1995). The FHA-program was financed by insurance premiums that were deposited in an in surance reserve fund. The program did bring homeownership within reach of wo rking class households, but it failed to target the poor that had no stable income. Public housing was the second element of dire ct government response, as set out in the Housing Act of 1937. The objective of the public housing program was to generate employment opportunities in the construction industry, to provide housing for families who were temporarily unemployed, and to clear slums and blighted areas. Local public housing authorities (PHAs) were created under state legislati ons to build and operate housing developments. At the outset of the program, the capital cost of land and building construction was paid for by the federal government through annual installment repayments of bonds that PHAs issued to finance the realization of projects. Operating costs were expe cted to be financed in full by the PHAs from rent collections (Quercia and Galster 1997). Bu t development was hindered by federal opposition against the public housing concept and by local opposition caused by negative perception of public housing tenants and resistan ce against poor curb appeal of the structures. By the 1960s, public housing authorities also faced financial challenges in supporting operating and maintenance costs. They were criticized for poor management, but according to Hays (1995) the financial distress was really caused by other fact ors. First, inflation was increasing expenses. Second, structures were aging a nd started to require repairs a nd capital improvements, but no cash reserves were available. Third, the compos ition of the public housin g tenants changed from the temporarily unemployed to the long-term poor resulting in declining tenant incomes and shortfall in rents to cover operating expens es. The government responded with the Brooke

PAGE 35

35 Amendments in 1969, which limited rents to 25% of household income and provided for federal subsidies to bridge the gap between re venues and operating costs (Schnare 1991). Between the 1930s and mid-1970s, housing bu ilt by the government public housing was the approach to address the need for affordab le rental housing. The fede ral attitude to public housing shifted in a fundamental way when Kennedy took office in 1961. Kennedy was in support of housing the poor and wanted to stimulate the economy through construction. But he was concerned with the always troubled public housing program (Orlebek e 2000). He started to explore alternate approaches to housing production that would ach ieve two goals: To develop a housing program that would target households w ith incomes too high to be eligible for public housing but too low to acquire housing with F HA support; and to develop a housing program on the principle of public-priva te cooperation (Hays 1995). Interest Rate Subsidy Programs: Sect ion 221(d)(3) BMIR and Section 236 Section 221(d)(3) Below Market Interest Rate The Kennedy Administration enacted the Sect ion 221(d)(3) Below Mark et Interest Rate (BMIR) program under the National Housing Act in 1961. It enabled private lenders to originate mortgages on rental housing prope rties at a 3% interest rate by allowing them to sell these mortgages to Fannie Mae at market rate. The objective of the program was to promote construction of affordable housing by offe ring non-profit and for-profit developers the opportunity to obtain subsidized loans at an inte rest rate below market. The loans were also insured by the Federal Housing Administration to lower the risk to lenders. The amortization term of the mortgages issued under this program was 40 years. In general, non-profit owners had no option to prepay the mortgage. They had to ma intain a 40-year use restriction until maturity. But most for-profit owners could prepay after 20 years and terminate the use restriction (Pedone 1991). The rate of return to propert y owners was restricted to 6% of their original equity, which

PAGE 36

36 was generally 10% of the initial cost of the project. Public en tities and non-profit developers were able to borrow up to 100% of the mortgage and were thereby not required to provide down payments. The first mortgage under this pr ogram closed in 1962 (U.S. Congress 1987). Tax benefits were also provided to property owners as additional incentives to construct affordable rental properties. Owners were allo wed to deduct mortgage interest payments and calculate accelerated deprecia tion, although a limited dividend re striction on their cash flow distribution was imposed (Achtenberg, 2002). The program was aimed at families with incomes too high for public housing, but not sufficient to afford housing in the private rental market or the FHAsupported owner-occupied market. To determine eligibility, income limits were generally established at 95% of the area median income (Millennial Housing Commission 2002). The rent was usually set between the upper income limit for the public housing rent le vel and the area median income (Hays 1995). But without layering a rental assistance program or other a dditional funding on the mortgage subsidy, affordability of the units proved difficult to achieve (Affordable Housing Study Commission 2005). Section 221(d)(3) BMIR required large capital outlays up front, since Fannie Mae would purchase each entire mortgage and absorb the di fference between 3% and the market rate. Due to this budgetary impact, the progr am was eliminated and replac ed by Section 236 in 1968 (Hays 1995). According to HUD data for Florida, nineteen properties with more than 2,100 were built under the Section 221(d)(3) BMIR program duri ng the 1960s and early 1970s (HUD 2009). Only one of these properties still has an original 221(d)(3) BMIR mortgage at present. The development contains 220 units, receives no othe r project-based funding and has a maturity date

PAGE 37

37 of September 2012 (HUD 2009). The original Se ction 221(d)(3) BMIR mortgages on the 18 properties were terminated mostly because of prepayment or assignment1. Section 236 The Section 236 program became law under th e Housing and Urban Development Act of 1968, which was enacted by the Johnson Administ ration. The government provided lenders with a monthly interest reduction payment (IRP) subsidy, which reduced the interest rate on loans to developers of rental housing from market rate to 1%. The amor tization term of the mortgages issued under this program was the same as that for Section 221(d)(3) BMIR, 40 years with eligibility for most for-profit owners to prepay after 20 years (Clay and Wallace 1990). In addition to the mortgage subsidy, the government provided special tax adva ntages to the property owners such as accelerated depreciation, as well as FHA mortgage insurance to reduce lenders risk. The rate of return to owners continued to be capped at 6% of original equity (Achtenberg 2002). Households were eligible if their income did not exceed 80% of the area median. Tenants were to pay the higher of 30% of their income or basic rent that was set at the amount of operating expenses plus debt serv ice at 1% (Achtenberg 2002). Th e rent could not exceed Fair Market Rent2 (FMR) (HUD 2007). In 1973, new funding under the Section 236 program came to a halt when Nixons government placed a moratorium on all housing progr ams in reaction to rising costs and subsidy commitments, cash flow problems of public housi ng developments, program scandals and public 1 Assignment means that HUD has taken over the loan after the property owner defaults. When this happens, HUD will attempt to work out the situation with the owner to get him back on track of mortgage payments. If this is not successful, the next st ep is foreclosure. 2 Fair Market Rent is HUDs estimate of the actual ma rket rent for a comparable apartment in the conventional marketplace. Every year, HUD develo ps and publishes FMRs for ever y MSA and every apartment type (Recapitalization Adviso rs, Inc. 2009, 1).

PAGE 38

38 criticism (Mitchell 1985). Sec tion 236 was abandoned and a ne w program (Section 8) was introduced to stimulate constructi on and rehabilitation of affordab le rental housing by the private sector. HUD data reported that 132 properties w ith more than 19,000 units were funded under Section 236 between 1969 and the early 1970s in Florida. Today, 49 properties of these properties with more than 8,000 units still have an original Section 236 mortgage. The majority will reach maturity by 2014; all mortga ges are due to mature by 2017 (HUD 2009). Rental Assistance Programs: Section 8 Under the Housing and Community Devel opment Act of 1974, the federal government created a new program to stimulate constructi on of affordable rental housing by the private sector, called Section 8. It encompassed several rental assistance initiatives, both demand-side subsidies and supply-side subsidies. Demand-side subsidies were provided unde r the Section 8 Existing Housing Program, which was considered the first national vouche r program (Schwartz 2006). It offered tenantbased rental assistance for hous eholds to find housing in the private market. This program followed an experimental rental allowance program called Section 23 that offered the first form of tenant-based assistance and that was crea ted by Congress in 1965 (Olsen 2000; Grigsby and Bourassa 2004). The Section 8 tenant-based prog ram supplied rental certificates to households with an income at or below 80% of the area me dian. It limited the househol ds rent and utilities to 25% of gross income (increased to 30% in 1981). The property owner received a subsidy to cover the gap between collected rent and Fair Market Rent. Vouchers were administered by public housing authorities that were also in char ge of waiting lists and inspections of units. A variant on the Section 8 certi ficate program was introduced in 1983, known as the Freestanding Voucher program (Schwartz 2006). This program also limited the households rent to 30% of

PAGE 39

39 income, although a household was allowed to spend more or less. The Section 8 certificate and voucher programs were merged into the H ousing Choice Voucher program in 1998, which required that at least 75% of all vouchers issu ed each year are distributed to extremely lowincome households that earn below 30% of the area median income (Schwartz 2006). The program has become the nations leading source of housing assistance for low-income elderly, people with disabilities, and families with chil dren (Rice, Sard and Coven 2007, 1). Vouchers currently serve about two million low-income families nationwide (Rice 2009). However, the number of households in need of financial s upport far exceeds the assistance that is available; more than eight million renter households were se verely cost burdened in the United States in 2007 (U.S. Bureau of the Census 2007). More th an 94,000 vouchers were administered by public housing authorities in Florid a in 2007 (Shimberg Center for Housing Studies 2009). Supply-side subsidies were introduced unde r various Section 8 project-based rental assistance programs. These programs provided a direct subsidy of rents through a contract between the property owner and the local public housing authority as the contract administrator. The subsidy covered the difference between the re nt and 25% of household income (increased to 30% in 1981). Eligibility was initially set at 80% of the area median income, but later 40% of tenants admitted annually had to be classified as extremely low income (Millennial Housing Commission 2002). The major Section 8 projectbased rental assistance programs were the Section 8 Loan Management Set Aside (LMSA) program, the Section 8 New Construction (NC) program and the Section 8 Substantia l Rehabilitation (SR) program. Properties built under Section 221(d)(3) BMIR and Section 236 were experiencing difficulty in meeting their mortgage obligations du e to shortcomings in th e rents received from tenants. The LMSA program was developed to supplement the rents on those properties (Bratt

PAGE 40

40 1989). This would protect very-low income families and save the Federal Housing Administration the cost of loan defaults (Kochera, Redfoot, and Citro 2001). Project rents were budget-based, which means based on operating cost s and a financial return on investment to owners. The initial term of the LMSA contracts was fifteen years with the option to renew or opt-out upon contract expiration (Recapitalizati on Advisors, Inc. 2009). In Florida, over 8,000 units continue to receive project-based rental-assistance under the LMSA program (HUD 2008), which is for the renewal of existing contract s; the program no longe r provides new funding (Millennial Housing Commission 2002). The Section 8 New Construction and Substant ial Rehabilitation programs provided rental assistance to newly constructe d or substantially rehabilita ted privately-owned rental developments that were funded by any FHA-insu red mortgage (HUD 1999). The contract rent was based on Fair Market Rent and subject to an annual adjustment factor The initial contract term for assistance was set at 20 to 40 years with the option to renew or op t-out at the end of the contract term. It was up to the developers to decide on the type of financing such as a conventional loan or a below-market-rate mort gage. The Section 8 NC and SR programs were designed as an alternative to the interest ra te subsidy programs (Sect ion 221(d)(3) BMIR and Section 236) in order to provide more flexibility to owners and to reach lower-income families (Schwartz 2006). The programs offered an attr active tax benefit; owne rs could reduce their taxable income by using accelerated depreciation allowances (Schwartz 2006). The Section 8 NC and SR programs were repealed in 1983, but the federal government continues to fund the renewal of existing contracts. Current contracts provide rental assistance to almost 22,000 units in Florida (HUD 2009).

PAGE 41

41 In Florida, the LMSA and NC/SR programs curre ntly assist almost 80% of units that are covered under project-based rental assistance contracts. Roughly 14% of assisted units receive rental assistance in the form of Project Re ntal Assistance Contracts (PRACs) (HUD 2009). PRACs subsidize operating expenses of deve lopments that are funded under the HUD Section 202 Capital Advance program for the elderly or the HUD Section 811 Capital Advance program for persons with disabilities (NLIHC 2009). The remaining 6% of project-based rental assistance units are funded by programs that are vari ants on the LMSA and NC/SR programs. Due to a lack of information as well as da ta discrepancies, it is unclear how many units with HUD project-based rental assistance have been lost as a result of opt-out by the owner or non-renewal by HUDs decision. Housing Preservation Act By the 1980s, federal housing policy had shif ted from a supply-side production approach to a demand-side subsidy system. After a substantia l increase in the supply of affordable rental housing units during the previous two decades, new construction slowed down steadily under President Reagans antiproduction, voucher-o nly housing policy (Orlebeke 2000, 509). Both homeowners and renters were st arting to experience difficulty in finding affordable housing. The problem was most serious for renters (Hays 1995). The 1980s also marked the 20th anniversary of the subsidi zed mortgages issued under the earliest federal programs, which placed assisted rental housing at risk of losing affordability restrictions if an owner would d ecide to prepay the mortgage. Th e extent of prepayments and the impact on residents were uncertain and had not been explored. In 1987, the Congressional Budget Office wrote, As the twentieth anniversar y dates on the first of [the 221(d)(3) and 236] projects have approached, c oncern is being expressed about whether the government should respond to the loss of a potentially significant nu mber of these units from the assisted housing

PAGE 42

42 stock and, if so, what form the response should take (U.S. Congress 1987, 1). In the same year, the federal government intr oduced the Emergency Low Income Housing Preservation Act (ELIHPA) to prevent owners from converting to market rents. The Act placed a two-year moratorium on prepayments, while Congress de veloped a permanent solution to protect affordable units. Various reports were produ ced by Congress and by gove rnment-initiated and independent task forces on the scope of the subs idized mortgage prepayment issue and the optout and expiry of Section 8 Project-based rental assistance contracts: The Potential Loss of Assisted Housing Units as Cert ain Mortgage-Interest Subsidy Programs Mature (1987) by the Congressional Budget Office of the U.S. Congr ess; Preventing the Disappearance of Low Income Housing (1988) by the National Low Income Housing Preserva tion Commission; The Preservation of Low and Moderate Income Housi ng in the United States of America (1988) by the National Housing Preservation Task Force; and A Decent Place to Live (1988) by the National Housing Task Force. One of the reports predicted that 81% of the older HUD-assisted inventory would likely be impacted by prepayments or defaults from 1988 to 2002, assuming the expiry of Section 8 subsidies (Pedone 1991). A nother overall conclusion wa s that the affected households were relatively poor an d the threat of displacement of current tenants was substantial. In 1990, ELIHPA was replaced by the Low-Inco me Housing Preservation and Resident Homeownership Act (LIHPRHA) that imposed per manent prepayment restrictions. This Act was aimed at preserving privatel y owned subsidized properties as low-income housing for their remaining useful life (50 years). This new regul ation placed several properties in challenging cash flow positions by prohibiting the increase to market rent. Owners were provided incentives to refinance their propertie s under the program or to sell to a qualified purchaser, an entity with a mandate to keep the units affordable. Among th e incentives were, insur ed or direct capital

PAGE 43

43 improvement financing, an equity takeout loan, an 8% return on preservation equity, access to reserves, increased Section 8 and non-Section 8 rent s, and insured acquisition loans and grants to qualified purchaser (Koebel and Baily 1992, 997) Conversion to market rate units or nonresidential use was only allowed when public funding for preservation was inadequate or when a qualified purchaser could not be found (Koebel and Bailey 1992). In 1996, the Clinton Administration adopted the Housing Opportunity Program Extension Act that restored the prepayment rights. In th e subsequent year, all federal preservation funding under LIHPRHA was terminated (Peiser 1999). Elig ible property owners were now allowed to prepay the mortgage at the 20th anniversary of the loan and prio r to the 40 year maturity. This shift in policy followed the change in the fe deral focus from preservation of housing to protection of tenants against di splacement (Achtenberg 2002). In recent years, the research of the pr eservation issue as well as the funding of preservation projects have become a highe r priority nationwide among state and local governments, foundations such as the MacArthur Foundation, research institutes such as the Center for Housing Policy and the Shimberg Center for Housing Studies, and advocacy bodies such as the National Low In come Housing Coalition. Programmatic Reasons for Loss of Affordability and Owner Choices A property is either lost as a result of deterioration and default, or conversion to marketrate housing, as discussed next. The risk of loss of affordable units as a resu lt of deterioration stem s from the age of the structures and the need for capital improvements. As buildings have aged, replacement of such items as roofs, plumbing fixtures and heatin g-cooling equipment has become necessary. HUD properties built during the 1960s through 1980s are struggling with physical deterioration and deferred capital improvements. However, many s ubsidized multifamily properties have none or

PAGE 44

44 limited capital reserves (Khadduri and Wilkin s 2007; Wilkins 2002). As explained by Recapitalization Advisors (2008, 1) Financial difficulties are in evitably tied up with physical deterioration: it is rare to find one without the other. An owner may default on a mortgage if it is no longer able to meet its mortgage obliga tion. Dilapidation and lo an default lead to foreclosure if the government agency such as HUD cannot restore financial viability of a property in cooperation with the owner (Pedone 1991). Foreclosure can result in displacement of tenants if a new property owner can not be found. The risk that a pr operty is not maintained is highest in financially and so cially distressed neighborhoods. Conversion to market-rate rentals or condomin iums can be a financially attractive option in a strong local housing market. An owner has th is option when a subsidized property reaches a discontinuity event, a term phrased by Recapitalization Advisors, Inc. (2002, 6). A discontinuity event is a point in time at which th e terms of the funding allow the owner to make a choice about the future of a propertys a ffordability. Discontinuity events include: Mortgage prepayment eligibility: A property owner with a mortgage under Section 221(d)(3) BMIR or Section 236 may be eligible to prepay the lo an after 20 years from the origination of the loan and any time prior to maturity. Mortgage maturity: When the loan obligations are met at the end of the term, all affordability restrictions are lifted if no othe r funding programs or agreements are in place to keep the development affordable for a longer term. Expiration of rental assistan ce: Each rental assistance c ontract has a limited timeframe. Upon expiration of the contract, an owner has the choice to extend the contract and continue to receive rental assistance, or to opt-out. Expiration of use restriction: Some funding programs place a long-term use restriction on the property to keep units affordable. It is possible that such an affordability period extents beyond a loan maturity date.

PAGE 45

45 Owners of subsidized properties have contr actual rights and obliga tions that allow or force them to make a choice about the future of a property at the time of a discontinuity event such as eligibility of prepayment or mo rtgage maturity. Among the choices are to: Retain the property and maintain affordabi lity by deferring prepayment of a mortgage, refinancing upon prepayment or mortgage ma turity, or renewing the rental assistance contract. Retain the property, terminate affordabilit y, and convert to mark et-rate rentals or condominiums by way of rehabilitation or by demolition and redevelopment. Sell the property to a preserving entity such as a non-profit or ganization that will maintain affordability and continue to serve the low-income tenants. Sell the property to a market buyer who will te rminate affordability and convert to marketrate rentals or condominiums. The choice that the owner will ultimately ma ke is impacted by a wide variety of interconnected motivations related to financial and market considerations, physical state of the property, availability of funding and relationship with the funding agency, and type of ownership. Financial considerations may motivate an owner to terminate affordability and are impacted by market conditions. Cash flow is critical to the financial viabili ty of a development. Marginal cash flow may be a motivation for a prope rty owner to convert to market rate rentals in order to substantially increase re nt revenues and improve the financial feasibility of a project. According to the National Housing Trust, the average rent hike in such a case is 45% (Bodaken and Heitlinger 2002). Although this is an average number from a number of years ago for conversions nationwide, it does provide a sense of the impact on rent and loss of affordability. A property owner may also choose to convert to condominiums as a way to quickly achieve a financial return on investment and abandon the long-term rental res ponsibilities of property management. Conversion to either market rate rentals or condominiums can be an attractive

PAGE 46

46 option in a strong housing market or a gentrifying neighborhood that will support high rents or high home prices relative to curr ent project rents. In a housing market where property values have appreciated, it can also be financially beneficial to sell a property. Especially if a property has a limited dividend restriction, the owner may be enticed to prepay or opt-out in order to improve return on equity, although substantial exit taxes on the capital gain may be due upon sale of a property. The physical state of a property can impact an owners decision. A structure may be deteriorating and in need of capital improvement s due to owners' neglect or lack of capital reserves. If the structure is located in a strong housing market and a low-poverty neighborhood, an owner has an opportunity to ra ise cash by conversion to market rate rentals or condominiums. Alternatively, an owner can sell the property to a third party who recognizes the prospect of increased property value. Another aspect of the physical state of a property is its design and functionality. An older apartment building may hold large units with three and more bedrooms, whereas the market may demand one or two bedrooms to accommodate smaller households. Functional obsolescence can motivat e an owner to sell the property or to redevelop and convert. The availability of funding and the relationshi p with the funding agency are factors that can motivate an owner to change its current c ourse. Additional funding might be necessary to improve the financial feasibility of a property and keep it from financial loss. Throughout the state of Florida, owners have seen their operating expenses incr ease as a result of higher property taxes and property insurance. But additional fina ncial support seems unavailable. After nearly 20 years of increases, growth in federal housing assistance ground to a halt in the second half of the 1990s" (Joint Center for H ousing Studies of Harvard Univ ersity 2006, 29). The relationship with the funding agency such as HUD can also be a factor in deciding wh ether to continue to

PAGE 47

47 operate a property as affordable housing. An ow ner may no longer want to work within the complex system of regulations and administ rative requirements. A recent study for HUD also cited difficult relations with local HUD offices (Finkel et al. 2006, 71) as a reason to opt-out. The type of ownership and an organizatio ns mission can also drive the decision about termination of affordability. The mandate of a no n-profit owner will be to serve lower income families and supply housing that is adequate a nd affordable (U.S. Government Accountability Office 2004). The risk of conversion to market rate rentals or condominiums is therefore considered marginal, if housing is owned by a non-profit entity. For-p rofit owners who are driven by the financial bottom line will make business decisions accordingly. They are more likely to exit a funding program if it makes fina ncial sense to do so. Personal motivations of management, such as retirement plans, can also impact the fate of a property.

PAGE 48

48 CHAPTER 4 RISK AND RISK ASSESSMENT METHODS Definition of Risk In the general area of finance, risk is define d as the variability of a realized value from the forecasted value of a variable (Weaver and Michelson 2004; Van Horne and Wachowicz 2001). Variability is the result of our imperf ect knowledge of the future, a consequence of change (Knight 1965, 198). While the knowledge about the future is imperfect, risk implies that previous knowledge about possible outcomes does exist. This previous knowledge makes risk measurable. Uncertainty, on the other hand, is not measurable and cannot be forecasted, since uncertainty assumes that any outcome is pos sible (UBC Real Estate Division 2001; Knight 1965). However, the concepts of risk and uncertainty are commonly used interchangeably (Slade 2006; Pyhrr 1973). In real estate finance and investment anal ysis, risk is the variance about a forecasted value of variables such as re nt, vacancy, operating e xpenses, cash flow, net present value and internal rate of return. Since the actual future values of variables can vary from the forecasted future values, risk is associated with th e forecast (Weaver and Michelson 2004). Financial analysts and investors perform ri sk analysis to determine which investment decisions to make based on the forecasted values, the risk associated with the forecasts and their tolerance to risk (Albright, Winston, and Zappe 2006). In the context of preservation, the concept of risk differs from its conventional meaning as applied in finance. In preser vation research, risk analysis or risk assessment is commonly described as an approach to identify the properties or types of properties at risk of loss to the affordable housing stock. In this context, the word risk takes on the definition as described by Knight (1965, 233): The word risk is ordinarily used in a loos e way to refer to any sort of

PAGE 49

49 uncertainty viewed from the standpoint of the unf avorable contingency; we speak of the risk of a loss. In other words, risk in preservation res earch does not refer to the variability in a value. Instead, it refers to the variabil ity in the expectation about th e continued affordability of an assisted property. More specifically, risk refers to the probability that a property loses its affordability as a result of: An owners decision to opt-out of a rental assistance contract, prepay a subsidized mortgage or terminate a use restriction, and to convert the property to market-rate housing. Physical deterioration and mortgage default. The Role of Property Data To effectively target resources and preser ve affordable housing, it is critical to know which properties are most likely to lose affordability. Therefore, the focus of data collection and analysis is to be on the property level. Acco rding to Recapitalization Advisors, Inc. (2002, 3), Collecting and distil ling data on the prope rties, the programs and financ ing structures that affect them, and establishing some basic definitions of prevalent terms would greatly facilitate the effective employment of our coll ective resources. But preservati on efforts have been hampered by a limited knowledge about the char acteristics of the subsidized housing stock and a lack of understanding of the motivations of property owners. Throughout the 1990s, several research papers identified the gap in information on multifamily housing and the weak multifamily housing datasets that lacked breadth and depth, especially compared to the research available on the single-family housing market (Follain 1994; Bogdon and Follain 1996). Follain (1994, 536, 564) stated that the combination of a growing demand for information and a literature that has not placed great emphasis on multifamily housing has produced an information gap. The study of multifamily housing is severely hampered by inadequate data. Researchers expressed the need for study of multifamily housing to address policy concerns such as loss of affordability

PAGE 50

50 (Galster, Tatian, and Wilson 1999). Recapitalization Advisors, Inc. (2002, 3) echoed the need for data for policy purposes, claiming that this lack of fundamental data handicaps the efforts of non-profit practitioners, governmental agencies and le nders who wish to assist in the effort to preserve the housing. It is difficu lt to know just where to apply re sources and which category of housing has the greatest need a nd will be the most responsive to cost-effective preservation. Policy therefore tends to be driven by anec dote and headline rather than by a thoughtful strategy. Michael Bodaken, President of the National Housing Trust, also recognized the knowledge gap. He identified expanded funding for research, education and data gathering as one of the critical steps to address the preservation issue (Bodaken 2002). Property-level data collection and data shari ng have improved in recent years as a result of the preservation debate and HUDs initiative to publicize information on expiring Section 8 rental assistance contracts (Southern Ca lifornia Association of Governments 2000). Organizations at the national, state and local levels, including state housing finance agencies, advocacy coalitions and university research centers, have started to collect data to take an inventory of assisted developmen ts within their jurisdictions. Su ch an inventory generally takes the shape of a development-level database that is populated with prop erty data (e.g., address information, number of units, subsidy programs) co llected from funding sources such as the U.S. Department of Housing and Urban Development, the U.S. Department of Agriculture Rural Development, state housing fina nce agencies and local governm ents. An inventory is often created to inform policy-makers, planners a nd advocates about the affordable housing supply. But several organizations have also started to use the inventory as a preservation tool by flagging the properties most likely to be lost. However, as expressed by Florid as Affordable Housing Study Commission (2005, 24), There are no widely available standardi zed risk analysis tools to

PAGE 51

51 assist states and local governments in identify ing and examining properties that may be facing expiration and/or opt-out situations so that pres ervation strategies can be built around the specific needs of each property. In order to develop a methodology to address the dissertation resear ch questions and to identify critical data variables, an extensive review was performed of risk assessment methods that were designed to analy ze the subsidized housing stock and the issue of preservation. Synthesis of the literature result ed in the classification of risk assessment methods by three types based on approach and objective. Th e sophistication of analysis and the depth of the data varied among the types of risk assessment methods, but prope rty-level data were at the core of all three types. The first type of risk assessment method iden tifies a handful of key indicators of risk (e.g., type of ownership, REAC Phys ical Inspection Score) and uses these to short list subsidized properties that meet the risk cr iteria. This type of method c ould also involve the rating of properties by level of risk. The objective of this approach is generally to flag specific properties in need of attention, or to assess the magnitude of properties at risk and in need of resources. The second type of risk assessment method is the statistical analysis of housing data for properties with terminated subsidie s and properties with current subs idies. The objective is to test which variables are significant indicators that a property is at risk of loss. The third type of method is the simulation of property owner choices with the objective to estimate the number of units at risk of loss a nd to inform policy-makers of the implications of this loss.

PAGE 52

52 Risk Assessment Method Type One: Target Inventory General Description and Purpose The most common risk assessment method is th e target inventory. This approach uses the development-level database of assisted housing to identify properties at hi ghest risk of loss as measured by the affordability expiration dates and a small number of other risk indicators. This method was found to be most common, because it is relatively simplistic and the data variables are most easily obtainable. The purpose of creating a target inventory is generally to inform policy-makers about the extent of the potential loss of affordable housi ng units, to advocate for preservation legislation and funding, to flag individual at-risk properties, and to prioritize the al location of preservation resources. The specific purpose is generally tied to the mission of the organization. For example, the Governors Task Force for Housing Preser vation in Wisconsin built an inventory of multifamily properties funded by HUD, RD and the Wisconsin Housing and Economic Development Authority with th e mission to identify and preser ve those affordable rental housing units at greatest risk of loss where the tenants residency is most threatened in order to maintain a positive impact on the stability of Wi sconsins residents and the continued sustained growth of Wisconsins economy and to make r ecommendations on how to best preserve those units (Governors Task Force fo r Housing Preservation 2004, 3). Methodology and Risk Indicators Most target inventories are based on a small number of key risk indicators: Subsidy or affordability expiration dates, ownership type and strength of the market. Some inventories only use the expiration dates to identify the properties at highest risk of loss. This was the approach taken by the Community Economic Developmen t Assistance Corporation (CEDAC). CEDAC (2008) listed all properties with federal or stat e subsidized or insured mortgages and properties

PAGE 53

53 with HUD rental assistance that we re at risk of leaving the stock by 2010 due to prepayment, full mortgage repayment or contract terminations. The City of Los Angeles ( 2002) also applied this method when analyzing at-risk housi ng for its Housing Element. It assessed the potential loss of federal, state and locally assisted housing between 2000 and 2010 according to the expiration year of affordability restrictions. The key indicators generally include the following: Subsidy or affordability expiration dates, ownership type, and strength of the market. Expiration dates include the date of eligibility for prepayment of a subsidized mortgage, maturi ty date, rental assistance contract expiration date, or use restriction e xpiration date. If an expiration date is imminent, the risk of loss to the affordable housing stock is higher, because a property owner will soon have the option to terminate affordability. When a property has multip le funding layers, the expiration date of the most restrictive program is generally applied in the analysis. For example, if a property owned by a non-profit has a HUD rental assistance cont ract that expires in 2009 and a HUD Section 221(d)(3) BMIR mortgage that matures in 2012, the end date of 2012 is used under the assumption that the non-profit is not eligible to pr epay the mortgage and that the rental assistance contract will get renewed (Shimberg Center for Affordable Housing 2007). Many target inventories define a timeframe for analysis in or der to focus on the properties at highest risk of loss. For example, the National Housing Trust (2006b) reported on Section 8 contracts due to expire by the end of fiscal year 2011. The type of ownership and an organiza tions mission can drive the decision about termination of affordability, hence its use as a key indicator. For-profit owners have a strong focus on the financial bottom line and aim for maximization of returns (Wallace 1995; Pedone 1991). A for-profit is more likely to exit the funding program and se ll the property or convert to

PAGE 54

54 market-rate housing if it makes financial sense to do so. The mandate of a non-profit owner is generally to serve lower income families in the community, and therefore the risk of conversion is considered marginal. A study prepared for HUD found that non-profits were less likely to opt out of a rental assistance contract compared to for-profits, be cause nonprofit owners are often mission-driven to continue to prov ide affordable housing (Finkel et al. 2006, ix). But the risk of deterioration was considered higher for pr operties owned by non-pr ofits (Recapitalization Advisors, Inc. 2002). The conditions in the local housing market can also impact the level of risk that a property will convert to market-ra te housing. Conversion risk is cons idered higher in tight rental markets (Recapitalization Advisors, Inc. 2002). If a property is located in a distressed area with high poverty, the risk of deterioration and default is generally higher. The strength of the market can be measured by various indicators: Ratio of project rents to market rents. A weak ratio (below 1) is an indication of higher risk of loss, because an owner has greater oppor tunity to improve rental revenue through conversion (Finkel et al. 2006; Southern California Associ ation of Governments 2000). Neighborhood characteristics such as area vacan cy rate, poverty rate and median income. If vacancy and poverty rates are relatively lo w and median income is relatively high or improving, the local market is considered re latively strong (Finkel et al. 2006; United States General Accounting Office 2004). Home price appreciation. The year-over-year change in median home sales price provides another proxy for strength of the market. In a 2002 report for Cook County, Illinois, Recapitalization Advisors, Inc. (2002) categoriz ed a market as strong if it had a positive change in median home sales price greater than 20% between 1997 and 2000, stable if it had a positive change of less than 20%, and weak in the case of a negative change. Physical condition of the prope rty is another indicator that was often mentioned as an important variable in the assessm ent of the risk of loss due to conversion or deterioration and default (United States General Accounting O ffice 2004). A property in good physical condition has a higher conversion potential (A chtenberg 2002). A deteriorated pr operty that is in need of

PAGE 55

55 capital improvements due to owners' neglect or lack of capital reserves is at a higher risk of mortgage default. Most target inventories do no t incorporate physical condition as an indicator, because of the lack of data. Unless an organi zation has access to capital needs assessments for the assisted housing that it is track ing, the only publicly available information that can be used as a proxy for physical condition is the physical inspection score for HUD properties. The HUD Real Estate Assessment Center (REAC) performs inspections and assigns this score, referred to as the REAC score. HUD only started to make th e REAC score publicly available on its website as of November 2007. But the United States Ge neral Accounting Office (GAO) had prior access to REAC data and incorporated this informati on in a state-by-state in ventory of HUD properties with maturing mortgages and expiring rental assistance contracts (U nited States General Accounting Office 2004). GAO created a target inve ntory and included all properties with a maturity or expiration date between 2003 and 2013. In addition to general property information such as the number of units a nd the actual date of subsidy expi ration, the inventory reported the REAC inspection score, ownership type, targ et population and economic occupancy rate.1 This allowed for anyone to conduct a more thorough risk assessment of properties at risk by 2013, for example by narrowing down the list of prope rties to those with low REAC scores. The year of construction or i ssuance of the certificate of occupancy could also be used as a proxy for the physical condition, assuming that olde r properties have greater capital needs. But this information is not commonly available from public datasets. In the case of HUD properties, the mortgage origination date could serve as a proxy. However, the ag e of the property is no indicator for the physical condition if a structure was rehabilitated. 1 GAO described the economic occupancy rate as the income received from the rented units in a property divided by the income that would be received if all units were occupied.

PAGE 56

56 Data Sources A target inventory can be created for one or several types of funding programs. For example, the Housing Development Center ( 2006) in Portland, Oregon completed a risk assessment of Oregons Low-Income Housing Tax Credit properties that were reaching the year fifteen of the use restriction between 2006 a nd 2011. More commonly, a target inventory is created for multiple programs such as HUD rent al assistance, HUD insured mortgages and Rural Development loans. The availability of data a nd the purpose of analysis seemed to drive the focus of an inventory, whether it report ed on one or several funding programs. The following are data sources that are co mmonly used to build a target inventory: HUD Insured Multifamily Mortgages Database. This dataset is available online and updated quarterly. Some entities (e.g., Shimberg Center for Housing Studies) receive supplemental data from state HUD offices. HUD Multifamily Assistance and Section 8 Contracts Database. This da taset is available online and updated abou t every two months. HUD Low-Income Housing Tax Credit Database This dataset is available online and currently reports on properties placed in service be tween 1987 and 2005. Rural Development. Data on RD loans and re ntal assistance are not available on the RD website, but can often be obtained from st ate RD offices or the Housing Assistance Council. State and local programs. Data on statefunded properties can be supplied by housing finance agencies (HFAs). State programs commonly include bonds, HOME and programs funded by housing trust funds. Some state HFAs provide data on tax credit properties, which are often more up to date than th e information available through the HUD LIHTC Database. Data on locally-funded properties are available from munici pal departments and local housing finance authoritie s. If the data collection enti ty has a broad geographic scope that spans multiple jurisdictions or even an entire state, gathering information on local programs and the properties that have b een funded can be an intense process. Portfolio data. An entity that owns or manage s a portfolio of assisted properties (e.g., a state housing finance agency) can use its own data to create a target inventory. It may use the public data sources to expand the inve ntory with information on other assisted developments in its jurisdiction.

PAGE 57

57 Market data. Information on home prices, me dian income, poverty rate, crime rate and other market and neighborhood ch aracteristics can be retrieved from sources such as the U.S. Census Bureau, local realtor as sociations and property appraisers. Output Format and Updates The output format of a target inventory takes the form of a list of individual properties or aggregate counts of types of properties at ri sk. For example, the United States General Accounting Office (2004) publishe d a state-by-state list of HUD properties with maturing mortgages and expiring rental assistance contracts by 2013. GAO also created tables and graphs to report the total number of properties and uni ts by HUD funding program and by the year of mortgage maturity or rental assi stance expiration. The out put of the target inventory can also be mapped. LISC (2005) created maps for metropol itan areas, which plotted the location of federally-assisted properties and identified the ty pe of ownership of each property (non-profit or for-profit) and timeframe of th e rental assistance contract expiration (2005-2009 or after 2009). The maps also included median house hold income categories by census tracts. While some target inventories are system atically updated on a re gular basis (at least annually), others are the result of one-time or episodic efforts. Risk Assessment Method Type One: Risk Rating General Description and Purpose Several entities have taken the target inventory approach one step further by categorizing each at-risk property by the level of risk of loss. The level of risk is generally determined by a small set of risk indicators such as su bsidy expiration date and ownership type. The purpose of rating properties by level of risk is the same as that of creating a target inventory: To inform policy-makers about the ex tent of the potential loss of affordable housing units, to advocate for preservation legislation an d funding, to flag indivi dual at-risk properties, and to prioritize the allocation of preservation resources. As a specific example, the California

PAGE 58

58 Housing Partnership Corporation (CHPC) bu ilt a state-wide inventory of HUD and RD properties and classified each property as low ri sk, lower risk or at risk. CHPCs purpose for creating this inventory and risk assessment was derived from its mission, which was to assist non-profit and government housing agencies to crea te, acquire and preserve housing affordable to lower income households, while providing leadership on housing preservation policy and funding (California Housing Partnership Corporation 2006, 1). Methodology and Risk Indicators The level of risk is determined by a small set of risk indicators, which are the same indicators as those app lied in the target inventory method: Subsidy or affordability expiration dates, ownership type and strength of the market. Some risk ratings may incorporate additional indicators. Most risk ratings classify properties as lower risk, medium risk and higher risk. A property is generally considered at lower risk of loss if the affordability end date is not imminent and if it is owned by a non-profit. A weak local housing market can also be used as an indicator of lower risk. As an example, the Washington Low Income Housing Network2 conducted a risk assessment of HUD properties with Section 8 pr oject-based assistance in Washington State. Properties owned by non-profits with a housing mission or properties with use restrictions of 20 years or more were classified as preserved. Pr operties located in non-tight housing markets were deemed at lower risk. The rental vacancy rate was used to measure the condition of the market (non-tight market if above 6%). Where vacancy da ta were not available, the percentage change in the median home price was calculated for the past year to gauge the strength of the market (non-tight market if less than 10% change). For counties where median home prices were also 2 In 2003, the Washington Low Income Housing Network merged with the Washington Low Income Housing Congress and formed the Washington Low Income Housing Alliance.

PAGE 59

59 not available, the following data variables were used: Percentage changes in the number of home sales and building permits issued, and number of households paying more than 35% of income on rent. The Washington Low Income Housing Network also classified properties as lower risk if they had undergone a debt restructuring and project re nt reduction under HUDs Mark-toMarket program (Farley 2002). A property is often categorized as medium risk if the subsidy expiration is either imminent or if it is coming up in the medium term, and if the ownership is for-profit. The California Housing Partnership Corporation (CHP C 2006) considers these two indicators in its risk assessment of HUD and RD properties. Accordi ng to CHPC, a property is at moderate risk if it can convert to market rate housing in five to ten years. If it is owne d by a non-profit entity, the risk is reduced by one level. For example, a proper ty with a subsidy expiring in five to ten years and with non-profit ownership is re duced to the lowest risk level. Some risk ratings also include the strength of the market as an indicator, whic h can be measured in various ways. CHPC (2001) performed a conversion risk analys is of tax credit properties th at were constructed during the first years of the program (1987-1989) when the us e restriction was only fifteen years. Due to a lack of other data, it used the county median income as a percentage of the statewide median income as a measure of the local market. Prope rties in medianor low-income counties were placed in the medium risk category if they also had non-profit ow nership and no other affordability restrictions. The Southern Calif ornia Association of Governments (SCAG 2000) used the project rent as a percen tage of market rent to measure the strength of the market. SCAG assessed the risk of loss of prope rties with HUD project-based rental assistance and classified the following property types as moderate risk: HUD re ntal assistance contract expiration scheduled

PAGE 60

60 to occur within five years, owner type is prof it-motivated, and the project rent is between 105 and 120% of the estimated potenti al market rent in the area. A property is typically classifi ed as high risk if the affordab ility period is due to expire within the short term, if it is owned by a for-profit organizati on and if it is located in a strong housing market. These are the indicators incorporat ed in the risk assessm ent of properties with HUD project-based rental assistance by th e Chicago Rehab Network (CRN 2003). CRN considered a property most at risk if it had a rental as sistance contract due to expire within the year, if the owner was a for-profit entity, a nd if the property was located in a booming or gentrifying area. SCAG (2000) took a similar ap proach. A high risk property had a rental assistance contract that e xpired within five years, a for-profit owner, and a project rent that is 105% or less of the estimated potential rent in the area. Data Sources The data sources used to rate the risk leve l of assisted properties are the same as those discussed in the section on the targ et inventory risk assessment method. Output Format and Updates Similar to the target inventory, the results of the risk rating of properties are either presented in aggregate form or ar e provided for each development. Risk ratings can be updated according to a regular schedule (at le ast annually), but are sometimes the result of one-time analysis. Risk Assessment Method Type Two: Cross Tabulations General Description and Purpose Finkel et al. (2006) prepared a national study for HUD that a ssessed the characteristics of properties that left the assisted housing stock through mo rtgage prepayment or rental assistance opt-out and compared these to pr operties that have remained in the stock. As part of the

PAGE 61

61 quantitative analysis, descriptiv e cross tabulations were crea ted in order to examine the properties according to the followi ng characteristics: Property, owner, financing, location, tenant, and physical and financial operating characteristics. The purpose of the study was to identify the ch aracteristics of propert ies that opted out or prepaid compared to those that remained assisted, and to assess the rents in lost properties. By gaining an understanding of the factors that impact the decision-making of owners, the study was able to present policy recommendations to HUD fo r preventing further loss of assisted housing. Cross tabulations were used as a method to determ ine the characteristics of properties that were more likely to opt out or prepa y. The purpose of the cross tabula tions was also to identify the explanatory variables to include in a regression analysis to asse ss the impact of variables on an owners decision to opt-in or opt-out of a Section 8 rental assistance contract. Methodology and Risk Indicators The first step in the cross tabulations method applied by Finkel et al. (2006) was to divide the universe of HUD properties into categories of property types according to the original subsidy program and the current funding status. Th e final categories for analysis were opt-ins, opt-outs/prepays, foreclosure/enforcement, and all other. The next step was to statistically describe the characteristics of each of the property types. The characteristics that were described were considered potential factors in the decisi on-making of owners. A ma ster file was created that included all the variables for 22,471 unique properties located throughout the country. These variables were organized as follows: Property: Development size in units, unit si ze in bedrooms, targ et population, building type, HUD program type, average pe rcentage of assisted units, pr oject rent to Fair Market Rent ratio, and building age. Owner: Ownership type and ma nagement review score. Financing: Primary form of financing and Housing Finance Agency-related properties.

PAGE 62

62 Location: Census division, metropolitan locat ion, and neighborhood characteristics (e.g., median income, poverty rate, vacancy rate). Tenant: Length of residence, household size percent minority-head ed, percent household heads with disabilities, per cent elderly-headed households, percent households with children, and household income as a percentage of area median income. Physical and financial oper ating: REAC physical inspection score, REAC financial performance score, financial ratios (e.g., expe nse-to-income ratio, debt-service-coverage), surplus cash level, reserve, vacan cy rate, and operating expenses. Percentages, means and medians were calcula ted for each of the four property types. The data were presented by characteristic in separate tables. Table 4-1 gives an example of a crosstab. Analysis of the cross tabulations concluded th at properties were more likely to opt out if they were older, if they were occupied by famili es, if the owner was a nonprofit entity, or if the project rents were substantially below the Fair Market Rent. Data Sources Since the study was commissioned by HUD, Finkel et al. had access to datasets and variables that are not publicly available. This allo wed for a more rigorous analysis of the assisted housing stock compared to the target inventory and risk rating methods that generally rely on publicly available property data. The data sources included the following: HUD Office of Housings (FHA) Real Estate Management System (REMS) Data. This contained propertyand c ontract-level information. HUD FHAs Multifamily DataMart (MPRD) file s. These files included mortgage and contract data for active properties. HUD FHAs Multifamily Insurance System (MFIS) or F-47 data. Mortgage financing data was reported in this dataset. HUD Real Estate Assessment Center (REAC) Data. This was the source for physical condition and financial op erating characteristics.

PAGE 63

63 Tenant Rental Assistance Certification Syst em (TRACS). This system contained data on tenant characteristics. PIH Information Center (PIC) data. These data were used to retrieve information on rents. 1990 and 2000 Census of Population and Housi ng data. Neighborhood char acteristics were based on Census data. FHAs List of Opt-out Propertie s (Opt-out List). This list re ported the properties that completed the rental assistance contract opt-out process. Output Format and Updates The results of the cross tabulations were a ggregated at the national level and presented by property type and characteristic for the enti re national sample of 22,471 properties. The study was performed under contract with HUD as a one-time analysis. Risk Assessment Method Type Two: Regression Analysis General Description and Purpose Regression analysis is anothe r method that can be used to explain the correlation between an owners decision (as the dependent variable ) and characteristics of a development, its ownership, terms of the funding program and lo cation (as the independe nt or explanatory variables). This method requires knowledge about the owners decision or intent, as well as information on numerous property characteristic s. Publicly available datasets contain only limited historical data that reflect owners decisions (e.g., prepayment of HUD subsidized mortgages), a limited number of data variables, and no information on owners intent. Because of these data limitations and becau se of the higher statistical comp lexity, regression analysis is not commonly performed by the entities that main tain assisted housing inventories. But this section will discuss two studies th at used regression analysis to analyze the affordable housing stock and the decision of owners. Finkel et al. (2006) prepared a study for HUD and was provided with multiple datasets that enabled them to construct a multivariate logistic regression

PAGE 64

64 model in order to analyze an owners decision to opt-in or opt-out of a project-based rental assistance contract. Melendez, Schwartz, and de Montrichard (2007) conducted a survey of owners and developers of properties built under the Low-Income Housing Tax Credit program during 1987 to 1989. They developed an ordered logit model based on the extensive data collected on the owners intent after expiration of the fifteen year compliance period, property characteristics, ownership struct ure and affordability restrictions. The purpose of the multivariate regression analysis for the HUD properties was to test the observations that resulted from th e cross tabulations (as discussed in the previous section), and to identify the characteristics of lost properties in order to enable policy-makers at HUD to predict and monitor which properties are most likely to opt-out of a rental a ssistance contract. The purpose of the ordered logit model for the tax credit properties was to examine the primary determinants of risk [of losing affordability] for properties with credit allocated between 1987 and 1989 (Melendez, Schwartz and de Montrichard 2007, 1). Methodology and Risk Indicators The multivariate regression analysis only incorp orated the decision to opt-in or opt-out of a HUD Section 8 project-based re ntal assistance contract. It excluded the decision about prepayment of a HUD subsidized mortgage. Fi nkel et al. (2006, 16) explained that by narrowing the focus in this way, we avoided ha ving to account for two different decisions (opting out of project-b ased Section 8 and mortgage prepay ment) with the same model. The owners decision was the dependent variable that took a value of 0 (opt-in) or 1 (opt-out). The explanatory variables in the model were derive d from the cross tabula tions and are listed in Table 4-2. The sample contained a total of 8,992 properties with non-missing values for all variables, of which 763 prope rties (8.5%) were opt-outs.

PAGE 65

65 The regression model analyzed the relationship between each explanatory variable (each property characteristic) and an owners decision, while keeping a ll other variables constant. The results were presented in odds ratio format, which means that a variable with an odds ratio estimate larger than 1.0 had a pos itive impact on the decision to opt-out; an odds ratio estimate smaller than 1.0 implies that the property characteris tic reduced the likelihood of opting out (see Table 4-3). The regression analysis found that most variables were st atistically significant. It also concluded that the key explanatory variable yi elded by the multivariate analyses appears to be the rent-to-FMR ratio: the lower the rent-to-FM R ratio, the higher the likelihood of opting out (Finkel et al. 2006, 33). When the project rent is relatively low compared to the Fair Market Rent, the owner has a greater opportunity to im prove rent revenues by opt ing out and converting to market rate housing. Another key variable wa s type of ownership; non-profit owners were significantly less likely to optout compared to other owners. Other findings included that properties with the following characteristics were more likely to opt-out (holding each other variable constant): 100% of th e units have rental assistance ; family-occupied; fewer than 50 units; unit mix with three or fewer bedrooms; ol der assisted properties; low-poverty rate census tracts; and central city or non-metropolitan locations. For the ordered logit model, a telephone survey was completed for 164 tax credit properties placed in service between 1987 a nd 1989 in metropolitan areas throughout the country. The level of risk of losing affordability was the dependent vari able and was ranked on a scale of 1 to 6. The level of risk was based on the owners intent to continue the ownership of the property and to maintain affordability after the ex piration of the fifteen year use restriction. The owners intent was identified through the survey and categor ized as follows (Melendez, Schwartz and de Montrichard 2007, 13):

PAGE 66

66 Affordability could be continued: o When the owner said that maintaining the propertys affordability was very important. o When the owner said continued affordability was only somewhat or not too important. o When the owner intends to sell the prope rty to an entity that will maintain affordability. Affordability may be lost: o When the owner plans to sell the propert y and is not intere sted in keeping it affordable or is undecided about what to do with the property. o When the owner plans to convert the propert y to market-rate occupancy, or when the property has already been sold and is at risk of converting to market rate. o When the property has already been sold without any affordability guarantees; this situation is thus ranked as havi ng the highest risk on the scale. The explanatory variables were also collect ed through the survey, and included data on the basic property characteristics, location, type of sponsor and ownership structure, additional affordability restrictions, occupancy rate, replacement reserves and rehabilitation needs. The model analyzed the relationship betw een the risk level and each explanatory variable, while keeping all other variables constant. It found that a property had a lower risk of losing affordability if it had a non-profit sponsor, if additional affordability restrictions were in place beyond the year fifteen, or if the property had extensive reha bilitation needs. Contrary to the expectations of the research ers, a tax credit property located in a high rent housing market by itself was not found to be a factor in an owners decision to convert. Data Sources Finkel et al. was able to ach ieve a relatively large sample size and conduct the regression analysis thanks to access to numerous HUD intern al data sources with detailed property-level

PAGE 67

67 information for both the lost and remaining assi sted housing stock. Many of these data elements, especially for the lost units, are not publicly av ailable. The data sources used for the regression analysis were the same as those for the cross-tabulations. Melendez, Schwartz and deMontrichard relied on data from the HUD Low-Income Housing Tax Credit Database and conducted a tele phone survey of owners and developers to collect detailed data on the owners intent and ch aracteristics of the prope rties. The researchers also held interviews with tax credit syndicators about acquisition, financing and rehabilitation of tax credit properties. Output Format and Updates The regression analysis was performed fo r the national sample. The results were presented in the table of coefficient estimates (Table 4-3). The study was carried out under contract with HUD as a one-time analysis. The results of the logit model and a descrip tive analysis of the ta x credit properties were reported in the study and presented in tables. The study was recently re leased (2007) and does not make any reference to plans of updating the analysis. Risk Assessment Method Type Three: Simulation Modeling General Description and Purpose Another method to assess the number of a ssisted properties at risk of loss is the simulation model, which is a model that simulates the decisions of owners to end or continue the affordability. This is a rather complex method, b ecause it has to incorporate a large number of possible interactions among owner decisions economic trends, and HUD rules and funding availability (Wallace 1995, 44). It also requi res access to many data variables and the understanding of simulation software. Therefore, the simulation model is not used as a common risk assessment method by state and local entities that maintain a ssisted housing inventories. But

PAGE 68

68 a simulation model was developed for the HUD portfolio of insured and assisted properties in the late 1980s; the model was refined during the early 1990s. In 1987, the National Low Income Housing Preservation Commission (1988) was created to assess the risk of loss of HUD-assisted properties built during the 1960s to early 1970s The Commission contracted with Abt Associates, Inc. to create an economic model to simulate owners decisions. Dr. James E. Wallace of Abt was the technical director for the Commission. The simulation model that was developed for the Commission was based on his doctoral thesis research (Wallace 1995). Wallace further refined the model under a cont ract with HUD in 1992, as discussed in this chapter. The purpose of the simulation model for the HUD properties was to assess the number of units at potential risk of loss a nd the impact on tenants, and to analyze the costs and effects of policy solutions (Wallace 1995; National Low In come Housing Preser vation Commission 1988). Methodology and Risk Indicators The model developed by Wallace was based on project-specific data and environment or context data to simulate the op tions that were available to pr operty owners, under the assumption that owners make economically ra tional decisions. The following data variables were collected for a sample of 570 HUD properties, representi ng a total of 13,271 properties with mortgages insured or held by HUD: Property-specific data, includi ng ownership type, mortgage am ount and status, section of the federal housing act, annual income and expenses, cost of meeting physical needs (backlog of repairs and replacements neede d, beyond normal maintenance, to restore a property to original working c ondition), expected accrual of fu ture repair and replacement needs, highest and best use (or condominium prices), costs of upgrading to unassisted market use, and tenant income distributions. Environment or context data, including infla tion rate for repairs and operating expenses, another inflation rate for rent s and prices, discount rates, lo an underwriting terms, and tax rates. Model parameters allow the user to specify a number of policy and budgetary conditions, such as the tenant income level eligible for new Section 8 Loan Management

PAGE 69

69 Set Aside assistance, the funding priority sc ore necessary to receive LMSA funding, and whether the Low-Income Housi ng Tax Credit is available. The input data were first used to make ba sic projections for future potential revenues, operating costs and capital needs for each property. Next, the model tested the options that were available to owners, which included the following: Continued operation, either status quo or with the acceptance of federal preservation incentives. Disposition of the property th rough conversion to market-rate rentals by the current owner, through sale and subsequent conversion to market-rate rentals or condominiums by a new owner, through abandonment and deed-in-lieu of foreclosure, through sale as a LowIncome Housing Tax Credit project, or through transfer to a preserving entity. For each property, the model simulated altern ative paths for all the possible combinations of operation and disposition decisions that an ow ner can make until the maturity of a mortgage. Then Wallace (1992, A-2) made the assumption that a for-profit owner choos es the path that yields the highest discounted present value of the stream of after-tax returns from annual operation and eventual disposition. Non-profit owners are assumed to operate through the mortgage term, if possible; othe rwise to sell the property as a Low-Income Housing Tax Credit project, if possible; and, as a last alternative, to resign the property (submit the deed in lieu of foreclosure), which is triggere d by cumulative cash deficits. The model also incorporated HUD rules and regulations related to the availability of supplemental assistance and the conditions for forecl osure. The ultimate path of a property over time was determined by the combination of ow ner and HUD choices. For this path, the model calculated the following three outpu ts: The predicted year of dis position, the characteristics of tenants, and the cost to the government.

PAGE 70

70 The model predicted that during the 20-year timeframe, 33% of older assisted properties3 would foreclose, 24% would sell to a non-prof it buyer, and 38% continue operation. Only 4% were expected to convert to market-rate housi ng, because many older assisted properties did not have this option. Among ne wer assisted properties4, 79% were predicted to continue operation and 20% were expected to convert to unassisted housing. Model parameters such as inflation rate s and the availability of government funding could be changed in order to assess the imp act on the future of assisted properties. Data Sources The development of the simulation m odel was only possible because of the comprehensive data that were made available by HUD and that were gath ered through intensive collection efforts. HUD provided se veral datasets from its comput erized data systems, which contained basic information on the assisted de velopments such as occupancy type, funding program and total units. HUD also supplied proper ty-level financial information related to mortgage terms, revenues and expenses, whic h was verified and supplemented by HUD field offices. To obtain information on the physical condition of properties, on-site inspections were performed. A survey of local real estate experts and HUD field offices was conducted to determine potential rents of unassisted propertie s and condominium sales prices, and to identify comparable properties. Property owners and managers were also surveyed to obtain missing financial information, tenant characteris tics and ownership structure information. 3 Properties built during the 1960s to mid-1970s under the following programs: Section 221(d)(3) Below Market Interest Rate, Section 236, Loan Management Set Aside, Rent Supplement or Rental Assistance Payment, and Section 8 Property Disposition. 4 Properties built since the mid-1970s through the 1980s under the following programs: Section 8 New Construction, Section 8 Substantial Rehabilitation and Section 8 Moderate Rehabilitation.

PAGE 71

71 Output Format and Updates The model made annual predictions of the stat us of each of the properties in the sample over a 20 year timeframe, starting in 1990. The results were weighted up to the universe of more than 13,000 HUD properties nationwide. The report produced tables with a count of properties and units by property status (e.g., market conversion, foreclosure) at five year intervals under the baseline condition of current HUD funding5 and under a full funding scenario6. Two excerpts of output tables are illustrated in Figures 4-1 and 4-2. The report also provided tables with predicted government costs for each property status unde r various scenarios of funding availability. The study was carried out under contract with HUD as a one-time analysis. Other Risk Indicators The literature noted other variables that are indi cators of risk of loss, in addition to those incorporated in the risk assessment methods discussed. But info rmation for these variables was often not available to the entities that maintain assisted housing inventories, because it requires direct interaction with the owners or access to documentation that owners do not want to make public or that funding agencies cannot share under privacy rules. Owner motivation is an example of anot her indicator. The Vermont Housing Finance Agency (1988, 7) suggested that it is important to continue an active dialogue with the owners of properties in the higher risk categories to determine their needs, concerns and long-term objectives. The Wisconsin Governors Task Force for Housing Preservation (2004, 5) also warned that owner participat ion is critical in achieving preservation of at-risk housing. Owners decisions are driven by their motivatio ns, which are impossible to pinpoint without 5 Current HUD funding relates to the HUD obliga tions and rental assistance contr acts that were in place at the time that the simulation analysis was performed. 6 The full funding scenario assumed the addition of new rent al assistance contracts and re duced-interest direct loans for all eligible properties.

PAGE 72

72 direct interaction. Motiva tions are impacted by financial consider ations such as tax benefits and ongoing federal funding, but also by personal motives such as retirement plans (Recapitalization Advisors, Inc. 2002). A property can be at higher risk of loss if tax benefits are exhausted, if ongoing funding is uncertain, or if an owner wants to scale back its portf olio or has plans to retire. Ownership structure can provide additional insight into the risk of loss. Most assisted developments are owned by limited partnerships th at consist of general partners who manage day-to-day operations and limited partners who are the investors (Achtenberg 2002). Limited and general partners can develop conflicting interests, which can impact the future of an assisted property. For example, the limited partner in a Low-Income Housing Tax Credit property may want to sell after the fifteen year compliance period expires a nd all tax credits have been received. But the general partner may be most in terested in continuing to operate the property and serve low-income households. The general partner will have to buy out the limited partner, which could require a large amount of capital that may not be available (Melendez, Schwartz and de Montrichard 2007). Portfolio risk is another indicator. Propert y owners with small por tfolios may lack the resources, access to capital and expertise to e ffectively manage their assets, thereby increasing the risk of deterioration and default (Recapitali zation Advisors, Inc. 2002). Owners with larger portfolios may have affiliated pr operty management companies that rely on the portfolio for fees and may therefore be reluctant to sell (Achte nberg 2002; Recapitalization Advisors, Inc. 2002). Information on the size of an owners portfo lio is not easily obta inable, even though HUD reports owner names in its public datasets. Many properties are single-p urpose entities with unique limited partnership names. The actual sp onsor can therefore not be identified. But

PAGE 73

73 Recapitalization Advisors (2002) explained that many properties are managed by the sponsor and made the broad assumption that the name of th e management company can be used as a proxy for owner. Compliance requirements and regulations can also impact an owners decision. Subsidized properties are subjec t to compliance requirements and regulations as imposed by the funding entity. A factor in the decision of a property owner to terminate affordability can be its reluctance to continue to operate under comp lex rules. Achtenberg (2002, 38) called it HUD fatigue in the case of HUD properties. The Hous ing Development Center (2006) reported that even non-profit entities ar e looking to sell tax credit properti es after the initi al fifteen year compliance period expires, because of ongoing compliance requirements and reporting burdens. The administration of funding is especially complex when a property has various subsidy layers (Governors Task Force for Housing Preservation 2004). Exit taxes can play a big role in an owners decision about the futu re use of a property. Exit taxes are payable by the owne r on the capital gain when a property is sold. As explained by Achtenberg (2002, 40), the capital gain consists of the cash proceeds realized on sale minus the owner's capital account. The capit al account is the original cash investment adjusted by cumulative profits and tax losses to date. After 20 years, properties that have provided generous depreciation and interest deductions but limited di vidends will typically have a negative capital account. In these cases, owners will owe taxes ev en if they realize no cash proceeds from the sale. Excessive exit taxes can provide an incentive to owners to retain the property and convert to market-rate rentals. If the local market is not strong enough to make conversion feasible, there is a risk of deterioration and default if the owne r is reluctant or unable to inject cash into the property for repairs and renovations (Governors Task Force for Housing Preservation 2004).

PAGE 74

74 Financial condition is also regarded as a ke y indicator. If a property suffers from poor cash flow and low reserves, it may be difficult fo r the owner to meet its mortgage obligations, thereby increasing the risk of default. In a strong housing market or gentrifying neighborhood, poor cash flow can motivate the ow ner to sell or convert to mark et-rate housing in order to increase rent revenue and im prove financial feasibility. Lastly, information on capital needs and reserv es would give importa nt insight into the risk of loss. A lack of reserves poses the risk that capital improvements cannot be made and that no resources are available to cover financial shortf alls. The risk of deterioration and default is heightened if a property has capital needs and if reserves are lim ited or non-existent. A survey of tax credit properties reaching year fifteen be tween 2006 and 2011 found that 30% of respondents did not know if they had adequate reserves a nd almost 14% claimed not to have sufficient reserves (Housing Development Center 2006). Table 4-1. Example of Cross Tabulations fo r Tenant Characteristics in HUD Properties Average Tenant Characteristics Opt-ins Opt-outs/ Prepays Foreclosure/ Enforcemen t All Other Total Number of properties 11,126 1,715 2,385 7,245 22,471 Percent of properties 49.5% 7.6% 10.6% 32.2% 100% Length of residence (years) 6.0 5.3 5.7 5.8 5.9 Household size 1.7 2.1 2.2 1.5 1.7 Percent minority-headed 42.1% 50.6% 72.7% 35.8% 42.4% Percent household heads with disabilities 18.5% 12.5% 13.6% 29.9% 21.6% Percent elderly-headed households 48.5% 27.9% 19.3% 47.5% 45.0% Percent households with children 25.0% 42.8% 48.6% 16.8% 24.9% Household income as a percentage of area median income (AMI) 27.7% 27.9% 23.8% 28.9% 27.8% Source: Finkel et al. (2006).

PAGE 75

75 Table 4-2. Regression Model Variables for the Logistic Regression Model of the Opt-out Decision Variable Variable Specification Expected Direction of Impact Development size in units Less than 50 units (reference category) 50-99 units 100-199 units 200+ units Unknown. On one hand, conversion to market rate may involve fixed costs; since larger projects have lower per-unit costs, this may increase their lik elihood of opting out. On the other hand, large projects tend to be associated with other physical features that are less attractive to unassisted tenants. Density Percent of 3-bedroom-plus units Negative. It may be harder to market projects with large units to unassisted tenants because these units may not be physically suitable for higher income singles and couples who could afford market rate units. Family occupancy type Family = 1 Elderly/disabled = 0 Positive. Elderly projects face competition from amenity-rich private market projects. Also, the income distribution among elderly and disabled households may not support many market rate units. In other words, family projects are more likely to opt out. Building type Detached or semi-detached = 1 Other = 0 Positive. Detached and semidetached projects tend to be associated with other amenities and physical characteristics that are attractive to unassisted tenants. Older Assisted HUD program types Older assisted = 1 Newer assisted = 0 Positive. The older projects often have rents that are below market rate. Ratio of rentto-FMR Rent-to-FMR ratio < 80% 80% < rent-to-FMR ratio < 100% 100% < rent-to-FMR ratio < 120% (reference category) 120% < rentto-FMR ratio < 130% 130% < rent-to-FMR ratio < 140% 140% < rent-to-FMR ratio < 160% Rent-to-FMR ratio > 160% Negative for projects with rents above local FMR. Projects with rents that are low relative to the FMR may be able to raise rents with little effect on vacancy rates. In other words, as rent-to-FMR ratio increases, we expect the property owner to be less motivated to opt out. Ownership type Nonprofit = 1 For-profit or limited dividend = 0 Negative. Nonprofits are less likely to opt out. By definition, for-profit owners are motivated to increase revenues.

PAGE 76

76 Table 4-2. Continued Variable Variable Specification Expected Direction of Impact Not federally financed mortgage Not federally financed = 1 Other = 0 Negative. This value is a proxy for projects financed by state Housing Finance Agencies (HFAs). HFAs may impose prepayment and/or optout restrictions. Neighborhood poverty rate Percent of persons in the surrounding census tract with incomes below poverty threshold in year 2000 Negative. Research has shown that tracts with high poverty rates typically have features that make them undesirable places to live and hence are less able to command high rents. 100-percent assisted Projects with 100-percent units receiving HUD assistance =1 Other = 0 Positive. A project with a high percentage of unassisted tenants risks high turnover upon conversion to private market status because these tenants will not have enhanced vouchers and may not be able or willing to afford the higher rents. A high percentage of assisted tenants implies more opportunity for the owner to raise rents to market levels. Metropolitan location Suburb (reference category) Central city Non-metropolitan Negative for central city. We expect owners in central cities to be less likely to opt out because markets may be unable to support unassisted housing. Positive for suburb. Suburban areas tend to have higher income renters to absorb market rate housing. Census division New England Mid Atlantic East North Central West North Central South Atlantic (reference category) East South Central West South Central Mountain Pacific Positive for high rent regions such as New England, Mid-Atlantic, and Pacific. Source: Finkel et al. (2006).

PAGE 77

77 Table 4-3. Coefficient Estimates of the Logist ic Regression Model of the Opt-out Decision Explanatory Variable Odds Ratio Tstatistic Development size Less than 50 units (reference category) 50-99 units 0.51 *** -6.04 100-199 units 0.38 *** -6.82 200 or more units 0.44 *** -3.06 Density Percent 3-bedroom-plus units in development 0.28 *** -5.88 Occupancy type Family occupancy type 2.30 *** 6.84 Elderly/disabled (reference category) Building type Detached or semi-detached building type 1.13 0.75 Ownership type Nonprofit sponsor type 0.16 *** -10.65 Program characteristics Older Assisted Section 8 2.37 *** 8.13 100% of project units are rece iving HUD assistance 13.92 *** 11.38 Not federally financed (proxy for HFA deals) 0.82 -1.47 Neighborhood characteristic Census tract poverty rate 0.97 *** -7.43 Rent-to-FMR ratio Rent-to-FMR ratio < 80% 11.56 *** 16.55 80% < rent-to-FMR ratio < 100% 2.91 *** 8.03 100% < rent-to-FMR ratio < 120% (reference category) 120% < rent-to-FMR ratio < 130% 0.53 *** -2.65 130% < rent-to-FMR ratio < 140% 0.48 ** -2.48 140% < rent-to-FMR ratio < 160% 0.19 *** -4.17 Rent-to-FMR ratio > 160% 0.22 *** -2.86 Metropolitan location Central city 1.49 *** 3.44 Non-metropolitan 1.29 1.65 Suburb (reference category) Census Division New England 0.95 -0.19 Mid Atlantic 1.32 1.17 East North Central 1.42 ** 2.23 West North Central 1.44 2.02

PAGE 78

78 Table 4-3. Continued East South Central 0.88 -0.57 West South Central 1.78 *** 3.19 Mountain 1.50 ** 2.00 Pacific 1.45 ** 2.33 South Atlantic (reference category) Other Regression Model Information Value Opt-out = 1; opt-in = 0 Total number of properties 8,992 Number of opt-out properties 763 Log-likelihood -1701.20 Pseudo R-square 0.35 Notes: *** indicates significance at the 0.01 level; ** indicates significance at the 0.05 level; indicates significance at the 0.10 level. Source: Finkel et al. (2006). Figure 4-1. Predicted Number of Older Assist ed HUD Properties and Un its by Property Status under Current HUD Obligations for 1990-1994. Source: Wallace (1992).

PAGE 79

79 Figure 4-2. Predicted Number of Older Assist ed HUD Properties and Un its by Property Status under Full Funding Scenario for 1990-1994. Source: Wallace (1992).

PAGE 80

80 CHAPTER 5 METHODOLOGY Research Overview The objective of the research was to address two main research questions related to failout risk and opt-out risk. What are the property, financial, subsidy and tena nt characteristics of properties identified at fail-out risk, as measured by the financial or physical condition? What are the property, financial, subsidy and tena nt characteristics of properties identified at opt-out risk, as measured by the opportunity to increase project rents and improve cash flow? The research methodology that was applie d to address the rese arch questions was simulation modeling of the net cas h flow of rental housing properties that receive HUD projectbased rental assistance in Duval and Miami-Dade County. A cash flow approach was used, because net operating income provides an indicati on of the financial and physical health of a property. Simulation modeling was applied, because many of the input vari ables to compose net cash flow statements were uncertain due to a lack information about the current and future financial and physical condition of subsidized properties. Missing property-level data included actual operating expenses, vacancy rates and capita l needs. Simulation modeling allowed for the estimation of uncertain values accord ing to probability distributions. Similar to the risk assessment methods disc ussed in Chapter 4, a database of properties was at the core of the net cash flow method. Data on assisted multifamily properties with projectbased rental assistance contracts were collected and merged to compose a database. From this database, development-level proformas were creat ed, using actual as well as simulated data for rent revenues, operating expenses and debt servic e. Descriptive analysis, significance tests and regression analysis were performed to analyze the simulated net operating income data and to

PAGE 81

81 characterize the properties that were identified at heightened risk of loss to the affordable housing stock. Net Cash Flow Approach to Fail-out Risk The study used proforma analys is to assess the risk of failout as a result of physical deterioration and mortgage default. This approach was based on the link between the financial condition and the physical state of a property. Ac cording to Recapitalization Advisors, Inc. (2008, 1), any propertys Net Operating Income (NOI) is strongly affected by its physical status. Wallace et al. (1993, 1-10) made the same link and stated that the current financial situation is particularly relevant in assessing wh ether a property is at risk of defaulting on its mortgage. As noted by Goodman (2004, 229), net ope rating income is used in several academic studies as a trigger for mortgage default. Goodman (2004, 243) went on to state that properties with negative cash flow have been shown to be the most likely to default on mortgages and ultimately be abandoned or othe rwise removed from the housing stock. Wallace et al. (1993) also found that ne t cash flow was a predictor of a distressed property, which was defined as a property whose combined physical and financial problem s are severe enough to jeopardize tenant well-being, impa ir sound operations, and (if not corrected) lead to financial failure of the property (Wallace et al. 1993, 1-2). The consequences of insufficient cash flow include missed payments to reserve accounts and deferred repairs and capital improvements (Recapitalization Advisors, Inc. 2002; Wallace et al. 1993). Hence, financial difficu lties are inevitably ti ed up with physical de terioration: It is rare to find one without the other (Recapitalization Advisors, Inc. 2008, 1). Pedone (1991) also believed that properties with insufficient cash flows and limited reserves were at risk of dete rioration and default. Under thes e financial conditions, owners are struggling to make mortgage payments and to address mounting repairs and capital needs.

PAGE 82

82 Pedone claimed that some owners may purposely avoid default and allow their properties to deteriorate by not incurring expens es for repairs and not infusing cash from sources external to the property. An owner may make this disinvestment decision if a property is located in a weaker housing market and does not have a higher and best use that would be more profitable, and if exit taxes are owed upon foreclosure. Achtenberg (2002, i) echoed that in weaker markets, subsidized housing is threatened by di sinvestment, default, and foreclosure. Net Cash Flow Approach to Opt-out Risk This study also used proforma analysis to asse ss the risk of opt-out of a rental assistance contract in order to convert th e property to market-rate housing. Th e reason for the net cash flow approach related to the link be tween an owners decision to te rminate affordability and the expected financial return that would result from this decisi on. In a study of the HUD-insured multifamily housing stock, Wallace et al. (1993, 2-57) st ated that the probability of an eligible owner converting from assisted to market use de pends on the revenues and costs associated with each [prepayment or opt-out] option. Recapitaliza tion Advisors, Inc. (2002) argued that owners were almost always motivated to prepay a subsidized mortgage or opt-out of a rental assistance contract if this action would result in a higher fi nancial return on the prope rty. This scenario was most realistic in very strong rental markets. Pedone (1991) also claimed that conversion to market-rate housing or another use can be a fina ncially attractive option in tight housing markets where property owners have the opportunity to ch arge higher rents and improve profitability. Another condition for market conversion in addition to a strong local market was the financial state of a property. In an analysis by Wallace (1992), prope rties that were predicted to be feasible for a market conversion we re in a healthy financial condition. Wallace et al. (1993) did point out that c onversion commonly required that costs were incurred for repairs and capital improvements. According to Pedone (1991, 247), converting to

PAGE 83

83 higher income use could be more profitable than continuing the current low-income use, even with rehabilitation costs, any othe r transition costs, and the costs of refinancing the mortgage at current interest rates, which will substantially exceed the subsidized interest rates. A 2006 study that was conducted for HUD found a correlation between an owners decision to terminate affordability and the financial motivation. Finkel et al. (2006) performed multivariate logistic regression analysis to asse ss the correlation between an owners decision to opt-out of a HUD rental assistan ce contract (dependent variable ) and characteristics of the property (independent variables). The study concluded that the key explanatory variable yielded by the multivariate analyses appears to be th e rent-to-FMR ratio: the lower the rent-to-FMR ratio, the higher the likelihood of opting out (Fin kel et al. 2006, 33). When the project rent was relatively low compared to the Fa ir Market Rent, the owner had a greater opportunity to improve rent revenues by opting out and conv erting to market rate housing. Simulation Modeling of Net Cash Flow Overview of Monte Carlo Simulation and Applica tion to Real Estate Simulation is a technique that can be used to quantitatively analyze decisions that involve uncertainty (Myerson 2005; Vose 1996). Monte Carlo is one type of simulation method. It differs from other simulation methods by attempting to in corporate the random uncertainty of the real world (Archer 2005, 1). In a Monte Carlo simulation computer model, a real-life situation is imitated and uncertainty is explicitly incorporat ed by assigning a range of possible values to input variables that are random, rather than using single-point estimates (Albright, Winston and Zappa 2006). Uncertain values are also referred to as stochastic values (Palisade Corporation 2008). The Monte Carlo method wa s developed for the nuclear industry during the 1940s (Schumann 2006; Vose 1996). Its name Monte Carlo originates from the city in Monaco in Southern France, known for its casino and gambling games such as roulette and dice, which are

PAGE 84

84 based on random selections of numbers (Sc humann 2006; UBC Real Estate Division 2001). Monte Carlo simulations are most commonly run in spreadsheets, as a result of th e popularity of simulation add-in software programs such as @RIS K and Crystal Ball that can be used in Excel (Albright, Winston and Zappa 2006; Vose 1996). Simulation is used as a risk analysis technique in real estate to ma ke investment decisions (Slade 2006; Pyhrr 1973). The use of simulation in real estate investment builds on the traditional discounted cash flow (DCF) method for the valuation of assets. Input variables that are common in real estate analysis relate to investment outlays (e.g., size and type of units, cost of land and construction), operations (e.g., rental revenue, operating expenses), financing (e.g., amount of debt, interest rate ) and reversion (e.g., property sa les price, holding period). The traditional DCF model is deterministic, because all input variables are single-point estimates and therefore the output variables ar e single-point estimates also (Pyhrr 1973). But in reality many input variables are random and ther efore uncertain. For example, the vacancy rate for next year could be estimated with a single value, but vacan cy is a random variable with uncertain future values. It is common for the real estate investor or analyst to perform sensitivity and scenario analysis based on the single-point estimates. Sensitivity analysis typically estimates minimum and maximum values of variables to assess th e impact on the outcome. But this approach does not recognize that the minimum and maximum valu es are less likely to occur compared to the single-point best guess value. Scen ario analysis is usually based on a small number of scenarios, which are combinations of single-point estima tes. However, several hundred scenarios could exist if the model has mu ltiple variables (Vose 1996). It is nex t to impossible to test the entire range of possible outputs (Kelliher and Mahoney 2000, 49).

PAGE 85

85 Simulation analysis addresses the limitations of deterministic, sensitivity and scenario analysis. The simulation risk analysis model inco rporates the uncertainty of random variables, considers the probabilities of values, and runs hundreds of what if sc enarios in a matter of seconds. Input and Output variables The simulation risk analysis model is proba bilistic in nature (Pyhrr 1973). In a simulation model, each random input variable has a probabi lity density function, which is a range of possible values and a probability distribution for these values, as determined by the analyst. The simulation software samples a random variable from every probability density function and calculates the output value for the combination of inputs. This pro cess is repeated a large number of times, e.g., 1,000 iterations. Each iteration can therefore be c onsidered a separate what if scenario. Since each random input variable ha s a range of possible values, the output of simulation analysis is not a si ngle-value, but a distribution of outcomes (Albright, Winston and Zappa 2006). While the simulation runs, all fo recasts stabilize toward a smooth frequency distribution (Schumann 2006, 13). The final outpu t after multiple iterations includes summary statistics of the iterations (e.g., mean, media n, standard deviation, minimum, maximum) and graphical displays (e.g., frequency charts) (Kelliher and Mahoney 2000). Probability Distributions For each uncertain variable, a type of probab ility distribution has to be selected. Vose (1996) identified three ways to categorize probability distributions: Continuous versus discrete; bounded versus unbounded; and parametric versus non-parametric. A continuous distribution is used when the variable can take any value within a defined range, implying that the value is infinitely divisible. The @RISK simulation software that was used in this research identified 31 types of continuous distributions such as the normal, logistic, gamma and uniform distributions.

PAGE 86

86 A discrete distribution is selected when the variable can only ta ke a finite number of values. Eight discrete distribution t ypes are defined by @RISK, including binominal and Poisson distributions. Probability distributions are also ei ther bounded or unbounded: A bounded distribution lies between two sp ecific values, which is the case for uniform and triangular distributions. An unbounded distribution has an infinite minimum value and an infinite maximum value. This applies to distributions su ch as the normal and logistic distribution. A partially bounded distribution is al so possible when either the mi nimum or the maximum value is constrained. The probability distributions can al so be distinguished by parametric versus nonparametric. A parametric distribution is specific to a problem and is theoretically derived, relying on mathematics to model the problem (Vose 1996) An exponential or l ognormal distribution can be parametric. A non-parametric distribution is considered a ge neral distribution such as a discrete, triangular or cumulative distribution, for which few a ssumptions about the population distribution are required (Agres ti and Finlay 1997). As desc ribed by Vose, the defining parameters for general distributions are features of the graph shape (1996, 56), which contrasts the parameters of parametric distributions whos e shape is borne of the mathematics describing a theoretical problem (1996, 56). Figure 5-1 illustrates the probability distributions as categorized by common, continuous and discrete distributions. The liter ature on the application of simula tion to real estate investment analysis most commonly used the following probability distributions: Normal, lognormal, triangular and uniform (Schumann 2006; Hoesli, Jani, and Bender 2006; French and Gabrielli 2005). These distributions are l east complex, widely applicable and generally understood among real estate analysts. However, both Schumann (2006) and French and Gabrielli (2005) made the argument that the triangular distribution was the most appropriate approach in real estate

PAGE 87

87 valuation, even though normal distribution was more statistically robust. The authors explained that the triangular distribution reflects the thought process of the analyst who thinks in terms of best, worst and most likely figures. Stochastic values have inherent uncertainty, hence the selection of probability distributions. To determine which type of probabili ty distribution to select, the uncertainty can be quantified through research of each variable, which often involves the analysis of current and historical data, as well as the i nput of expert opinion. When a reli able data source is available for an uncertain variable, the simulation software ca n fit a probability distribution to the observed data. But Pyhrr expressed that in real estate, where object ive data is sparse, or often nonexistent, the decision maker is forced to use probability estimates that are high ly subjective (1973, 62). Kelliher and Mahoney also pointed out that even when historical data is available, much of the distribution selection process is driven by subjective judgment, moderated by experience, with a final check for reasonableness (2000, 51). Vose (1996, 153) presented the following reasons for data limitations and the difficu lty in obtaining data to assess the uncertainty of variables: The data have simply never been collected in the past. The data are too expensive to obtain. Past data are no longer rele vant (new technology, changes in political or commercial environment, etc.). The data are sparse, requiring expe rt opinion to fill in the holes. The area being modeled is new. Rationale for Simulation Modeling Both Archer (2005) and Vose (1996) outlin ed criteria to be met if Monte Carlo simulation analysis is applied as a methodology. Archer explained that one of the following two conditions must exist. First, if the stochastic behavior is complex and difficult to represent by

PAGE 88

88 statistical models, then the flexib ility of Monte Carlo can enable a closer fit to the real process. Second, if the main question depends on a threshol d, such as with any option behavior (default, prepayment, lease renewal, etc.), Monte Carlo is well suited to examine the probability of crossing the decision threshold (Archer 2005, 4). According to Vose (1996), two criteria must be met. First, it must be possible to model th e problem. Second, the variables in the model have to be quantifiable. Simulation was considered a suitable method to address the research questions in this study, because all conditions as outlined by Archer and Vose were met: There were many uncertain input variables, which made the fail -out and opt-out proces ses complex; the main research questions depended on thresholds relate d to subsidy renewal and default; it was possible to model the risk of loss of affordable housing by using a cash flow appr oach; and the variables were quantifiable. Research Design Unit of Analysis, Population and Sampling Frame The property was the unit of anal ysis. It was the unit of analys is for several reasons. First, project-based housing subs idies are made at the property leve l. Second, the property is the level at which owners commonly make decisions c oncerning investment, dis position and opt-out of subsidy programs. Third, physical de terioration directly or indirec tly affects all units and thereby the entire property. Fourth, mortgages are secure d by the real property an d therefore mortgage default occurs at the property level. The total population size was 119 properties. The population was defined as the properties that were located in Miami-Dade and Duval County, and that had an active contract under any HUD project-based rent al assistance program (with one exception), which included the following programs

PAGE 89

89 Loan Management Set Aside Section 8 New Construction Section 8 Substantial Rehabilitation Section 8 Moderate Rehabilitation Property Disposition Preservation Rent Supplement Excluded from this list are the Project Re ntal Assistance Contracts that subsidize operating expenses of developments that are funded under the HUD Section 202 Capital Advance program for the elderly or the HUD Section 811 Capital Advance program for persons with disabilities. These programs were outside the scope of the research. The geographic focus was Miami-Dade and D uval, because the number of properties and units with project-based rental assistance in these counties trumpe d that of all other individual counties in Florida (HUD 2008a). More than on e third of properties and units covered under HUD rental assistance contracts we re located in Miami-Dade or Duval. Preventing the loss of properties with rental assistance was considered important, because both counties housed renters that made less than 60% of the area median income and were paying more than 40% of income on housing; this group made up more than 30% of renter households in Miami-Dade and more than 21% in Duval (Shimberg Ce nter for Housing Studies 2007). The sampling frame was the Assisted Housing Inventory (AHI), a property-level database of privately-owned rental propert ies in Florida that were funde d under federal, state or local housing programs. These housing programs impose inco me and/or rent restri ctions for all or a portion of the units, thereby offering affordable housing to lower income households. The AHI is a product of the Florida Housing Data Clearinghouse, which is part of the Shimberg Center for Housing Studies at the University of Florida. The AHI as the sampling frame was broader than

PAGE 90

90 the population of research intere st, because the AHI includes properties throughout the state of Florida and properties subsidized un der programs other than HUD Section 8. Data Collection and Database Design The aim was to create a property-level database (herein referred to as the database) that contained the variables that were required for the analysis. The database was created in Excel with one row for each property. The Assisted Ho using Inventory (Shimberg Center for Housing Studies 2008) was used as the starting point to identify the properties with HUD project-based rental assistance in Miami-Dade and Duval Count y. All variables that th e AHI reported on were initially incorporated into the database. These in cluded property characteristics such as address, number of units, bedroom configuration, target population, year built and housing programs. The AHI also reports on preservation-re lated variables such as type of ownership, Fair Market Rent, project rent to FMR ratio and expira tion dates of the subsidy programs. An advantage of the AHI was that it incorpor ated data from four major funding sources: HUD, the U.S. Department of Agriculture Ru ral Development, Florida Housing Finance Corporation and Local Housing Finance Author ities. As a result, the number and types of funding layers was known for each property in the AHI. Excluded from the AHI were detailed data about the HUD rental assistance contracts such as the contract effective date and the type of rental assistance program. These were essential variables for the research, because they provided in sight into the history of contract renewal and past owner decisions. Therefore, two datasets that were public ly available from HUD were merged with the database: The HUD Section 8 Contracts dataset and HUD Section 8 Properties dataset (HUD 2008a). Merging required the ma tching of properties by the HUD identification number, which is a unique ID that is assigned to each HUD-funded property.

PAGE 91

91 Other important information not captured by the AHI concerned the mortgage data for HUD insured multifamily mortgages. This inform ation was captured by the publicly available dataset called HUD Insured Mortgages (HUD 2008b), which provided details regarding the terms of the loan, the mortgage amount and the monthly payments. This information can add to the understanding of the financia l status of a property. The insured mortgages dataset was merged with the database. Another publicly available dataset that was important for the verification of year built data was called HUD Terminated Mortgages (HUD 2008c). Both the HUD Insured Mortgages and HUD Terminated Mortgage s datasets did not have a data field for HUD ID. Therefore, the merging with the databa se was more of a cumbersome process. It involved the matching of the proper ties to a third dataset that func tioned as a reference dataset, because it contained the HUD ID. This other da taset was an annual Excel report from HUD in Jacksonville (HUD 2008d), which listed HUD-funded properties in Florida. The properties were matched by HUD Project Number (a number that is assigned to each HUD loan) or by property name. The HUD ID was then added to the HUD in sured and terminated datasets. Then the variables in these datasets we re merged into the database by matching properties by HUD ID. HUD Physical Inspection Scor es are part of the AHI. However, at the time of the database design, a recent dataset of scores was released, which was not yet reported in the AHI. Therefore, the recent scores were merged in to the database, which was done through matching by HUD ID (HUD 2008e). The properties in the database were also matc hed to the tenant and unit characteristics as published in the HUD Picture of Subsidized Households for 2000. This is the most recent year for which tenant data were available.

PAGE 92

92 Once all the various datasets were merged, it was necessary to edit and clean the data, and to perform quality control. Each data field was examined by reviewing all cell contents. A pivot table of the database was created as an ai de for quality control. This enabled a quick view of the data entries in order to detect anomalie s and missing information. Several edits were made to the data fields and data entr ies. For example, the database in itially contained one data field with the names of all housing programs that fund ed each property. Since this format was not conducive to efficient analysis, the names of th e housing programs were parsed and placed in separate cells. As another example, discrepanc ies were found between AHI data and entries in the HUD datasets. For instance, for some propert ies the year built as re ported by the AHI was almost a decade later than the year of HUD mortgage endorsement and start of loan payments for the terminated mortgage. Subsequent research on those properties found th at the reported year built was actually the year of refinancing. The year built for those properties was therefore changed to the year of mortgage endorsement. Another finding during the quality control st age was that the bedr oom configuration and market rent data were not avai lable for units in properties that had less than 100% of the units covered by project-based rental assistance. This was the case fo r 20 properties in the population. To fill the data gap, each property was contac ted by telephone to collect data on the breakdown by bedroom size and on the rents by bedroom si ze for the non-rental assistance units. As part of the data editing process, an adjustment was made to the monthly loan payment amounts for properties funded under the HUD Secti on 236 program. According to the terms of this program, property owners receive an interest subsidy that reduces the mortgage payments by lowering the interest rate to 1%. However, during the data review it was found that HUD calculated the Section 236 mortgage payments based on a market interest rate, rather than the

PAGE 93

93 reduced rate. The actual debt se rvice was therefore overstated. Since the HUD data that were incorporated into the database included figures on loan amount a nd the term of the mortgage, the monthly mortgage payments were recalculated for all Section 236 proper ties in the population, based on a 1% interest rate. Table 5-1 lists all da tasets that were used to compose the database for the research. Fail-out Risk: Structure of the Mo del and Methodological Assumptions The Fail-out Risk Model was designed to a ddress the research que stion concerning failout risk. This model made the initial assumpti on that all properties, regardless of type of ownership, were at risk of failout. Therefore, monthly net income statements were simulated for all 119 properties. Then the assumption was made th at properties were at he ightened risk of failout if they met at least one of the following conditions: REAC Physical Inspection Sc ore of less than 60, or Median debt coverage ratio below 1.0, or Median Net Operating Income of less than $200 per unit per month. As found in the research, all three conditions were considered indicators of financial or physical challenges. The first condition was based on the notion that a REAC Physical Inspection Score below 60 is a failing score. Prop erties with a score under 60 are referred to the HUD Departmental Enforcement Center (Achte nberg et al. 2005). The second condition was included, because a debt coverage ratio below 1.0 indicates that a prop erty is experiencing difficulty covering the mortgage payments out of net income. The third condition was based on a benchmark adopted from a study completed for HUD by Abt Associates, Inc. (Finkel et al. 1999). The study included an analysis of the fina ncial condition of HUD-insured properties. The analysis compared the mean and median annual net cash flow after debt service per 2-bedroom unit. Properties were classified as negative cash flow, low positive cash flow between $0 and

PAGE 94

94 $500, high positive cash flow between $500 and $1,000, and very high positive cash flow over $1,000. For the purpose of the dissert ation research, these benchmarks were converted to per month figures before debt service, since the simulation results for NOI were per month and debt service was missing for the majority (76%) of properties in the population. The first step in converting the benchmarks was to add back th e mean debt service of $2,201 per year for HUDinsured properties, as was estimated by Abt (Finkel et al. 1999). The next st ep was to divide the figures by 12 to determine the benchmarks for th e monthly NOI, as outlined in Table 5-2. Lastly, the assumption was made that a monthly per uni t NOI of less than $200 was considered low. At least one of the three c onditions of fail-out risk was met by 32 properties or almost 27% of the population, as summarized in Table 53. These figures cannot be interpreted to mean that all these properties will be lo st to the assisted housing stock w ithin the next few years. These numbers were based on estimates of net cash flow and probability distributions. But they can be used to provide insight into the magnitude of prope rties that could be lost and that should receive attention from policy-makers and advocates. Th e analysis also inform s the housing community of the characteristics of higher risk properties. Opt-out Risk: Structure of the Mo del and Methodological Assumptions The Opt-out Risk Model was designed to address the research question related to opt-out risk. This model relied on the initial assumption that properties under non-profit ownership were not subject to this type of risk, because non-prof it entities have a mission to serve low income households. Therefore, monthly ne t income statements were simulated for properties with forprofit or limited-dividend ownershi p (both classified as for-profit for the purpose of analysis), a total of 83 properties. The simulation calcula ted two types of NOI values for for-profit properties: NOI values under the sc enario that the properties contin ued to operate status quo with the same number of assisted (and non-assisted) un its; and NOI values under the scenario that all

PAGE 95

95 properties converted all units to market rate rents. The percentage change in NOI values was calculated in order to assess wh ich properties would experience an increase in the bottom line and at what percentage. Next, th e assumption was made that propert ies were at heightened risk of opt-out if they met all of the following conditions: NOI increase of at least 20%, and Expiration of the rental assistance contract by December 31, 2014, and Original contract term, and Not located in a low poverty census tract. All these conditions were consid ered indicators of heightened risk of loss due to opt-out. An increase in NOI by at least 20% was established as an ar bitrary benchmark above which properties would be enticed to convert to market -rate housing. The contract expiration by year end 2014 was selected as a condition, because owners of these contracts have an option to optout in the short-term. Properties th at have a longer term contract in place were considered at a lower risk of imminent loss, because the c ontractual constraints extended beyond 2014. Original contract term was also one of the conditions, b ecause the owners have not yet had an option to opt-out of the contract. They could decide to op t-out once they have the first opportunity to make such a decision. This was the experience in th e late 1990s when the first Section 8 rental assistance contracts reached the e nd of the original contract term ; a wave of opt-outs took place (HUD 2007). The assumption was made that properti es that have had a chance to renew the rental assistance contract have not been interested in conversion. They would have opted out and converted to market rate housing when Floridas markets were booming in the first half of the current decade. Properties were al so classified at greater risk of opt-out and conversion to market-rate housing if they were located in a low poverty census tract where they could command market rents.

PAGE 96

96 The comparative cash flow approach (com paring status quo to 100% market units) was also taken in a recent paper that analyzed Low -Income Housing Tax Credit developments that were built during the early days of the program when the use restriction was only fifteen years (McClure and Grube 2007). The paper compared tw o scenarios of the operating performance for a tax credit development that was about to reach year fifteen of the us e restriction. The first scenario was resyndication of the tax credits and continuation as a tax cr edit property with use restrictions. The second s cenario was conversion to market rate housing. A total of 27 properties or almost 33% of the for-profit properties met all conditions (Table 5-4). As was mentioned under the description of the failout risk model, these figures should not be interpreted to imply that all these properties will be lost to the affordable housing inventory within several years. These numbers were based on es timates of net cash flow and probability distributions. But they can be used to estimate the scale of the preservation challenge and to act as guide for policy-makers and advocat es to determine the level and type of required resources to prevent loss of housing. Input Variables and Probability Distributions The input variables used to compose the net income statements had two types of values: Single-point or deterministic values and uncertain or probabilis tic values. Single-point values were based on actual data as obtained through prim ary or secondary data collection (e.g., number of units). Probabilistic values were assigned to random variables (e.g., oper ating expense ratios). All random variables were expressed as a ra nge of possible values and the likelihood of occurrence of each of these values as modeled by a probability distribution. All input variables and probability distributions that went into the simulation models are discussed in this section and summarized in Table 5-5.

PAGE 97

97 The first element of the net income statem ent was the gross potential rental income, which was the sum of the potential rents on all o ccupied and vacant units There were six input variables to calculate the gross potential rental income. The first two input variables were the number of units with project-based rental assistance by bedroo m size and the project rents for these units. The number of units by bedroom configuration was provide d by HUD and included in the database. The project rent was calculated as the product of the Fair Market Rent by bedroom size and the project rent to the FMR ratio for a propert y, as reported by HUD. Since the data on the units with project-based rental assistance we re derived from actual HUD figures, these input variables were included in the models as single-point estimates. The next two rent input variables were th e number of units with income and/or rent restrictions under other subsidy programs by be droom size and the corresponding rents. If the rental assistance contract did not cover 100% of the units a nd if another subsidy program imposed restrictions on the balance of the un its, the number of other restricted units was calculated as the difference between total units and rental assistan ce units. Rent data for the other restricted units were not avai lable through a public data source and were therefore collected directly at the property level. Property mana gement provided rents e ither as single-point estimates or as a range with a minimum and a maximum rent value. The lit erature reported that rental income was commonly assumed to follow a normal, uniform or triangular distribution (Schumann 2006; French and Gabrielli 2005; Kelliher and Mahoney 2000). But for the dissertation research, the uniform distribution was deemed most a ppropriate for units for which a range of restricted rents was provided, because no data were available on most likely or average rents and standard deviations.

PAGE 98

98 The last two rent input vari ables were the number of unrestr icted units and market rents. If the rental assistance contract did not cover 100% of the units and if no other subsidy program funded the property, the number of unrestricted units was calcul ated as the difference between total units and rental assistance units. Actual market rent data for each property were not available through a public data source. For the fail -out model as well as the status quo scenario under the opt-out model, actual market rents were collected directly from the property. Rents by bedroom size were either provided as single-point estimates or as a range with a minimum and a maximum rent. A uniform probability distribution wa s selected if a range of rents was available. For the market-rate scenario unde r the opt-out model, the assump tion was made that all units were rented at market. Government data sources for market rent (e.g., HUD, U.S. Census) have limitations in terms of the geographical scale (e. g., only county-level market rents or only market rents for a limited number of counties) and th e time period (e.g., only market rents for several years ago). Private data sources may be available from appraisers or property management firms, but these data are often restricted to a small number of geographies, are kept for internal purposes only, or cannot be obtai ned through purchase. Due to these limitations, an alternate source of market rent informa tion was relied upon, Zilpy.com. Z ilpy is described as a free online rental market facts and analysis service dedicated to help you make better rental and investment decisions (Zilpy LLC 2009). It collects rental data from a variety of sources, including newspaper classified ads, online classi fied ads and apartment rentals. Zilpy reported median rents by bedroom size, the number of rentals by bedroom size and average square footage by bedroom size for a specific address, zip code or city. For each query, Zilpy also provided a list of the top 100 mo st recent rentals with address details, unit configuration and building structure type. For each property in the opt-out model, Zilpy was used to find median

PAGE 99

99 rents by zip code by bedroom size. The simulation software was then used to fit a probability distribution to the observed data for each be droom size by county. The lognormal distribution was found to be the best fit. In calculation of the gross pot ential rental income, other re ntal income was excluded due to a lack of data. Other income may include commercial rent, interest income from reserve accounts or forfeited tenant deposits (Wallace et al. 1993). Another input variable was vacancy loss and bad debt allowance, which was calculated as a percentage of the gross potential rental income. This was an un certain input variable for which no actual property-level data were publicly available. Hence, the models made the assumption that this input variable was uniformly distri buted between 5-10%. Similar to the comparative analysis conducted by McClure and Grube (2007), the assumption was the same for the status quo scenario and market-rate scenario under the opt-out model. The uniform distribution was selected, because it was applied in other simu lation research (Hoesli, Jani, and Bender 2006; Baroni, Barthlmy, and Mokrane 2005) and beca use it models Wilkins suggestion that the allowance typically be 7-9% and not lower than 5% (2002). The range of 5-10% also considered the industry standard of 7%. A ccording to Stan Fitterman at the Florida Housing Coalition (2007), a vacancy rate of 7% was the benchm ark. Achtenberg (LISC 2005, 10) also explained that the minimum vacancy and bad debt allowa nce required for first mortgage underwriting is generally 7% for properties that were rest ructured under the HUD Mark-to-Market program. Slade (2006) used a triangular distribution to es timate vacancy rate, because he was modeling only one property and was able to make an assumption about the most likely vacancy rate; such an assumption was not deemed appropriate in this study.

PAGE 100

100 The gross potential rental income less the vacancy loss and bad debt allowance equaled the effective gross income. Operating expenses, another input variable, were calculated as a percentage of effective gross income. None of the public data sources contained actual data on the operating expenses at the property level, beca use these data are considered proprietary. In lieu, operating expenses were estimated based on data from the Institute of Real Estate Management (IREM). Hecht (2006) in his book Developing Affordable Housing, A Practical Guide for Nonprofit Organizations suggested to rely on IREM data to project and compare operating costs if historical info rmation is not available. IREM data are commonly used in real estate analysis to estimate operating expenses (Goodman 2004; Finkel et al. 1999; Bogdon and Follain 1996). As explained by IREM, the data can serve as inputs for feasibility studies and as benchmarks for comparison purposes (IREM 2007) IREM publishes an income and expense analysis report annually for various types of real estate, including federa lly assisted apartments funded under subsidy programs such as Secti on 236 and Section 8 project-based rental assistance. The report contains operating ratio data which were used as the input variables in the risk models. Operating expenses divided by act ual collections equaled the operating ratio. The IREM report presented the opera ting ratio data in two ways: Operating ratio by property age group and s ubsidy type. The propert ies in this study fall within two of the property age groups, 1965 to 1977 and 1978 to date. Operating ratio by region and subsidy types. Fl orida is part of Region 4, which consists of Kentucky, Tennessee, North Carolina, South Ca rolina, Mississippi, Al abama, Georgia and Florida. Since operating expenses are random with uncertain values, the simulation analysis considered both sets of operating ratios. A dditionally, the operating ratios were obtained for three consecutive years, rather than for one year, to take into account year over year fluctuations in operating expenses (Harvard University Gr aduate School of Design 2001; Finkel et al. 1999)

PAGE 101

101 and in variations in the IREM sample base (IREM 2007). From among both sets of operating ratios over three years, the minimum and maximu m values were recorded by subsidy type and age group. Depending on a propertys funding layer a nd year built, the simulation analysis drew from the range of operating ratio values and a ssumed that these follow a uniform probability distribution. While other studies applied a triangular distribution to operating expenses (Hoesli, Jani, and Bender 2006; Kelliher and Mahoney 2000), th is was not considered a suitable approach in this research due to the l ack of property-specific information about operating expenses and most likely values. The assumpti ons pertaining to the operating e xpense estimates were the same for each model and all scenarios in this study, which was similar to the method use by McClure and Grube (2007) in their cash flow comparison of the status quo scenario to the market-rate scenario of a tax credit property. Replacement reserves were another cost item and random input variable. Replacement reserves were estimated to follow a triangular distribution with a minimum value of $0, a most likely value of $26.50 and a maximum value of $62.50 per unit per month. These estimates were derived from an analysis of capital needs asse ssments and replacement reserve studies conducted by On-site Insight, Inc. during 1999 and 2000, which are the most recent studies. The analysis was conducted for a sample of 183 assisted pr operties funded by HUD or state housing finance agencies and located in eighteen different states On-site Insight, Inc. (2001) reported a median annual contribution to the replacement rese rve account of $26.50 per unit per month. The company published a histogram with annual replacement reserve contribution categories by number of properties. The cat egories ranged from $0-99 to $2, 100-2,199. In order to estimate a probability distribution for the simulation analysis a frequency table was created with the data from the histogram. The simulation software was then used to fit probability distributions to the

PAGE 102

102 data. The triangular distribution provided the best fit. The minimum value was set to $0 and the most likely value was determined to be $26.50 (the median replacement contribution). The maximum value was established at $62.50 per unit per month, even though the histogram displayed observations with higher reserves. This maximum was considered appropriate for two main reasons. First, from $0 to $62.50 in annual re placement reserves cove red almost 90% of all observations of the On-site Insight data. Second, at this maximum value, the mean value is $29.67, which approximates the benchmarks used within the industry. Kh adurri and Wilkins (2007) assumed a replacement reserve of $25 per unit per month in simulation analysis of a tax credit property; Finkel et al (1999) calculated average repl acement reserve deposits at $27.58 per unit per month for HUD-assisted properties, compared to median re serves at $20.33 per unit per month; and LISC (2005) discussed a ru le of thumb of $30 pe r unit per month. The assumptions for replacement reserves were the sa me for each model and a ll scenarios, which was also assumed in the comparative analysis by McClure and Grube (2007). Debt service was the last input variable Data on monthly mortgage payments were publicly available for propertie s funded with a HUD mortgage. Th ese data were incorporated into the database and used to establish singlepoint estimates of the debt service for those properties. For all other properties, no debt se rvice information was available. No inferences were made about the monthly mortgage payments for those other properties due to the high level of uncertainty concerning this random variable. The following input variables were the same for the fail-out risk model and the opt-out risk model with its two scenar ios: Vacancy loss and bad debt allowance, operating expense ratios, and replacement reserve. All three input variables were modeled only once. The cash flow calculation for every property under each model li nked to the same input cells for these random

PAGE 103

103 variables. This was required in order to pr operly compare the outputs among all properties. If these three input variables were simulated for each property cash flow, the same variable (e.g., replacement reserve) would have a different va lue for every property in each single iteration (Vose 1996). Instead, the random values of these three input variables should be the same for all properties in each iteration. As explained by Vose (1996, 96), every uncertain variable must be represented once only in the spreadsheet and an y other cell that needs its value must make reference to the cell in whic h the distribution resides. Output Variables Each of the models had two output variable s, net operating income and debt coverage ratio. Effective gross income less the operating expe nses and replacement re serves totaled the net operating income, which was estimated for each pr operty. Net operating income is a measure of financial condition (Wallace et al. 1993). Debt coverage ratio was calcula ted as the net operating income divided by debt service. It was considered an indicator of financial healt h, because it measured if the net cash flow was sufficient to cover the mortgage payments (Bradl ey, Cutts, and Follain 2 001). Since debt service information was not available for all properties, the debt coverage ratio could only be estimated for a subset of properties. Data Analysis and Analytical Tools Descriptive Analysis and Significance Tests A simulation was run with 1,000 iterations to estimate the values of four categories of output variables: Net operating inco me at current rents; debt covera ge ratio at curre nt rents; net operating income at market rents; and de bt coverage ratio at market rents. For each property, the simulation analysis cal culated the NOI based on the current project rents. For those properties for which mortgage data were available, the s imulation also generated

PAGE 104

104 values for the debt coverage ratio based on curr ent project rents. The results of the simulation that was run with current project rents were used for the analysis of the fail-out risk. The results for the properties with for-profit ownership were al so used in the assessment of the opt-out risk. The simulation also calculated the NOI for fo r-profit properties with all market rents as the input variables. Additionally, the simulation estimated the debt coverage ratio under the scenario of market rents for for-profit properties for which mortgage payment information was available. The outcomes of the simulation ba sed on market rents we re compared to the simulation results based on current rents in order to analy ze the opt-out risk. The simulation results were summarized in ta bles and described for each risk model and by output variable. In addition to descriptive analysis, significan ce tests were performed for the opt-out risk model to compare th e mean NOI values and to analyze the change in NOI if market rents are charged rather than contract rents. For each risk model, the characteristics of th e properties classified at a higher risk were compared to those of properties that were identif ied at lower risk. Four types of characteristics were described and summarized in cross-tabulat ions: Property, financial, subsidy and tenant characteristics. Where the difference in charact eristics between the highe r risk and lower risk groups seemed notable, a significance test wa s performed to assess if the difference was statistically significant. If the data that described the characteris tic were quantitative, the two groups were compared using a t-test of the null hypothesis that the means of the groups were the same (H0: 1 = 2; Ha: 1 2). The t-test was selected as the significance test, because it is considered a robust statistical method. As explained by Agresti and Finlay (1997, 187), even if the population is not normally distributed, two-side d tests and confidence intervals based on the t

PAGE 105

105 distribution still work quite well. The P -values and confidence coefficients are fairly accurate, the accuracy being quite good when n exceeds about 15. If a characteristic was described by qual itative data, the population proportions were compared by performing a z-test of the null hypothesis that the proportions were equal (H0: 1 = 2; Ha: 1 2). Agrasti and Finlay (1997) suggested th e z-test to compare population proportions if each category in each group contai ned more than five observations. The significance level for all significance test s was set at 0.05 (5%). The P-value that was calculated by each test was compared to this si gnificance level; a null hypot hesis was rejected if a P-value was less than or equal to 0.05. Correlation and Multiple Regression Analysis Since NOI was considered a major indicator of the financial condition of a property, multiple regression analysis was conducted to as sess which characteristics caused a lower NOI. The estimated mean NOI was the response (dep endent) variable. The following explanatory (independent) variables were se lected from the property and s ubsidy characteristics that were studied: Total number of units Target population (1=elderly; 0=family) Average unit size by number of bedrooms Year built Type of ownership (1=for-profit; 0=non-profit) REAC Physical Inspection Score Project rent to Fair Market Rent ratio Number of program layers Contract effective year Contract expiration year Original contract (1=not original contract; 0=original contract) Prior to running the multiple regression anal ysis, each explanatory variable was regressed on the others to calculate the co rrelation between each pa ir of variables. Correlation statistics

PAGE 106

106 were summarized in a correlation matrix and were analyzed to test for multicollinearity. The condition of multicollinearity exists when explan atory variables are highly correlated. A variable that is highly correlated becomes redundant in the regression model, because it has limited unique explanatory power. The effects of multicollinearity are problematic and can cause inflated standard errors, overly large P-values and wrong signs of the regression coefficients (Albright, Winston, and Zappe 2006; Agresti and Finlay 1997). Variables that were highly correlated were excluded from further analysis. A multiple regression was run with the remaining explanator y variables. A stepwise regression was also performed. The stepwise procedure adds one explan atory variable to the model at a time. At each step, the variable that contribute s most to R square is selected to be added. After a variable is added, the stepwise method also removes any variables that are no longer significant. The steps are repeated until no more variables can be adde d to improve R square (Agresti and Finlay 1997). Scenario Analysis For lack of a public data source with detailed and current market rents, the input values for the market scenario under the opt-out model were derived from the website Zilpy.com. The estimated market rents seemed reasonable, based on a comparison to the Fair Market Rent data. But since the reliability of Z ilpy.com was unknown, the market sc enario under the opt-out model was rerun with the 2009 Fair Market Rents as the input values for all units. The change in mean NOI was recalculated, analyzed and compared to the outcomes of the simulation that was based on the market rent data from Zilpy. Analytical Tools The cross-tabulations for the descriptive statistics and the significance tests were conducted in Microsoft Excel 2003. The analytical tool used to perf orm the t-test was the t-Test:

PAGE 107

107 Two-Sample Assuming Unequal Variance in Ex cel. The correlation matrix and the multiple regression analysis (includi ng the stepwise regression) were prepared in SPSS 16.0. The simulation was performed with @RISK Ve rsion 5.0, which is a software system for risk analysis and Monte Carlo simulation. It is an add-in for Microsoft Excel. @RISK is produced by Palisade Corporation. Limitations of Net Cash Flow Approach and Simulation Modeling The net cash flow approach that was taken in this research had seve ral limitations. One of the limitations was that random variables had to be used as input data. For most random variables no actual historical or current property-level data we re available on which to base assumptions for the range of input values and pr obability distributions. This meant that if the quality of the inputs was low, the outputs w ould not be realistic (Li 2000; Vose 1996). To optimize the quality of the outputs, the uncertain inputs were ther efore carefully researched in order to make realistic assumptions that could be justified. Another limitation of the net cash flow appr oach was that this did not consider other major indicators of a propertys viability. Wallace et al. (1993, 2-25) warned that in assessing a propertys viability, net cash flow must be exam ined concurrently with physical needs and property management. A property could have deceptively positive cash flow by failing to make necessary expenditures for repa irs and replacements. Convers ely, a property could have deceptively negative cash flow because a new owner or manager has begun a crash repair program to eliminate an accumulated backlog of physical needs. The fail-out model incorporated the REAC Physical Inspection Scores in lieu of publicly available data on capital needs, repairs and maintenance. The research did not consider any property management variables, due to a lack of information.

PAGE 108

108 Type of ownership was a defining factor in the opt-out model, which only included properties with for-profit ownership. However, ow nership type is only a general proxy of owner intent. Even though research showed that non-profit owners were less likely to opt-out compared to for-profit entities (Fin kel et al. 2006), it would be possible for non-profit entities to decide to opt-out of a subsidy. Conversely, for-profits ma y have a double bottom line and want to continue to serve low-income households, rath er than converting to market-rate housing. An owners decision can be impacted by othe r factors that were not accounted for in the analysis. This may have resulted in an overstatement of the estimate of units at risk. It is clearly not possible to precisely predict an owners decision. [A]rmchair assessment can never be highly reliable as a predictor of owner behavior. Interacting with the owner and the property is, ultimately, the way to gauge the operative dynamic s of the owners decision making. Even then, predicting actual outcomes is difficult (Recapitalization Advisors, Inc. 2002, 36).

PAGE 109

109 Figure 5-1. Probability Distributions. Source: Palisade Corporation (2008).

PAGE 110

110 Table 5-1. Datasets Used to Create the Database Dataset Source Data as of Date Assisted Housing Inventory Shimberg Center for Housing Studies, University of Florida September 2008 Multifamily Assistance and Section 8 Contracts Database HUD DC September 2008 Multifamily Assistance and Section 8 Properties Database HUD DC September 2008 Insured Multifamily Mortgages Database HUD DC June 2008 Terminated Multifamily Mortgages Database HUD DC June 2008 REAC Physical Inspection Scores HUD DC September 2008 Annual Database for Florida HUD Jacksonville January 2008 Picture of Subsidized Households HUD DC 2000 Source: Shimberg Center for Housing Studies ( 2008b); U.S. Department of Housing and Urban Development (2008abcde); U.S. Department of Housing and Urban Development (2006). Table 4-2. Net Operating Income Benchmarks Benchmark and Amount Benchmarks of Annual Net Cash Flow after Debt Service per 2-Bedroom Unit: Low positive cash flow $0-500 High positive cash flow $500-1,000 Very high positive cash flow >$1,000 Annual Mean Debt Service for Total HUD-Insured Properties per Property: $2,201 Benchmarks of Annual Net Cash Flow before Debt Service: Low positive cash flow $2,200-2,700 High positive cash flow $2,700-3,200 Very high positive cash flow >$3,200 Benchmarks of Monthly Net Cash Fl ow before Debt Service per Unit: Low positive cash flow $183-225 High positive cash flow $225-266 Very high positive cash flow >$266 Research Assumption for Low Benchmark of Monthly Net Operating Income per Unit: <$200 Source: Adapted from Finkel et al. (1999). Table 5-3. Number of Properties by Condition for Higher Fail-out Risk Condition Properties Mean NOI <$200 20 Mean DCR <1.0 1 REAC < 60 8 Mean NOI <$200 AND REAC < 60 2 Mean DCR <1.0 AND REAC < 60 1 Total properties that meet at least one condition 32

PAGE 111

111 Table 5-4. Number of Properties by Condition for Higher Opt-out Risk Condition Properties For-profit ownership 83 Not in low poverty census tract 74 Contract expiration by December 31, 2014 64 Mean NOI at least 20% higher 59 Original contract term 35 Total properties that meet all conditions 27 Table 5-5. Input Variables, Dist ributions and Parameters for th e Fail-out and Opt-out Risk Models Input Variable Dist ribution Parameters Number of rental assistance units Single-point Property specific Project rent (Fair Market Rent multiplied by project rent to FMR ratio) Single-point Property specific Number of other restricted units Single-point Property specific Rent for other restricted units Singl e-point, uniform Property specific Number of unrestricted units Single-point Property specific Market rent for unrestricted units 1) Fail-out Model and Opt-out Model/Assisted Scenario: Single-point, uniform 2) Opt-out Model/Market Scenario: lognormal 1) Fail-out Model and Opt-out Model/Assisted Scenario: Property specific 2) Opt-out Model/Market Scenario: Specific to county and bedroom size Vacancy loss and bad debt allowance Uniform Minimum 5%, maximum 10% Operating Expenses Uniform Sp ecific to subsidy type and age group Reserves Triangular Minimum $0, most likely $26.50, maximum $62.50 Debt Service Single-po int Property specific

PAGE 112

112 CHAPTER 6 DATA ANALYSIS Input Variables and Simulation R esults: Descriptive Analysis A total of 41 input variables were si mulated according to selected probability distributions. The table in Appendi x 1 lists all input variables and the values that were generated in the 1,000 iterations performed by the simulation analysis. The ta ble also notes the type of probability distribution and displays distribution graphs for all variables. Five categories of input variables were simulated: Vacancy loss and bad debt allowance; operating expense ratios; replacement reserve; current non-Sec tion 8 rents; and market rents. Vacancy loss and bad debt allowance was an input variable used in all models. A uniform probability distribution was assumed for this rando m variable. This distribution required that a minimum and maximum value were established. The simulation calculated a mean value of 7.5%. The percentiles show that there was an 80% probability that the vacancy loss and bad debt allowance is at least 6%; the probability was 20% that this variable amoun ted to more than 9%. The operating expense ratios were a compone nt of all models. The operating expense ratios were estimated by a uniform probability di stribution rather than a triangular, normal or lognormal distribution, due to the lack of inform ation that would provide insight into actual distribution patterns and most likely values. Five types of operating expense ratios were simulated. Each ratio was specific to a HUD subs idy program and year of construction. The minimum and maximum values for each type were determined based on a range of ratios calculated by the Institute of Real Estate Management for 2004 to 2006 (IREM 2007). The simulation calculated mean values that rang ed from 55 to 62% for Section 8 properties, depending on target population and year built. For properties that also had a Section 236 mortgage, the mean operating expense ratio was estimated at 74%.

PAGE 113

113 The replacement reserve was an input variable for each model. A triangular distribution was selected to compute the replacement rese rve. The minimum, most likely and maximum values were based on the findings of a study of replacement reserves by On-site Insight, Inc. (2001). The simulation analysis calculated a mean replacement reserve of $29.67 per unit per month. According to the percentiles, the probabilit y was 75% that the replacement reserve is at least $20.33 per unit per month. While the maximum value was set at $62.50, the results of the simulation show that the probability was 95% that the reserve is lower than $52.00. For properties that contained units not c overed by a HUD rental assistance contract, rental data were collected in the form of singl e-point estimates or range s of rents with minimum and maximum values. Non-Section 8 rents were either market rents or rents restricted under the HUD Section 236 program. Where a range of rent s was collected, a uniform distribution was assumed and the rental data were simulated at the property level for a total of 24 properties. The table in Appendix 1 summarizes the simulation resu lts for those properties such as the mean rent. The simulation of these rental data was performed as part of the fail-out model and the assisted scenario under th e opt-out model. The market scenario under the opt-out mode l assumed that all units were leased at a market rent. Therefore, market re ntal data were the input variable s that were estimated for each county by bedroom configuration. Market data were collected for all the zip codes of the properties in the population. A lognor mal distribution was found to be the best fit to the market data. While this type of distribution is unbounded with an infinite minimum and maximum value, the simulation results showed that none of the minimum values were unreasonably low considering the unit type (lowest minimum va lue $427 for studio apartment in Duval); the maximum values were relatively high for the largest unit types in Miami ($8,395 for four

PAGE 114

114 bedroom apartment), but the probability was less th an 5% that the market rents would reach such levels. NOI and DCR Output Analysis Descriptive Analysis and Significance Tests The NOI simulation results for the entire popul ation are reported at the property-level in the table in Appendix 2 and summarized in Tabl e 6-1. The estimated mean NOI ranged from $84 to $403 per unit per month, as graphed by Figure 6-1. The average of a ll 119 estimated mean NOI values amounted to $242; the median was $245. Table 6-2 compared these mean and median values to the mean and median NOI va lues that were reported for non-market rent properties in the metropolitan ar eas of Jacksonville and Miami (Urban Land Institute 2006). The mean and median NOI in these two areas ranged from $176 to $385. Close to 19% of the 119 properties had an estimated mean NOI of less than $200 per unit per month; almost half of these had a mean NOI below $150, but not less than $84. More than 66% of the properties had a mean NOI between $200 and $299 and 15% had a mean NOI greater than $300. Table 6-3 summarized the number and percentage of properties with an estimated NOI of below $200 by probability. For more than 68% of the properties the minimum NOI value that was estimated by the simulation was less than $ 200 and as low as $21. However, the percentiles showed that the probability was very small (l ess than 5%) that such a large proportion of properties had an NOI below this benchmark. At the 25th percentile, almost 32% of properties had an NOI less than $200. This implies that th e probability was 75% that these properties had an NOI that exceeded the benchmark. Both the m ean and median NOI calculation showed that almost 19% of the properties were below $200. Fo r roughly 14% of the properties, there is a 75% probability that the NOI is below the benchmark and a 25% probability that the NOI will

PAGE 115

115 rise above $200. For 10% of the properties, the probability is 95% that the NOI will be relatively low; a 5% probability exists that the NOI will reach at least $200. The simulation analysis also estimated the de bt coverage ratio, as outlined in Table 6-4. However, mortgage data were available for onl y 28 properties, which to taled less than 24% of the population. The majority of those properties had a mean DCR of at least 1.0; only two properties had a DCR less than 1.0, but not lower than 0.83. The mean DCR was less than 1.2 but greater than 1.0 for two properties. The calcu lation of maximum debt coverage ratios showed an outcome of 0.97 for one property and 1.05 for a nother property. The simulation estimates for the minimum DCR value resulted in a larger pro portion of properties with a low debt coverage ratio; more than 39% of properties with debt serv ice information (total of eleven properties) had a minimum DCR between 0.60 and 0.99. But the percen tiles indicated that the probability was small that all eleven properties had a DCR below 1.0. For example, at the 5th percentile, only three properties have a DCR less than 1.0. This m eans that the probability was 95% that these properties had a DCR of at least 1.0. Under the opt-out model, the simulation calc ulated the NOI values based on market rents for all units in properties with for-profit ownership. The simulation results were compared to the NOI values generated under the scenario of cu rrent project rents. Ta ble 6-5 summarized the output values under each rent scen ario. Under the assisted scenari o, the majority of the properties had a mean NOI between $149 and $402. In comparis on, the majority of the properties under the market scenario had a mean NOI that range d from $190 to $529. A t-test was performed to compare the mean NOI values under both scenarios. The conclusion of the t-test was that the mean NOI values were statistically different (P -value of < 0.000), as presented in Table 6-6;

PAGE 116

116 properties that charged market re nts for all units (as opposed to re stricted rents) yielded a higher net income. To assess the impact on the bottom line of conve rting project rents to all market rents, the percentage change in estimated NOI values wa s calculated, as displayed in Table 6-7 and graphed in Figure 6-2. The mean NOI improved for more than 95% of the properties. For almost quarter of the properties, the mean NOI increased by up to 19.9%. Forty percent of properties experienced a rise in mean NOI between 20 and 39.9%, and for 30% of properties the estimated mean net income went up by 40 to 59.9%. Accordi ng to the minimum values, there was a very small probability (less than 5%) for almost 41% of properties that the NOI dropped when project rents were calculated at to market rents. But le ss than 11% of properties experienced a decline in NOI at the 25th percentile, which means that the probability was 25% that the NOI was lower if all market rents were charged. A drop in NOI would be possible for two reas ons. First, some properties received project rents that were higher than the Fair Market Re nt. Depending on the projec t rent to FMR ratio, a conversion to market rents can result in a sma ller NOI. Second, for three of the bedroom sizes in Duval, the mean estimated market rents were lo wer than the Fair Market Rent. If the project rents for a property approximated FMR, the NOI can drop when revenues were calculated based on the simulated market rents. Properties with a percentage change in mean NOI of at least 20% were considered at higher risk of conversion. According to the st atistics presented in Table 6-8, roughly 71% of properties had an increase in mean NOI of at least 20%. For 60% of properties the probability was 50% that the NOI improved by at least 20% For almost half of the properties, the probability was 75% that the NOI reached at least the 20% benchmark. The maximum NOI

PAGE 117

117 values calculated by the simulation estimated that all properties experienced an increase in NOI of at least 32.8%. However, the prob ability of this was less than 5%. The simulation could only estimate the debt coverage ratio for about 15% or 13 of the for-profit properties (Table 6-9) Under the assisted scenario, the mean DCR was below 1.0 for one property and was at a low of value betw een 1.0 and 1.19 for two properties. Under the market scenario, none of the properties had a mean DCR below 1.0, although the ratio was estimated at only 1.01 for one property. Correlation and Multiple Regression Analysis Multiple regression analysis was conducted to gain insight into the variables that impact net operating income. But first, a correlation matrix was generated to assess the correlation between each set of the follo wing explanatory variables: Total number of units Target population (1=elderly; 0=family) Average unit size by number of bedrooms Year built Type of ownership (1=for-profit; 0=non-profit) REAC Physical Inspection Score Project rent to Fair Market Rent ratio Number of program layers Contract effective year Contract expiration year Original contract (1=not original contract; 0=original contract) Analysis of the correlation statistics resu lted in the exclusion of three explanatory variables from the regression analysis due to higher levels of correlation: Bedroom configuration, contract effective year and original contra ct. Bedroom configuration showed a very strong negative correlation of -0.803 with target population, im plying a negative association between the average unit size by number of bedrooms and elderl y as the target population. The bedroom variable was therefore removed from furt her analysis. The contract effective year was

PAGE 118

118 also excluded from the regression analysis, because it was highly correlated with the original contract variable at 0.950. To a weaker degree, it was also co rrelated with total units at 0.529, number of program layers at 0.432 and year built at -0.412. The third explanatory variable that was not included in further analysis was the or iginal contract variab le, because it had a correlation of 0.524 with total units and -0.492 with year built. A revised correlation matrix was prepared th at excluded the three explanatory variables that had a relatively high co rrelation. The recalculated statis tics indicated that no strong correlations were present; none of the correlations exceeded 0.379 (Table 6-10). The first regression model included all eight explanatory variables. The results reported a coefficient of determination (R Square) of 0.513. In other words, 51.3% of the variation in the mean NOI was explained by this model. Howe ver, only two of the eight variables were calculated to significantly affect the NOI: Project rent to FMR and year built. Total units, target population, ownership type, REAC sc ore, number of program layers and contract expiration year each had a relatively large P-va lue and small t-value, indicati ng no statistical significance. To improve the regression model, a stepwi se regression was completed with all eight explanatory variables. While six of the explanatory variables pr oved insignificant in the first regression, blindly excluding all th ese variables from a rerun of the regression analysis would not be appropriate. As explained by Albright, Winston, and Zappe (2006, 654), it is possible that when one of these variables is excluded, another one of them will become significant. The stepwise procedure added one expl anatory variable to the model at a time. At each step, the variable that contributes most to R square is selected to be ad ded. After a variable is added, the stepwise method also removes any variables that are no longer signi ficant. The steps are repeated until no more variables can be added to improve R square (Agresti and Finlay 1997). Table 6-11

PAGE 119

119 revealed the outcome of the stepwise regressi on, reporting two variables that were significant: Project rent to FMR and year built. Compared to the first regression model, the R square dropped to 0.467 from 0.513. One of the statistical properties of R square is that it only increases when explanatory variables are added, hence the larger value in th e first regression model. An alternative measure is the adjusted R square which adjusts R square for the number of explanatory variables in the equa tion. It is used primarily to monitor whether extra explanatory variables really belong in the equation (Albri ght, Winston, and Zappe 2006, 590). The adjusted R square that resulted from the stepwise re gression was 0.458, which was only slightly lower than the adjusted R square value of 0.477 as measured by the first regression model. In other words, 45.8% of the variation in the mean NOI was explained by only two explanatory variables. The estimated regression coefficients of the st epwise model can be interpreted as follows: The predicted mean operating income will increase $221.88 when the project rent to FMR ratio is increased by 1%, keeping all other variables constant. The predicted mean operating income will increase $4.33 when the year built is increased by one year, keeping all ot her variables constant. Fail-out Model: Descriptive Analysis and Significance Tests Almost 27% of the properties (32 properties) were identified at hi gher risk of fail-out; 73% (87 properties) were considered at lower risk. Property characteristics seemed to differ between the higher and lower risk groups (Table 6-12). The majority of the properties in the higher risk group had either less than 50 units (2 5%) or more than 200 units (31%); more than half of the lower risk properties contained less than 100 units. On average, properties at higher risk had 124 units, compared to 99 units fo r the lower risk group. A significance test was performed to compare the groups. The P-value of the two-tail test was 0.123 (Table 6-13), which was greater than the significance level of 0.05. Ther efore, the conclusion was made that there is no statistical difference in property size betw een the higher and lower risk properties.

PAGE 120

120 The properties with higher fail-out risk seem ed to contain smaller units by bedroom size. Seventy percent of units were studio or one-bed room apartments; this figure amounted to 45% for the group of lower risk properties. The aver age number of bedrooms for the higher risk group was 1.1, compared to 1.5 for the lower risk properti es. A t-test was run to examine the statistical difference in bedroom configura tion. The P-value of the two-ta il test was 0.023 (Table 6-14). Therefore, the null hypothesis was rejected and the conclusion was made that there was a statistical difference in the mean unit size; properti es classified at higher fail-out risk had smaller units as measured by average number of bedrooms per unit. The target population for the higher risk prop erties was fairly evenly distributed between family (44%) and elderly (56%). Almost 60% of the properties at lower risk targeted families. A significance test to compare the pr oportion of properties serving th e elderly returned a P-value at 0.052 (Table 6-15), which meant no significant difference in target population at the 0.05 level. The type of ownership for properties at higher risk of fail-out wa s evenly distributed between non-profits and for-prof its. For properties at lower ris k, 77% had for-profit ownership. The proportion of properties owned by non-profits was compared by applying a significance test. The P-value was calculated at 0.002 (Table 6-16), which implied that the null-hypothesis was rejected and that the difference in ownership t ype between higher and low risk properties was significant. On average, properties in both groups were built in the 1970s; properties at higher risk had an average year built of 1974 compared to 1 979 for those at lower risk. But the proportion of properties built by time period differed between th e groups. More almost 69% of higher risk properties were constructed during 1970 to 1974 and almost 22% were erected between 1980 and 1984. In contrast, more than 56% of properties at lower risk were built during 1980 to 1984;

PAGE 121

121 nearly 22% had a year of construction between 1970 and 1974. A t-test to compare the year built between the groups reported a Pvalue < 0.000 (Table 6-17). Ther efore, the null hypothesis was rejected and the conclusion was dr awn that the year bui lt was statistically different; properties at higher risk of fail-out ha d an earlier year built. The descriptive statistics showed that each risk group contained more than one third of properties with a REAC Physical Inspection Scor e of at least 90. But th e proportion of properties for the score categories below 90 was different be tween the groups. All lower risk properties had a passing score of at least 60, while more than 34 % of higher risk properties had a failing score below 60. This was the result of the approach that was taken to gene rate the list of properties at heightened risk of fail-out; a property was identif ied at risk if the REAC score fell below 60. A significance test confirmed the statistical differe nce between the risk groups; the P-value was 0.010 (Table 6-18). The financial characteristics as summarized in Table 6-19 reflected the methodology of flagging properties at higher fa il-out risk. Properties with a mean net operating income below $200 per unit per month or a mean debt coverage ra tio below 1.0 were determined to be at higher risk. More than 31% of higher risk properties had an estimated mean NOI between $84 and $149; for almost 38% the mean NOI ranged between $150 and $199. Almost one third of the properties in the higher risk gr oup were yielding a mean NOI grea ter than $200; th ese properties met either one of the other two conditions for inclusion in the higher risk group (REAC < 60 or DCR < 1.0). Almost 82% of the properties at lower fail-out risk had a mean NOI between $200 and $299.

PAGE 122

122 The debt coverage ratio c ould only be calculated for 43% of the higher risk properties and less than 16% of the lower risk properties. Only two properties were estimated to have a DCR below 1.0. One of these also had a REAC score below 60. The subsidy characteristics (Table 6-20) show ed that for almost 91% of the properties in each risk group, the project rent to FMR ratio was below 100%. The proportion of properties with a rent to FMR below 80% was relatively la rge for properties at higher fail-out risk; it amounted to almost 66%, which contrasted to 31% for the lower risk group. On average the project rent to FMR was 78.7% for the higher ri sk group and 85.8% for the lower risk group. The significance test found that the pr oject rent to FMR was statis tically different between the groups; the P-value was calcu lated at 0.008 (Table 6-21). The Loan Management Set Aside and Secti on 8 Substantial Rehabilitation programs were the largest categories of rental assistance programs for both hi gher and lower risk properties. More than 64% of properties at higher risk ha d a contract under the LMSA program and more than 18% were funded under Section 8 SR. Section 8 SR was the largest category for properties at lower risk (33%), followed by the LMSA program (almost 30%) and Section 8 New Construction (28%). A significance test was r un to analyze the difference between the risk groups in the proportion of properties with a LMSA contract At a P-value < 0.000, the test concluded that the difference was statistically significant; propertie s identified at higher risk had a larger proportion of contracts unde r the LMSA program (Table 6-22). More than three quarters of the properties at higher risk no longer had a rental assistance contract in place under the original term; the contra ct was renewed at least once. In contrast, the lower risk group was fairly evenly divided between original contracts (more than 45%) and

PAGE 123

123 renewed contracts (more than 54%). A statistica lly significant difference was detected between the groups in the contract renewal hi story (P-value at 0.017, Table 6-23). The difference in current contra ct effective year was also found to be statical ly significant between the risk groups (P-value at 0.029, Table 6-24). More than 75% of properties at higher risk of fail-out had a contract effective year between 2000 and 2008; the contract effective year ranged between 1980 and 1989 for more than 18% of the higher risk properties. For lower risk properties, nearly 40% of the current contracts originated between 1980 and 1989 and more than 53% had a contract effective year of 2000 or later. The current contract ef fective year is related to the contract renewal history; contracts with an earlier effective ye ar are still under their original term and have not yet been renewed. The expiration year of the cont ract also varied between the higher and lower risk groups. Almost 94% of contracts for prope rties at higher risk of fail-ou t were due to expire by 2014. For lower risk properties, the percentage of expiring properties by 2014 was about 78%; almost 22% of properties had a contract expiration year between 2020 and 2025. The difference in expiration year was statistically signif icant between the groups at a P-value of 0.003 (Table 6-25). A greater proportion of propert ies at higher fail-out risk had multiple program layers. Half of the higher risk propertie s had at least one other housing program in place in addition to the HUD rental assistance contract, most notably the HUD Sec tion 236 program. About 31% of the properties in the lo wer risk group were covered by anot her housing program. The number of funding programs was compared between the risk groups. The significance test produced a Pvalue of 0.050 (Table 6-26). Therefore, it was inferred that the number of housing programs between the groups was statistically different; pr operties at higher fail-out had a greater number of housing programs in place.

PAGE 124

124 Tenant characteristics were available for almo st 67% of properties at higher risk of failout and 59% of properties in the lower risk gr oup. The proportions in Table 6-27 were for the properties for which tenant data were available. The majority of propertie s classified at higher fail-out risk had a large proporti on of tenants at age 62 and olde r and a smaller proportion of residents with a female head of household and ch ildren. Significance tests of the characteristics data confirmed a statistically significant differe nce between the risk group s in the proportion of households with a female head and children (P-val ue at 0.010, Table 6-28), and in the proportion of households at age 62 and older (P-value at 0.033, Table 6-29); the properties at higher risk of fail-out had a relatively greater number of households that were older than 61 and a relatively smaller number of single-mom family households. At least 82% of households in each property in both risk gr oups were considered very low income at or below 50% of area median income. Both the higher risk and lower risk properties also served extremely low income hous eholds, those that were at or below 30% of AMI. Fifty to 74% of households were ELI for 40% of higher risk properties and more than 39% of lower risk properties; 75 to 100% of households were ELI fo r 50% of the higher risk group and almost 61% of the lower risk group. A t-test concluded that no signif icant difference existed between the proportions of ELI house holds served by the properties in each risk group (P-value at 0.092, Table 6-30). Household income as a percentage of local median family amounted to less than 25% for almost 77% of lower risk properties, compared to 50% of those at higher risk. Income as a percentage of local median family income was between 25 and 49% for half of all higher risk properties and almost 24% of lower risk prop erties. Significance tests were performed to compare the household income as a percentage of local median family income (P-value at 0.346,

PAGE 125

125 Table 6-31), as well as the annual household income (P-v alue at 0.101, Table 6-32). The conclusion was drawn that there was no significant difference in household income between the groups. Two other variations between the risk gr oups were observed, which related to the proportion of minority households and overhoused households. From 75 to 100% of households were considered minority for nearly 73% of lowe r risk properties, compared to 45% of higher risk properties. A t-test was run and found a statistical differe nce in minority households (Pvalue at 0.033, Table 6-33); properties at higher risk of fail-out housed a lower proportion of minority households. Less than 25% of households were overhoused in 65% of higher risk properties and in 33% of lower risk properties. The condition of overhousing may be explained by unit size; properties at higher fail-out risk are smaller in terms of number of bedrooms per unit. Therefore, overhousing is less likely to occur. A signifi cance test concluded th at the difference in overhousing between the risk groups was statisti cally significant (P-val ue at 0.018, Table 6-34); a smaller percentage of households in the higher fail-out group was overhoused. Opt-out Model Descriptive Analysis and Significance Tests Table 6-35 with property charac teristics showed that almost 78% of properties classified as higher risk of opt-out contai ned less than 50 units. The properti es at lower opt-out risk were more evenly distributed among pr operty size. On average, th e higher risk group contained 47 units compared to 120 units in the lower risk gr oup. A significance test confirmed that there was a statistical difference in the number of units between the groups (P-value < 0.000, Table 6-36); the properties in the higher risk group had a smaller number of units.

PAGE 126

126 A large variation in target population also seemed appare nt. The elderly were targeted by more than 66% of the higher risk properties and only 23% of the lower risk properties. A z-test was performed to compare the proportion of elderly by risk group. At a P-value of less than 0.000 (Table 6-37), the conclusion was drawn that the proportion of elderly was significantly different; a larger proportion of pr operties with a higher risk of opt-out targeted the elderly. The difference in target population could also explain the variation in bedroom configuration between higher and lo wer risk properties. The major ity of propertie s in the higher risk group had studio, 1 and 2 be droom units, with an average number of bedrooms at 1.3. The smaller units are more appropriate for elderly ho useholds. Most of the properties in the higher risk group contained 1, 2 and 3 bedroom apartments, averaging 1.9 bedrooms per property. The difference in unit size was found to be statis tically significant (P-v alue < 0.000, Table 6-38); higher risk properties had smaller units. The majority of all for-p rofit properties were built during 1970-1974 or 1980-1984. The lower risk properties were mostly built during both time periods and average a year built of 1976. The higher risk properties were mainly constructed during 1980-1984 and had an average year built of 1981. The t-test concluded that a significant difference existed between the years built of the risk groups (P-value < 0.000, Table 639); properties at higher risk of opt-out were built in later years. It is likely that this finding was impacted by the condition that properties in the higher risk group had original contract terms. These could not have been built during the 1960s or 1970s, because most of those contracts would have already had an option to renew. High REAC scores were achieved by both hi gher and lower risk pr operties. Nearly 67% of properties at higher opt-out ri sk and 79% of those at lower ri sk scored at least 80. A failing score of below 60 was observed for roughly 22% of the higher risk group and less than 4% of the

PAGE 127

127 lower risk group. According to th e results of a significance test the REAC scores between the risk groups were statistically diffe rent (P-value at 0.04, Table 6-40); properties at higher risk had lower REAC scores. The financial characteristics (Table 6-41) di ffered between the risk groups as a result of one of the criteria that were esta blished to compose the list of pr operties at heightened risk of loss. Properties had to experien ce at least a 20% estimated incr ease in the mean NOI. For more than 55% of the higher risk prope rties, an increase of 20 to 39.9% was estimated; the increase in mean NOI was 40 to 59.9% for almost 41% of the pr operties. In the lower risk group, an increase of less than 20% was calculated for almost 45% of the properties. A significance test confirmed a statistical difference in the mean NOI change be tween the risk groups (P-value at 0.031, Table 642); the properties at higher risk of opt-out had a larger percentage increase in NOI when project rents are converted to market rents. According to the subsidy characteristics (Table 6-43), the rent to Fa ir Market Rent ratio was below 100% for most higher risk and lower ri sk properties. Less than 19% of properties in the lower risk group had a project rent that was less than 80% of FMR, which compared to more than 39% of the properties in the lower risk group. On average, the project rent to FMR amounted to 86.4% for properties at higher risk a nd 84.7% for those at lower risk. At a P-value of 0.391 (Table 6-44), the difference in the projec t rent to FMR was not statistically different between the risk groups. The prevailing types of rental assistance programs varied by risk group. More than 85% of properties at higher risk of opt-out were funded under the Sec tion 8 Substantial Rehabilitation program and 11% under the Section 8 New Cons truction program. LMSA was the governing

PAGE 128

128 program type among lower risk properties; al most 51% were funded under a LMSA contract. Section 8 NC and Section 8 SR each funded nearly 18% of lower risk properties. Approximately 93% of higher risk properties had a current rental assistance contract effective year of 1980 to 1989 and a contract expiration year of 2008 to 2014. This large proportion was the result of the methodology that was applied to flag properties at higher risk of opt-out. To be identified at ris k, properties had to be effective under the original contract term and have an expiration by year end 2014. In contra st, the majority of lower risk properties had a current contract year of 2000-2008 and an e xpiration year that varied from 2008 to 2025. All but one of the properties in the hi gher risk group had the HUD rental assistance program as the only funding layer. More than 28% of lower risk propert ies had at least two housing programs, most notably a HUD insured mort gage without income restrictions in addition to the rental assistance program The difference in number of funding programs between the groups was statistically significan t (P-value 0.001, Table 6-45). As illustrated by Table 6-46, tenant characteri stics data were only av ailable for four of the higher risk properties and 43 of those at lower risk. At least ha lf of the households in almost 70% of the properties at lower risk of opt-out housed singlemoms. The proportion of households with elderly tenants was small fo r the lower risk group; more than 69% of properties housed less than 25% elderly. This was in lin e with the observation that family not elderly was the major target population group for the lower risk proper ties. No further comparative analysis was conducted, because the sample of higher risk proper ties for which data were available was small. Correlation and Scenario Analysis An increase of at least 20% in NOI when assi sted rents are converted to market rents was one of the conditions to be met for a property to be included in the group of properties at higher risk of opt-out. For more than 95% of the for-p rofit properties, the mean net operating income

PAGE 129

129 improved under the scenario of al l market rents. The reason fo r the increase in income was twofold. First, almost 93% of the properties had a project rent that was lower than the Fair Market Rent. Second, estimated mean market rents exceeded the Fair Market Rent for all the unit types in Miami-Dade and for two of the bedroom sizes in Duval. For lack of a public data source with detailed and current market rents, the input values for the market scenario under the opt-out model were derived from Zilpy.com. The estimated market rents seemed reasonable, based on a compar ison to the Fair Market Rent data. But since the reliability of Zilpy.com was unknown, the mark et scenario under the opt-out model was rerun with the 2009 Fair Market Rents as the input valu es for all units (in place of Zilpy rents). The mean NOI under the scenario of 2009 FMR was compared to the mean NOI under the scenario of assisted rents by calculating the change in mean NOI. Almost 88% of the properties saw an increase in NOI of at least 20%. This compared to 70% of the properties with market rents based on Zilpy, because the estimated Zilpy rents were lower than 2009 FMR for al l of the unit sizes in Duval. Under the scenario of all 2009 FMR rent s, only 2% of the prope rties (two properties) experienced a drop in mean NOI, compared to a lmost 5% (four propertie s) under the calculation with Zilpy-derived rents. The pr operties that had a project rent to FMR ratio that exceeded 100% did not achieve a positive change in NOI by at least 20% under the s cenario with 2009 FMR rents. The percentage change in the recalculated m ean NOI was correlated to the project rent to FMR ratio. The outcome was a strong negative correlation at -0.876; a standard deviation increase in project rent to FMR ratio corresponded to a 0.876 decrease in the mean NOI change. Table 6-47 compared the Fair Market Re nts for 2008 and 2009 to the simulated mean rents (based on Zilpy) by county and bedroom size.

PAGE 130

130 Table 6-1. Summary of Estimated Net Operating Income Values per Unit per Month for All Properties Mean Properties Percentage $83-149 10 8.4% $150-199 12 10.1% $200-249 42 35.3% $250-299 37 31.1% $300-349 13 10.9% $350-403 5 4.2% 25th Percentile $66-99 7 5.9% $100-149 6 5.0% $150-199 25 21.0% $200-249 43 36.1% $250-299 28 23.5% $300-349 7 5.9% $350-387 3 2.5% 75th Percentile $100-149 7 5.9% $150-199 10 8.4% $200-249 30 25.2% $250-299 45 37.8% $300-349 18 15.1% $350-399 6 5.0% $400-419 3 2.5%

PAGE 131

131 0 5 10 15 20 25 30 35 40 45$83-99$100-149$150-199$200249$250-299$300-349$350-403Mean NOINumber of Properties Figure 6-1. Mean Net Operating Income Table 6-2. Comparison of the Simulated Mean an d Median NOI Values to Reported NOI Values for Non-Market Rent Properties Per Unit per Month in Jacks onville MSA and Miami MSA Mean Median Properties Units Simulated NOI $242$245119 12,570 Garden Properties, Jacksonville MSA $223$27216 1,758 Elevator Properties, Jacksonville MSA $176$27611 1,181 Garden Properties, Miami MSA $261$25715 1,579 Elevator Properties, Miami MSA $385$33459 6,969 Source: Urban Land Institute (2006).

PAGE 132

132 Table 6-3. Number of Properties with an Estim ated Net Operating Income Below $200 per Unit per Month by Probability for All Properties PropertiesPercentage of 119 Properties Mean 22 18.5% Minimum 81 68.1% 5th percentile 56 47.1% 25th percentile 38 31.9% 50th percentile 22 18.5% 75th percentile 17 14.3% 95th percentile 12 10.1% Maximum 7 5.9% Table 6-4. Summary of Estimated Debt Coverage Ratio Mean Properties Percentage 0.83-0.99 2 7.1% 1.0-1.19 2 7.1% 1.2-1.49 3 10.7% 1.5-1.99 2 7.1% 2.0-2.99 7 25.0% 3.0-3.99 7 25.0% 4.0-4.99 3 10.7% 5.0-6.68 2 7.1% 25th Percentile 0.80-0.99 3 10.7% 1.0-1.19 1 3.6% 1.2-1.49 4 14.3% 1.5-1.99 5 17.9% 2.0-2.99 7 25.0% 3.0-3.99 6 21.4% 4.0-4.99 0 0.0% 5.0-6.03 2 7.1% 75th Percentile 0.87-0.99 2 7.1% 1.0-1.19 0 0.0% 1.2-1.49 3 10.7% 1.5-1.99 3 10.7% 2.0-2.99 5 17.9% 3.0-3.99 7 25.0% 4.0-4.99 5 17.9% 5.0-7.03 3 10.7%

PAGE 133

133 Table 6-5. Summary of Estimated Net Operating Income Values per Unit per Month by Rent Scenario for Opt-out Risk Model @ Assisted @ Market Mean Properties Percentage Properties Percentage $149-199 89.6%1 1.2% $200-249 35 42.2% 14 16.9% $250-299 26 31.3% 17 20.5% $300-349 12 14.5% 17 20.5% $350-399 1 1.2% 20 24.1% $400-449 1 1.2% 11 13.3% $450-499 0 0.0% 0 0.0% $500-529 0 0.0% 3 3.6% 25th Percentile $134-149 3 3.6% 0 0.0% $150-199 19 22.9% 6 7.2% $200-249 32 38.6% 23 27.7% $250-299 23 27.7% 21 25.3% $300-349 4 4.8% 20 24.1% $350-399 2 2.4% 10 12.0% $400-450 0 0.0% 3 3.6% 75th Percentile $165-199 4 4.8% 0 0.0% $200-249 25 30.1% 3 3.6% $250-299 34 41.0% 18 21.7% $300-349 14 16.9% 14 16.9% $350-399 5 6.0% 20 24.1% $400-449 1 1.2% 16 19.3% $450-499 0 0.0% 9 10.8% $500-549 0 0.0% 0 0.0% $550-588 0 0.0% 3 3.6%

PAGE 134

134 Table 6-6. Significance Test of Mean NOI by Rent Scenar io for Opt-out Risk Model Mean @ Assisted Mean @ Market Mean 252.0556965 329.4029134 Variance 2247.691523 5532.405399 Observations 83 83 Hypothesized Mean Difference 0 df 139 t Stat -7.988978161 P(T<=t) one-tail 2.33339E-13 t Critical one-tail 1.655889868 P(T<=t) two-tail 4.66678E-13 t Critical two-tail 1.977177694 Table 6-7. Percentage Change in Estimated NOI Value from Project Rents to Market Rents for the Opt-out Risk Model Mean 25th Percent ile 75th Percentile Properties PercentageProperties PercentageProperties Percentage -0.1%-16.0% 4 4.8% 9 10.8% 3 3.6% 0%-19.9% 20 24.1% 35 42.2% 18 21.7% 20.0%-39.9% 34 41.0% 21 25.3% 31 37.3% 40.0%-59.9% 17 20.5% 16 19.3% 23 27.7% 60.0%-79.0% 8 9.6% 2 2.4% 8 9.6%

PAGE 135

135 0 5 10 15 20 25 30 35 40 $149-199$200-249$250-299$300-349$350-399$400-449$450-499$500-529Mean NOINumber of Properties Mean NOI @ Assisted Mean NOI @ Market Figure 6-2. Mean Net Operating Income under Two Scenarios of Opt-out Model Table 6-8. Number of Properties with an Estimated Change of Net Operating Income of At Least 20% by Probability for the Opt-out Risk Model PropertiesPercentage of 83 Properties Mean 59 71.1% Minimum 21 25.3% 5th percentile 30 36.1% 25th percentile 39 47.0% 50th percentile 50 60.2% 75th percentile 62 74.7% 95th percentile 77 92.8% Maximum 83 100.0%

PAGE 136

136 Table 6-9. Summary of Estimated Debt Cove rage Ratio for the Opt-out Risk Model Mean Properties Percentage Properties Percentage 0.90-0.99 17.7% 0 0.0% 1.0-1.19 215.4% 1 7.7% 1.2-1.49 215.4% 3 23.1% 1.5-1.99 00.0% 1 7.7% 2.0-2.99 00.0% 0 0.0% 3.0-3.99 430.8% 0 0.0% 4.0-4.99 323.1% 1 7.7% 5.0-7.91 17.7% 7 53.8% 25th Percentile 0.86-0.99 215.4% 1 7.7% 1.0-1.19 17.7% 3 23.1% 1.2-1.49 215.4% 0 0.0% 1.5-1.99 00.0% 1 7.7% 2.0-2.99 17.7% 0 0.0% 3.0-3.99 646.2% 1 7.7% 4.0-4.99 00.0% 2 15.4% 5.0-6.81 17.7% 5 38.5% 75th Percentile 0.94-0.99 17.7% 0 0.0% 1.0-1.19 00.0% 1 7.7% 1.2-1.49 215.4% 3 23.1% 1.5-1.99 215.4% 0 0.0% 2.0-2.99 00.0% 1 7.7% 3.0-3.99 17.7% 0 0.0% 4.0-4.99 538.5% 1 7.7% 5.0-9.14 215.4% 7 53.8%

PAGE 137

137 Table 6-10. Correlation Matrix NOI Total_ Units Target_ Pop Year_ built Ownership Reac_score FMR Program_ lyr Expiration_ dt NOI Pearson Correlation 1 -.262 ** .128 .522 ** .244 ** -.029 .577 ** -.152 .007 Sig. (2-tailed) .004 .166 .000 .008 .754 .000 .098 .940 N 119 119 119 119 119 119 119 119 119 Total_Units Pearson Correlation -.262 ** 1 -.120 -.286 ** -.194 .110 -.098 .379 ** -.060 Sig. (2-tailed) .004 .194 .002 .035 .235 .290 .000 .518 N 119 119 119 119 119 119 119 119 119 Target_Pop Pearson Correla tion .128 -.120 1 .299 ** -.220 .038 .272 ** .075 -.304 Sig. (2-tailed) .166 .194 .001 .016 .679 .003 .417 .001 N 119 119 119 119 119 119 119 119 119 Year_built Pearson Correlation .522 ** -.286 ** .299 ** 1 .109 -.009 .299 ** -.114 -.268 ** Sig. (2-tailed) .000 .002 .001 .237 .918 .001 .218 .003 N 119 119 119 119 119 119 119 119 119 Ownership Pearson Correlation .244 ** -.194 -.220 .109 1 -.023 .162 -.287 ** .239 ** Sig. (2-tailed) .008 .035 .016 .237 .806 .077 .002 .009 N 119 119 119 119 119 119 119 119 119 Reac_score Pearson Correlation -.029 .110 .038 -.009 -.023 1 .189 .103 -.024 Sig. (2-tailed) .754 .235 .679 .918 .806 .039 .266 .793 N 119 119 119 119 119 119 119 119 119 FMR Pearson Correlation .577 ** -.098 .272 ** .299 ** .162 .189 1 -.013 .034 Sig. (2-tailed) .000 .290 .003 .001 .077 .039 .885 .715 N 119 119 119 119 119 119 119 119 119 Program_lyr Pearson Correlation -.152 .379 ** .075 -.114 -.287 ** .103 -.013 1 -.130 Sig. (2-tailed) .098 .000 .417 .218 .002 .266 .885 .160 N 119 119 119 119 119 119 119 119 119 Expiration_dt Pearson Corre lation .007 -.060 -.304 ** -.268 ** .239 ** -.024 .034 -.130 1 Sig. (2-tailed) .940 .518 .001 .003 .009 .793 .715 .160 N 119 119 119 119 119 119 119 119 119 **. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).

PAGE 138

138 Table 6-11. Results of Stepwise Regression for A ll Properties with Mean Net Operating Income as the Dependent Variable Variables Entered/Removeda Model Variables Entered Variables Removed Method 1 FMR Stepwise (Criteria: Probability-ofF-to-enter <= .050, Probability-ofF-to-remove >= .100). 2 Year_built Stepwise (Criteria: Probability-ofF-to-enter <= .050, Probability-ofF-to-remove >= .100). a. Dependent Variable: NOI Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate 1 .577a .333 .328 50.055 2 .683 b .467 .458 44.945 a. Predictors: (Constant), FMR b. Predictors: (Consta nt), FMR, Year_built ANOVAc Model Sum of Squares df Mean Square F Sig. 1 Regression 146501.335 1 146501.335 58.471 .000a Residual 293146.469 117 2505.525 Total 439647.804 118 2 Regression 205323.497 2 102661.748 50.822 .000 b Residual 234324.307 116 2020.037 Total 439647.804 118 a. Predictors: (Constant), FMR b. Predictors: (Consta nt), FMR, Year_built c. Dependent Variable: NOI

PAGE 139

139 Table 6-11. Continued Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 9.947 30.730 .324 .747 FMR 276.825 36.202 .577 7.647 .000 2 (Constant) -8498.348 1576.951 -5.389 .000 FMR 221.884 34.063 .463 6.514 .000 Year_built 4.326 .802 .383 5.396 .000 a. Dependent Variable: NOI Excluded Variablesc Model Beta In t Sig. Partial Correlation Collinearity Statistics Tolerance 1 Total_Units -.208a -2.821 .006 -.253 .990 Target_Pop -.031a -.397 .692 -.037 .926 Year_built .383a 5.396 .000 .448 .911 Ownership .154a 2.042 .043 .186 .974 Reac_score -.143a -1.885 .062 -.172 .964 Program_lyr -.145a -1.938 .055 -.177 1.000 Epriration_dt -.013a -.165 .869 -.015 .999 2 Total_Units -.117 b -1.666 .098 -.154 .918 Target_Pop -.129 b -1.791 .076 -.165 .874 Ownership .131 b 1.922 .057 .176 .970 Reac_score -.118 b -1.715 .089 -.158 .959 Program_lyr -.104 b -1.531 .129 -.141 .987 Epriration_dt .103 b 1.457 .148 .135 .914 a. Predictors in the Model: (Constant), FMR b. Predictors in the Model: (Constant), FMR, Year_built c. Dependent Variable: NOI

PAGE 140

140 Table 6-12. Property Characterist ics for the Fail-out Model Higher Risk Lower Risk Total Property Characteristics: Number of Properties 32 26.9% 87 73.1% 119 100.0% Number of Total Units 3,978 31.6% 8,592 68.4% 12,570 100.0% Number of Assisted Units 3,978 32.9% 8,112 67.1% 12,090 100.0% Property Size: Less than 50 units 8 25.0% 29 33.3% 37 31.1% 50-99 units 5 15.6% 19 21.8% 24 20.2% 100-149 units 4 12.5% 12 13.8% 16 13.4% 150-199 units 5 15.6% 13 14.9% 18 15.1% more than 200 units 10 31.3% 14 16.1% 24 20.2% Average number of units 124 99 106 Bedroom Configuration: 0 bedroom units 801 20.1% 689 8.0% 1,490 11.9% 1 bedroom units 1,987 49.9% 3,179 37.0% 5,166 41.1% 2 bedroom units 818 20.6% 2,924 34.0% 3,742 29.8% 3 bedroom units 344 8.6% 1,468 17.1% 1,812 14.4% 4 or more bedroom units 28 0.7% 332 3.9% 360 2.9% Average number of bedrooms 1.1 1.5 1.4 Target Population: Family 14 43.8% 52 59.8% 66 55.5% Elderly 18 56.3% 34 39.1% 52 43.7% Persons with Disabilities 0.0% 1 1.1% 1 0.8% Type of Ownership: Non-profit 16 50.0% 20 23.0% 36 30.3% For-profit 16 50.0% 67 77.0% 83 69.7% Year Built: 1967-1969 2 6.3% 4 4.6% 6 5.0% 1970-1974 22 68.8% 19 21.8% 41 34.5% 1975-1979 0.0% 10 11.5% 10 8.4% 1980-1984 7 21.9% 49 56.3% 56 47.1% 1985-1989 1 3.1% 3 3.4% 4 3.4% 1990-1994 0.0% 2 2.3% 2 1.7% Average Year Built 1974 1979 1978

PAGE 141

141 Table 6-12. Continued REAC Physical Inspection Score: 13-29 3 9.4% 0.0% 3 2.5% 30-59 8 25.0% 0.0% 8 6.7% 60-69 1 3.1% 9 10.3% 10 8.4% 70-79 3 9.4% 10 11.5% 13 10.9% 80-89 6 18.8% 36 41.4% 42 35.3% 90-100 11 34.4% 32 36.8% 43 36.1% Average REAC Score 71 85 81 County: Duval 17 53.1% 30 34.5% 47 39.5% Miami-Dade 15 46.9% 57 65.5% 72 60.5% Metropolitan Location: Metropolitan/central city 14 43.8% 44 50.6% 58 48.7% Suburb 16 50.0% 34 39.1% 50 42.0% Non-metropolitan 2 6.3% 9 10.3% 11 9.2% Low Poverty Census Tract: No 30 93.8% 76 87.4% 106 89.1% Yes 2 6.3% 11 12.6% 13 10.9% Table 6-13. Significance Test of Total Units pe r Property by Risk Group for the Fail-out Risk Model Total units higher risk Total units lower risk Mean 124.3125 98.75862069 Variance 6395.899194 5714.534082 Observations 32 87 Hypothesized Mean Difference 0 Df 53 t Stat 1.568117703 P(T<=t) one-tail 0.061402686 t Critical one-tail 1.674116237 P(T<=t) two-tail 0.122805372 t Critical two-tail 2.005745949

PAGE 142

142 Table 6-14. Significance Test of Average Unit Si ze by Risk Group for the Fail-out Risk Model Average br size higher risk Average br size lower risk Mean 1.14439346 1.53190707 Variance 0.60584367 0.76796959 Observations 32 87 Hypothesized Mean Difference 0 Df 62 t Stat -2.32583226 P(T<=t) one-tail 0.01165651 t Critical one-tail 1.66980416 P(T<=t) two-tail 0.02331302 t Critical two-tail 1.9989715 Table 6-15. Significance Test of Proportion of Elderly Properties by Risk Group for the Fail-out Risk Model Elderly Family Total % High risk 18 14 32 0.5625 Low risk 34 52 86 0.3953 Ho = 21=0 Estimate null hypothesis value -0.167151163 proportion of total sample 0.440677966 standard error 0.102803623 z= -1.625926768 P-value 0.051982617

PAGE 143

143 Table 6-16. Significance Test of Proportion of Non-profit Owners by Risk Group for the Fail-out Risk Model Non-profit For-profit Total % High risk 16 16 32 0.5000 Low risk 20 67 87 0.2299 Estimate null hypothesis value -0.270114943 proportion of total sample 0.302521008 standard error 0.094969073 z= -2.844241109 P-value 0.002225867 Table 6-17. Significance Test of Year Built by Risk Group for the Fail-out Risk Model Yr built higher risk Yr built lower risk Mean 1974.25 1978.701149 Variance 23.41935484 26.30499866 Observations 32 87 Hypothesized Mean Difference 0 Df 58 t Stat -4.376909501 P(T<=t) one-tail 2.54442E-05 t Critical one-tail 1.671552763 P(T<=t) two-tail 5.08885E-05 t Critical two-tail 2.001717468

PAGE 144

144 Table 6-18. Significance Test of REAC Physical Inspection Score by Risk Group for the Fail-out Risk Model REAC higher risk REAC lower risk Mean 71.21875 84.551724 Variance 738.24093 90.017642 Observations 32 87 Hypothesized Mean Difference 0 Df 34 t Stat -2.715664 P(T<=t) one-tail 0.00516 t Critical one-tail 1.6909242 P(T<=t) two-tail 0.0103201 t Critical two-tail 2.0322445 Table 6-19. Financial Characteri stics for the Fail-out Model Higher Risk Lower Risk Total Mean NOI: $84-149 10 31.3% 0.0% 10 8.4% $150-199 12 37.5% 0.0% 12 10.1% $200-249 4 12.5% 38 43.7% 42 35.3% $250-299 4 12.5% 33 37.9% 37 31.1% $300-349 1 3.1% 12 13.8% 13 10.9% $350-403 1 3.1% 4 4.6% 5 4.2% Mean DCR: 0.83-0.99 2 6.3% 0.0% 2 1.7% 1.0-1.19 0.0% 2 2.3% 2 1.7% 1.2-1.49 0.0% 3 3.4% 3 2.5% 1.49-1.99 1 3.1% 1 1.1% 2 1.7% 2.0-2.99 7 21.9% 0.0% 7 5.9% 3.0-3.99 2 6.3% 5 5.7% 7 5.9% 4.0-4.99 1 3.1% 2 2.3% 3 2.5% 5.0-6.68 1 3.1% 1 1.1% 2 1.7% Not available 18 56.3% 73 83.9% 91 76.5%

PAGE 145

145 Table 6-20. Subsidy Characteristics for the Fail-out Model Higher Risk Lower Risk Total Percentage HUD RA Units: 0-24% 0.0% 1 1.1% 1 0.8% 25-49% 1 3.1% 2 2.3% 3 2.5% 50-74% 4 12.5% 2 2.3% 6 5.0% 75-99% 5 15.6% 5 5.7% 10 8.4% 100% 22 68.8% 77 88.5% 99 83.2% Average % HUD RA Units 92.4% 96.5% 95.4% Rent to FMR Ratio: 50-59.9% 2 6.3% 2 2.3% 4 3.4% 60-69.9% 6 18.8% 2 2.3% 8 6.7% 70-79.9% 13 40.6% 23 26.4% 36 30.3% 80-89.9% 5 15.6% 36 41.4% 41 34.5% 90-99.9% 3 9.4% 16 18.4% 19 16.0% 100-109.9% 3 9.4% 4 4.6% 7 5.9% 110-119.9% 0.0% 2 2.3% 2 1.7% 120-129.9% 0.0% 1 1.1% 1 0.8% 130-139.9% 0.0% 0.0% 0.0% 140-142.2% 0.0% 1 1.1% 1 0.8% Average Rent to FMR Ratio 78.7% 85.8% 83.9% Number of HUD RA Contracts per Property: 1 31 96.9% 86 98.9% 117 98.3% 2 1 3.1% 1 1.1% 2 1.7% HUD RA Program Type (by contract): Loan Management Set Aside 21 63.6% 26 29.5% 47 38.8% Section 8 Substantial Rehab 6 18.2% 29 33.0% 35 28.9% Section 8 New Construction 1 3.0% 25 28.4% 26 21.5% HFDA/8 New Construction 1 3.0% 4 4.5% 5 4.1% Property Disposition/8 Existing 2 6.1% 3 3.4% 5 4.1% Property Disposition/8 Mod. Rehab 0.0% 1 1.1% 1 0.8% Section 8 Preservation 1 3.0% 0.0% 1 0.8% Rent Supplement 1 3.0% 0.0% 1 0.8% Contract Renewal History (by contract): Original contract term 8 24.2% 40 45.5% 48 39.7% Renewed contract 25 75.8% 48 54.5% 73 60.3% Current Contract Effective Year: 1970-1979 1 3.0% 4 4.5% 5 4.1% 1980-1989 6 18.2% 35 39.8% 41 33.9% 1990-1999 1 3.0% 2 2.3% 3 2.5% 2000-2008 25 75.8% 47 53.4% 72 59.5%

PAGE 146

146 Table 6-20. Continued Higher Risk Lower Risk Total Expiration of Current Contract: 2008-2009 13 39.4% 14 15.9% 27 22.3% 2010-2014 18 54.5% 54 61.4% 72 59.5% 2015-2019 0.0% 1 1.1% 1 0.8% 2020-2025 2 6.1% 19 21.6% 21 17.4% Program Source: HUD 32 100.0% 87 100.0% 119 100.0% RD 0.0% 0.0% 0.0% FHFC 5 15.6% 8 9.2% 13 10.9% LHFA 2 6.3% 4 4.6% 6 5.0% No. of Program Layers: 1 16 50.0% 66 75.9% 82 68.9% 2 11 34.4% 14 16.1% 25 21.0% 3 4 12.5% 4 4.6% 8 6.7% 4 1 3.1% 3 3.4% 4 3.4% Average No. of Program Layers 1.7 1.4 1.4 Funding Program Combinations: HUD RA Only 16 50.0% 66 75.9% 82 68.9% HUD RA; HUD Section 236 8 25.0% 0.0% 8 6.7% HUD RA; HUD Section 236; Elderly Housing Community Loan 2 6.3% 0.0% 2 1.7% HUD RA; HUD Insured Mortgage Unassisted 1 3.1% 11 12.6% 12 10.1% HUD RA; HUD Insured Mortgage Unassisted; Housing Credits 9%; SAIL 0.0% 2 2.3% 2 1.7% HUD RA; HUD Insured Mortgage Unassisted; State HOME; Elderly Housing Community Loan 1 3.1% 0.0% 1 0.8% HUD RA; HUD Insured Mortgage Unassisted; Bonds 2 6.3% 1 1.1% 3 2.5% HUD RA; Housing Credits 9% 0.0% 1 1.1% 1 0.8% HUD RA; Housing Credits 9%;State HOME 0.0% 1 1.1% 1 0.8% HUD RA; Housing Credits 4%; Bonds 0.0% 1 1.1% 1 0.8% HUD RA; State HOME 1 3.1% 0.0% 1 0.8% HUD RA; SAIL 1 3.1% 1 1.1% 2 1.7% HUD RA; Elderly Housing Community Loan; Bonds 0.0% 2 2.3% 2 1.7% HUD RA; Bonds 0.0% 1 1.1% 1 0.8%

PAGE 147

147 Table 6-21. Significance Test of Rent to Fair Market Rent Ra tio by Risk Group for the Fail-out Risk Model Rent to FMR higher risk Rent to FMR lower risk Mean 0.787127 0.857636 Variance 0.015299 0.01545 Observations 32 87 Hypothesized Mean Difference 0 Df 56 t Stat -2.753588 P(T<=t) one-tail 0.003967 t Critical one-tail 1.672522 P(T<=t) two-tail 0.007933 t Critical two-tail 2.003241 Table 6-22. Significance Test of Proportion of Properties with a LMSA contract by Risk Group for the Fail-out Risk Model LMSA Other Total % High risk 21 12 33 0.6364 Low risk 26 62 88 0.2955 Estimate null hypothesis value 0.340909091 proportion of total sample 0.388429752 standard error 0.09948871 z= 3.426610844 P-value 0.000305582

PAGE 148

148 Table 6-23. Significance Test of Proportion of Renewed Contract s by Risk Group for the Failout Risk Model Contract Renewed Original Contract Total % High risk 25 8 33 0.7576 Low risk 48 40 88 0.5455 Estimate null hypothesis value -0.212121212 proportion of total sample 0.603305785 standard error 0.099859884 z= -2.124188447 P-value 0.016827194 Table 6-24. Significance Test of Contract Effective Year by Risk Group for the Fail-out Risk Model Yr effective higher risk Yr effective lower risk Mean 1998.75 1993.64368 Variance 114.83871 140.185512 Observations 32 87 Hypothesized Mean Difference 0 Df 61 t Stat 2.23926221 P(T<=t) one-tail 0.01439799 t Critical one-tail 1.67021948 P(T<=t) two-tail 0.02879599 t Critical two-tail 1.99962357

PAGE 149

149 Table 6-25. Significance Test of Contract Expiration Year by Risk Group for the Fail-out Risk Model Expiration yr higher risk Expiration yr lower risk Mean 2010.8438 2013.1494 Variance 8.5877016 25.733226 Observations 32 87 Hypothesized Mean Difference 0 df 95 t Stat -3.069735 P(T<=t) one-tail 0.0013966 t Critical one-tail 1.6610518 P(T<=t) two-tail 0.0027931 t Critical two-tail 1.985251 Table 6-26. Significance Test of Number of Fund ing Layers by Risk Group for the Fail-out Risk Model Funding layers higher risk Funding Layers lower risk Mean 1.6875 1.356322 Variance 0.673387 0.534349 Observations 32 87 Hypothesized Mean Difference 0 Df 50 t Stat 2.008606 P(T<=t) one-tail 0.024997 t Critical one-tail 1.675905 P(T<=t) two-tail 0.049995 t Critical two-tail 2.008559

PAGE 150

150 Table 6-27. Tenant Characteris tics for the Fail-out Model Higher Risk Lower Risk Total Properties with data 20 66.7% 51 58.6% 71.0 59.7% Properties without data 10 33.3% 36 41.4% 46.0 38.7% % Female Head with Children: 0-24% 11 55.0% 17 33.3% 28.0 39.4% 25-49% 3 15.0% 3 5.9% 6.0 8.5% 50-74% 6 30.0% 19 37.3% 25.0 35.2% 75-100% 0.0% 12 23.5% 12.0 16.9% % of All Persons with Disability: 0-24% 19 95.0% 49 96.1% 68.0 95.8% 25-49% 1 5.0% 2 3.9% 3.0 4.2% 50-74% 0.0% 0.0% 0.0% 75-100% 0.0% 0.0% 0.0% % 62 Year of Age or More: 0-24% 7 35.0% 30 58.8% 37.0 52.1% 25-49% 2 10.0% 4 7.8% 6.0 8.5% 50-74% 1 5.0% 5 9.8% 6.0 8.5% 75-100% 10 50.0% 12 23.5% 22.0 31.0% % Minority: 0-24% 6 30.0% 1 2.0% 7.0 9.9% 25-49% 2 10.0% 5 9.8% 7.0 9.9% 50-74% 3 15.0% 8 15.7% 11.0 15.5% 75-100% 9 45.0% 37 72.5% 46.0 64.8% Annual Household Income (rounded to $000's): $3,000 0.0% 1 2.0% 1.0 1.4% $4,000 0.0% 0.0% 0.0% $5,000 0.0% 2 3.9% 2.0 2.8% $6,000 2 10.0% 1 2.0% 3.0 4.2% $7,000 1 5.0% 16 31.4% 17.0 23.9% $8,000 5 25.0% 9 17.6% 14.0 19.7% $9,000 4 20.0% 9 17.6% 13.0 18.3% $10,000 4 20.0% 7 13.7% 11.0 15.5% $11,000 1 5.0% 3 5.9% 4.0 5.6% $12,000 3 15.0% 3 5.9% 6.0 8.5% % Very Low Income: 0-24% 0.0% 0.0% 0.0% 25-49% 0.0% 0.0% 0.0% 50-74% 0.0% 0.0% 0.0% 75-100% 20 100.0% 51 100.0% 71.0 100.0%

PAGE 151

151 Table 6-27. Continued Higher Risk Lower Risk Total % Extremely Low Income: 0-24% 0.0% 0.0% 0.0% 25-49% 2 10.0% 0.0% 2.0 2.8% 50-74% 8 40.0% 20 39.2% 28.0 39.4% 75-100% 10 50.0% 31 60.8% 41.0 57.7% Household Income as % of Local Median Family Income: 0-24% 10 50.0% 39 76.5% 49.0 69.0% 25-49% 10 50.0% 12 23.5% 22.0 31.0% 50-74% 0.0% 0.0% 0.0% 75-100% 0.0% 0.0% 0.0% Average Monthly Rent with Utilities: $97-149 3 15.0% 26 51.0% 29.0 40.8% $150-199 10 50.0% 21 41.2% 31.0 43.7% $200-249 5 25.0% 4 7.8% 9.0 12.7% $250-275 2 10.0% 0.0% 2.0 2.8% Average Years Since Moved in: 2-5 12 60.0% 28 54.9% 40.0 56.3% 6-9 7 35.0% 19 37.3% 26.0 36.6% 10-13 1 5.0% 4 7.8% 5.0 7.0% % of Households Overhoused: 0-24% 13 65.0% 17 33.3% 30.0 42.3% 25-49% 2 10.0% 11 21.6% 13.0 18.3% 50-74% 4 20.0% 16 31.4% 20.0 28.2% 75-100% 1 5.0% 7 13.7% 8.0 11.3%

PAGE 152

152 Table 6-28. Significance Test of Percentage of Female Heads of Household with Children by Risk Group for the Fail-out Risk Model % Female head with kids higher risk % Female head with kids lower risk Mean 23.6 44.98039 Variance 803.7263 1132.42 Observations 20 51 Hypothesized Mean Difference 0 df 41 t Stat -2.706799 P(T<=t) one-tail 0.004929 t Critical one-tail 1.682878 P(T<=t) two-tail 0.009858 t Critical two-tail 2.019541 Table 6-29. Significance Test of Percentage of Heads of Households at Age 62 or Older by Risk Group for the Fail-out Risk Model % 62 or more higher risk % 62 or more lower risk Mean 59.3 36.352941 Variance 1626.4316 1217.7929 Observations 20 51 Hypothesized Mean Difference 0 df 31 t Stat 2.2372753 P(T<=t) one-tail 0.0162983 t Critical one-tail 1.6955187 P(T<=t) two-tail 0.0325967 t Critical two-tail 2.0395134

PAGE 153

153 Table 6-30. Significance Test of Percentage of Extremely Low Income Households by Risk Group for the Fail-out Risk Model % ELI higher risk % ELI lower risk Mean 70.6 76.94117647 Variance 191.2 193.5764706 Observations 20 51 Hypothesized Mean Difference 0 df 35 t Stat -1.73515334 P(T<=t) one-tail 0.045756311 t Critical one-tail 1.68957244 P(T<=t) two-tail 0.091512622 t Critical two-tail 2.030107915 Table 6-31. Significance Test of H ousehold Income as a Percentage of the Local Median Family Income by Risk Group for the Fail-out Risk Model HH inc. as % higher risk HH inc. as % lower risk Mean 22.9 21.470588 Variance 31.884211 33.094118 Observations 20 51 Hypothesized Mean Difference 0 df 35 t Stat 0.9544026 P(T<=t) one-tail 0.1732128 t Critical one-tail 1.6895724 P(T<=t) two-tail 0.3464255 t Critical two-tail 2.0301079

PAGE 154

154 Table 6-32. Significance Test of Annual Household Income by Risk Group for the Fail-out Risk Model HH inc. higher risk HH inc. lower risk Mean 9.1 8.2941176 Variance 3.2526316 3.4117647 Observations 20 51 Hypothesized Mean Difference 0 df 36 t Stat 1.6821043 P(T<=t) one-tail 0.0506022 t Critical one-tail 1.6882977 P(T<=t) two-tail 0.1012045 t Critical two-tail 2.028094 Table 6-33. Significance Test of Percentage of Minority Households by Risk Group for the Failout Risk Model % minority higher risk % minority lower risk Mean 60 81.98039216 Variance 1664.842105 563.7796078 Observations 20 51 Hypothesized Mean Difference 0 df 24 t Stat -2.263535982 P(T<=t) one-tail 0.016462116 t Critical one-tail 1.710882067 P(T<=t) two-tail 0.032924233 t Critical two-tail 2.063898547

PAGE 155

155 Table 6-34. Significance Test of Percentage of Overhoused Households by Risk Group for the Fail-out Risk Model % of hh overhoused higher risk % of hh overhoused lower risk Mean 21.1 40.6470588 Variance 873.357895 921.232941 Observations 20 51 Hypothesized Mean Difference 0 df 36 t Stat -2.4878762 P(T<=t) one-tail 0.00880874 t Critical one-tail 1.68829769 P(T<=t) two-tail 0.01761749 t Critical two-tail 2.02809399

PAGE 156

156 Table 6-35. Property Characteri stics for the Opt-out Model Higher Opt-out Risk Lower Opt-out Risk Total Property Characteristics: Number of Properties 2732.5%5667.5% 83 100.0% Number of Total Units 125815.8%669484.2% 7952 100.0% Number of Assisted Units 125816.6%634183.4% 7599 100.0% Property Size: Less than 50 units 2177.8%1017.9% 31 37.3% 50-99 units 311.1%1323.2% 16 19.3% 100-149 units 13.7%916.1% 10 12.0% 150-199 units 00.0%1221.4% 12 14.5% more than 200 units 27.4%1221.4% 14 16.9% Average number of units 47 120 96 Bedroom Configuration: 0 bedroom units 21617.2%2944.4% 510 6.4% 1 bedroom units 60347.9%209331.3% 2696 33.9% 2 bedroom units 35328.1%276841.4% 3121 39.2% 3 bedroom units 745.9%138520.7% 1459 18.3% 4 or more bedroom units 121.0%1542.3% 166 2.1% Average number of bedrooms 1.3 1.90.0% 1.8 Target Population: Family 933.3%4376.8% 52 62.7% Elderly 1866.7%1323.2% 31 37.3% Persons with Disabilities 00.0%00.0% 0 0.0% Type of Ownership: Non-profit 00.0%00.0% 0 0.0% For-profit 27100.0%56100.0% 83 100.0% Year Built: 1967-1969 00.0%610.7% 6 7.2% 1970-1974 00.0%2341.1% 23 27.7% 1975-1979 27.4%58.9% 7 8.4% 1980-1984 2592.6%1730.4% 42 50.6% 1985-1989 00.0%35.4% 3 3.6% 1990-1994 00.0%23.6% 2 2.4% Average Year Built 1981 1976 1978 REAC Physical Inspection Score: 13-29 27.4%00.0% 2 2.4% 30-59 414.8%23.6% 6 7.2% 60-69 13.7%712.5% 8 9.6% 70-79 27.4%35.4% 5 6.0% 80-89 1140.7%2137.5% 32 38.6% 90-100 725.9%2341.1% 30 36.1%

PAGE 157

157 Table 6-35. Continued Higher Opt-out Risk Lower Opt-out Risk Total Average REAC Score 74 84 81 County: Duval 00.0%3358.9% 33 39.8% Miami-Dade 27100.0%2341.1% 50 60.2% Metropolitan Location: Metropolitan/central city 1555.6%2951.8% 44 53.0% Suburb 829.6%2544.6% 33 39.8% Non-metropolitan 414.8%23.6% 6 7.2% Low Poverty Census Tract: Yes 00.0%916.1% 9 10.8% No 27100.0%4783.9% 74 89.2% Table 6-36. Significance Test of Total Units pe r Property by Risk Group for the Opt-out Risk Model Total units low risk Total units high risk Mean 119.5357143 46.59259259 Variance 4667.744156 4996.404558 Observations 56 27 Hypothesized Mean Difference 0 Df 50 t Stat 4.452350773 P(T<=t) one-tail 2.38696E-05 t Critical one-tail 1.675905026 P(T<=t) two-tail 4.77392E-05 t Critical two-tail 2.008559072

PAGE 158

158 Table 6-37. Significance Test of Proportion of Elderly Propertie s by Risk Group for the Opt-out Risk Model ElderlyFamily Total % High risk 18 9 27 0.6667 Low risk 13 43 56 0.2321 Ho = 21=0 Estimate null hypothesis value -0.4345238 proportion of total sample 0.37349398 standard error 0.11333592 z= -3.8339461 P-value 6.3052E-05 Table 6-38. Significance Test of Average Unit Size by Risk Group for the Opt-out Risk Model Average br size per property low risk Average br size per property high risk Mean 1.767384792 0.803010198 Variance 0.484078749 0.433132739 Observations 56 27 Hypothesized Mean Difference 0 df 54 t Stat 6.137881397 P(T<=t) one-tail 5.12083E-08 t Critical one-tail 1.673564907 P(T<=t) two-tail 1.02417E-07 t Critical two-tail 2.004879275

PAGE 159

159 Table 6-39. Significance Test of Year Built by Risk Group for the Opt-out Risk Model Yr built lower risk Yr built higher risk Mean 1976.428571 1980.925926 Variance 40.57662338 0.686609687 Observations 56 27 Hypothesized Mean Difference 0 df 59 t Stat -5.193054216 P(T<=t) one-tail 1.34339E-06 t Critical one-tail 1.671093033 P(T<=t) two-tail 2.68677E-06 t Critical two-tail 2.000995361 Table 6-40. Significance Test of REAC Physical Inspection Score by Risk Group for the Opt-out Risk Model REAC lower risk REAC higher risk Mean 84.14285714 73.59259259 Variance 123.0701299 613.4045584 Observations 56 27 Hypothesized Mean Difference 0 Df 31 t Stat 2.113591139 P(T<=t) one-tail 0.021349135 t Critical one-tail 1.695518742 P(T<=t) two-tail 0.04269827 t Critical two-tail 2.039513438 Table 6-41. Financial Characteri stics for the Opt-out Model Higher Opt-out Risk Lower Opt-out Risk Total Mean NOI Change: -0.1%-8.0% 00.0%47.1%4 4.8% 0%-19.9% 00.0%2137.5%21 25.3% 20.0%-39.9% 1555.6%1526.8%30 36.1% 40.0%-59.9% 1140.7%1017.9%21 25.3% 60.0%-75.6% 13.7%610.7%7 8.4%

PAGE 160

160 Table 6-42. Significance Test of the Mean NOI Change by Risk Group for the Opt-out Risk Model Mean % change lower risk Mean % change higher risk Mean 0.278078992 0.37676551 Variance 0.04577512 0.018419853 Observations 56 27 Pooled Variance 0.036994417 Hypothesized Mean Difference 0 Df 81 t Stat -2.189911565 P(T<=t) one-tail 0.015703493 t Critical one-tail 1.663883913 P(T<=t) two-tail 0.031406986 T Critical two-tail 1.989686288

PAGE 161

161 Table 6-43. Subsidy Characteristics for the Opt-out Model Higher Opt-out Risk Lower Opt-out Risk Total Funding Source: HUD 27100.0%56100.0% 83100.0% RD 00.0%00.0% 00.0% FHFC 13.7%47.1% 56.0% LHFA 00.0%47.1% 44.8% No. of Program Layers: 1 2696.3%4071.4% 6679.5% 2 00.0%1119.6% 1113.3% 3 13.7%35.4% 44.8% 4 00.0%23.6% 22.4% Average No. of Program Layers 1.1 1.4 1.3 Percentage HUD RA Units: 0-24% 00.0%00.0% 00.0% 25-49% 00.0%23.6% 22.4% 50-74% 00.0%11.8% 11.2% 75-99% 00.0%58.9% 56.0% 100% 27100.0%4885.7% 7590.4% Average % HUD RA Units 100.0% 96.0% 97.3% Number of HUD RA Contracts per Property: 1 27100.0%5598.2% 8298.8% 2 00.0%11.8% 11.2% HUD RA Program Type: HFDA/8 NC 00.0%35.3% 33.6% PD/8 MR 13.7%00.0% 11.2% LMSA 00.0%2950.9% 2934.5% PD/8 Existing 00.0%47.0% 44.8% Sec 8 NC 311.1%1017.5% 1315.5% Sec 8 SR 2385.2%1017.5% 3339.3% Preservation 00.0%11.8% 11.2% Rent to FMR Ratio: 67-79.9% 518.5%2239.3% 2732.5% 80-89.9% 1140.7%2341.1% 3441.0% 90-99.9% 1037.0%610.7% 1619.3% 100-109.9% 13.7%35.4% 44.8% 110-121.0% 00.0%23.6% 22.4% Average Rent to FMR Ratio 86.4% 84.7% 85.2% Funding Program Combinations: HUD RA Only 2696.3%4071.4% 6679.5% HUD RA; Bonds 00.0%11.8% 11.2% HUD RA: Bonds;HUD Insured Mortgage Unassisted 00.0%23.6% 22.4% HUD RA; Housing Credits 4%;Bonds 00.0%11.8% 11.2% HUD RA; Housing Credits 9% 00.0%11.8% 11.2% HUD RA; Housing Credits 9%;SAIL;HUD Insured Mortgage Unassisted 00.0%23.6% 22.4% HUD RA; Housing Credits 9%;State HOME 13.7%00.0% 11.2%

PAGE 162

162 Table 6-43. Continued Higher Opt-out Risk Lower Opt-out Risk Total HUD RA; HUD Insured Mortgage Unassisted 00.0%814.3% 89.6% HUD RA; HUD Section 236 00.0%11.8% 11.2% Contract Renewal History: Original contract term 27100.0%814.0% 3541.7% Renewed contract 00.0%4986.0% 4958.3% Current Contract Effective Year: 1970-1979 13.7%00.0% 11.2% 1980-1989 2592.6%712.3% 3238.1% 1990-1999 13.7%11.8% 22.4% 2000-2008 00.0%4986.0% 4958.3% Expiration of HUD RA Contract: 2008-2009 27.4%915.8% 1113.1% 2010-2014 2592.6%2849.1% 5363.1% 2015-2019 00.0%11.8% 11.2% 2020-2025 00.0%1933.3% 1922.6% Table 6-44. Significance Test of Rent to Fair Market Rent Ra tio by Risk Group for the Opt-out Risk Model Rent to FMR lower risk Rent to FMR higher risk Mean 0.84655797 0.863616578 Variance 0.011479136 0.005000712 Observations 56 27 Hypothesized Mean Difference 0 Df 73 t Stat -0.863579079 P(T<=t) one-tail 0.19532386 t Critical one-tail 1.665996224 P(T<=t) two-tail 0.390647721 t Critical two-tail 1.992997097

PAGE 163

163 Table 6-45. Significance Test of Number of Fund ing Layers by Risk Group for the Opt-out Risk Model Funding Layers low risk Funding layers high risk Mean 1.410714286 1.074074074 Variance 0.573701299 0.148148148 Observations 56 27 Hypothesized Mean Difference 0 Df 81 t Stat 2.683978715 P(T<=t) one-tail 0.004409663 t Critical one-tail 1.663883913 P(T<=t) two-tail 0.008819326 t Critical two-tail 1.989686288

PAGE 164

164 Table 6-46. Tenant Characteris tics for the Opt-out Model Higher Opt-out Risk Lower Opt-out Risk Total Properties with data 414.8%4376.8% 47 56.6% Properties without data 2385.2%1323.2% 36 43.4% % Female Head with Children: 0-24% 250.0%920.9% 11 23.4% 25-49% 125.0%49.3% 5 10.6% 50-74% 00.0%2251.2% 22 46.8% 75-100% 125.0%818.6% 9 19.1% % of All Persons with Disability: 0-24% 4100.0%4195.3% 45 95.7% 25-49% 00.0%24.7% 2 4.3% 50-74% 00.0%00.0% 0 0.0% 75-100% 00.0%00.0% 0 0.0% % 62 Year of Age or More: 0-24% 125.0%3069.8% 31 66.0% 25-49% 00.0%49.3% 4 8.5% 50-74% 125.0%37.0% 4 8.5% 75-100% 250.0%614.0% 8 17.0% % 85 Year of Age or More: 0-24% 4100.0%43100.0% 47 100.0% 25-49% 00.0%00.0% 0 0.0% 50-74% 00.0%00.0% 0 0.0% 75-100% 00.0%00.0% 0 0.0% % Minority: 0-24% 00.0%00.0% 0 0.0% 25-49% 00.0%511.6% 5 10.6% 50-74% 00.0%818.6% 8 17.0% 75-100% 4100.0%3069.8% 34 72.3% Annual Household Income: 3000 00.0%12.3% 1 2.1% 4000 00.0%00.0% 0 0.0% 5000 00.0%24.7% 2 4.3% 6000 00.0%24.7% 2 4.3% 7000 250.0%1125.6% 13 27.7% 8000 00.0%818.6% 8 17.0% 9000 125.0%818.6% 9 19.1% 10000 125.0%716.3% 8 17.0% 11000 00.0%12.3% 1 2.1% 12000 00.0%37.0% 3 6.4%

PAGE 165

165 Table 6-46. Continued Higher Opt-out Risk Lower Opt-out Risk Total % Very Low Income: 0-24% 00.0%00.0% 0 0.0% 25-49% 00.0%00.0% 0 0.0% 50-74% 00.0%00.0% 0 0.0% 75-100% 4100.0%43100.0% 47 100.0% % Extremely Low Income: 0-24% 00.0%00.0% 0 0.0% 25-49% 00.0%00.0% 0 0.0% 50-74% 250.0%1841.9% 20 42.6% 75-100% 250.0%2558.1% 27 57.4% Household Income as % of Local Median Family Income: 0-24% 375.0%3479.1% 37 78.7% 25-49% 125.0%920.9% 10 21.3% 50-74% 00.0%00.0% 0 0.0% 75-100% 00.0%00.0% 0 0.0% Average Monthly Rent with Utilities: $97-149 250.0%1944.2% 21 44.7% $150-199 250.0%2148.8% 23 48.9% $200-249 00.0%37.0% 3 6.4% $250-275 00.0%00.0% 0 0.0% Average Years Since Moved in: 2-5 00.0%2865.1% 28 59.6% 6-9 375.0%1227.9% 15 31.9% 10-13 125.0%37.0% 4 8.5% % of Households Overhoused: 0-24% 250.0%1023.3% 12 25.5% 25-49% 125.0%1227.9% 13 27.7% 50-74% 125.0%1637.2% 17 36.2% 75-100% 00.0%511.6% 5 10.6% Table 6-47. Comparison of Rent s by County and Bedroom Size 0 br 1br 2br 3br 4br Duval FMR '08 6167018161024 1173 Simulated Market 628628790976 1205 FMR '09 6857799071138 1304 MiamiDade FMR '08 75385310351324 1547 Simulated Market 85195412941701 2160 FMR '09 84295311561479 1728

PAGE 166

166 CHAPTER 7 SUMMARY, CONCLUSIONS AND RE COMMENDATIONS Summary and Conclusions: Fail-out Risk A property was classified as higher risk of fail-out if it met one of the following three criteria: Mean net operating income be low $200 per unit per month, or Mean debt coverage ratio below 1.0, or REAC Physical Inspection Score below 60. The majority of the higher risk group was composed of properties that only met the NOI condition. The significance tests f ound statistically significant diffe rences in characteristics between the higher and lower risk groups. The anal ysis concluded that pr operties at higher risk of fail-out had the following characteristics compared to the lowe r risk properties: Smaller unit sizes by bedrooms Non-profit ownership Earlier year built Lower project rent to FMR ratio Larger proportion of contracts under LMSA program Contract renewed at least once (not original term) More recent contract effective year Earlier contract expiration year More additional program layers (Section 236) More households at age 62 and older Fewer households with a female head and children Smaller proportion of minority households Smaller proportion of overhoused households Several of these variables were related. The LMSA program was designed to supplement the income of Section 236 properties, which were built by non-profits and limited dividend (forprofit) entities in th e late 1960s and early 1970s. Contracts under the LMSA program generally had shorter terms than those under the Secti on 8 NC/SR programs, hence the finding that contracts have been renewed, which in turn can explain the more recent c ontract effective year.

PAGE 167

167 The finding that the properties at higher risk served a larger number of tenants at age 62 can explain the smaller unit sizes and the lower number of single-mom families. The smaller unit size can explain the smaller propo rtion of overhoused households. Higher risk properties were also found to have the following characteristics, which were similar (not statistically different) to those of properties at lower risk of fail-out: Family as well as elderly target population Smaller (less than 50 units) as well as larger (more than 200 units) property size All very low income households Large proportion of extremely low income households To gain additional insight into the differen ces in characteristics between properties at higher risk of fail-out and those at lower ris k, multiple regression analysis was conducted with the mean NOI as the dependent variable. Since almost 69% of the properties flagged at higher risk met the lower NOI condition, the regression anal ysis was expected to provide further insight into the variables that impacted the level of NOI The analysis calculated that project rent to FMR and year built were statistica lly significant; the lower the project rent to FMR or the earlier the year built, the lower the mean NOI. The signif icance tests also found that the project rent to FMR and year built were statistica lly different between the propert ies at higher risk of fail-out and those at lower risk; the higher risk properties had a lower proj ect rent to FMR and an earlier year built. The findings that project rent to FMR and year built were significant were consistent with the outcomes of analysis by Abt Associates, Inc. In a national study for HUD (Finkel et al. 2006) and a Florida study for the University of Flor ida (Finkel and Lam 2008), Abt concluded that a larger percentage of HUD propertie s in distress or foreclosure had a lower project rent to FMR of below 80% and an earlier year built of before 1975. In both st udies, Abt also found that the majority of these distressed properties were funded under older HUD mortgage programs such as

PAGE 168

168 Section 236, had a rental assistance contract under the LMSA program, and had the lowest REAC Physical Inspection Score compared to other categories of properties. To identify properties with project-based rent al assistance at fail -out risk, the argument can be made that a risk assessment method can simply be based on a small number of key indicators: Project rent to Fair Market Rent ratio and year bui lt. Based on the results of the significance tests, the stepwise re gression analysis and the studies by Abt, these indicators could be sufficient to create a shortlis t of properties at heightened ri sk. In addition, properties with a failing REAC score should be considered at risk A more sophisticated or complex approach to flag properties may not add much value. It may also not be realistic, considering the current data limitations concerning the actua l financial and physical conditions of properties. A shortlist based on the three key indicators is a first step towards actual preservation of properties and protection of tenants. It could provide insight in to the magnitude of properties that could be lost and that should receive attention from policy-make rs and advocates. The next step is for policy-makers to acknowledge the need for preservation and to allocate resources. The next step is also for advocates to approach properties on the short list to assess the feasibility of preservation and the willingness of the owner to part icipate; it is during this step that detailed property-level data can be co llected concerning items such as operating expenses, reserve accounts and capital needs, provided an owner is interested in particip ating or selling the property to a preserving entity. Summary and Conclusions: Opt-out Risk For the opt-out model, a property was considered at higher risk if it met all of the following criteria: For-profit ownership, and Increase in mean NOI by at least 20% when all market rents were charged, and

PAGE 169

169 Expiration of the rental assistan ce contract by year end 2014, and Original contract term, and Not located in a low poverty census tract. The descriptive analysis and significance te sts of the opt-out risk model found that the properties at higher risk had the following charac teristics compared to those at lower risk:1 Smaller property size Larger proportion of elderly households Smaller unit sizes by bedrooms Later year built Lower REAC scores More contracts funded under th e Section 8 NC/SR program Smaller number of program layers The properties at higher risk of opt-out were more likely to serve the elderly, which could also explain the smaller units. The later year built could be related to the smaller property size; in later years, properties were built with fewer units. A significance test found no statistical diffe rence in the project rent to FMR ratios between the risk groups. The majority of properties had project re nts that were lower than FMR. The finding that more properties at higher risk of opt-out had el derly as the target population was in contrast to the fi nding in the national report by Ab t that elderly properties were less likely to opt-out than fam ily properties (Finkel et al. 2006). However, the Abt analysis for the Florida sample did not indicat e statistical significance of the target population to the opt-out decision (Finkel and Lam 2008). The large number of elderly properties flagge d at higher risk of opt -out can be explained by one of the conditions that had to be met as pa rt of the opt-out risk model: Original contract term. This condition was established based on th e assumption that properties that have had a chance to renew their current rent al assistance contract at least once are not interested in the 1 Not enough data were available to compare tenant characteristics.

PAGE 170

170 conversion to market-rate housing, or they would have opted out of the contract. The condition was also based on the assumption that a wave of opt-outs will occur when rental assistance contracts with original terms hit their expiration dates and th eir first opportunity to make the optout/renew decision, as was experienced in the late 1990s. Analysis of the properties that had an original contract term revealed that almost 69% had elderly as the target population. This would explain the larger number of elderly properties in the higher risk group. The characteristics found to be significantly different betwee n the properties identified at higher and lower risk of opt-out were driven by the conditions that were established to determine level of risk. Therefore, these characteristics (e .g., elderly target populatio n) should not be used as criteria to shortlist prop erties at risk of opt-out. As was confirmed by the correlation statistics the lower the project rent to FMR ratio, the larger the change in net operating income when all units are leased at market rent rather than project rent. Therefore, when identifying propertie s at risk of opt-out, the first key indicator should be the project rent to Fair Market Rent. Once a shortlist is created of properties by project rent to FMR, additional criteria can be applied to further specify th e properties at risk of opt-out such as the expiration year a nd contract renewal history. Project rent to FMR was found to be the ke y explanatory variable in both the national study and the analysis for Florid a conducted by Abt. The key e xplanatory variable is the rent-to-FMR ratio. It explains the largest share of variations in the pr obability of opting out, suggesting that a propertys preopt-out rent relative to the lo cal market rent is the most important determinant of the owne rs opt-out decision, controlling for all other ch aracteristics. When the Section 8 rent is signi ficantly below the market leve l (proxy by FMR), owners realize

PAGE 171

171 that a conversion to market rate units can increas e the rental revenues (and therefore profits) with little effect on vacancy rate s (Finkel and Lam 2008, 17). Recommendations for Policy The first policy recommendation is that th e federal government allocate more funding to the Section 8 project-base d rental assistance program. This is essential, considering the need for housing and the risk of losing part of the a ffordable stock. Additional funding will have two purposes. The first purpose is to provide incent ives to current owners to renew the rental assistance contract upon expiration. Incentives could include financial resources for rehabilitation. Another incentive co uld be to adjust the project re nt to 100% of the Fair Market Rent, if the project rent to FMR is currently lo wer. The second purpose is to provide funding to new owners for acquisition and rehabilitation of pr operties with a project-based rental assistance contract. While additional funding is a valid recomme ndation, it may not be realistic considering todays economic climate and federal budget deficit. The Center on Budget and Policy Priorities estimated that approximately $8.2 billion is requi red for the Section 8 project-based rental assistance program in 2010; $7.9 billion for the rene wal of existing contracts and the remaining funds to expand the program. In lieu of the allocati on of additional federal funds to expand the program, the recommendation is for the federal government to sustain the current level of funding for the renewal of contracts and to commit to making subsidy payments on time and in full. During 2008, HUD experienced insufficient appropriations to fund the renewal of rental assistance contracts for a full year and was also maki ng late subsidy payments (Bodaken 2008). These conditions add another reason for pr operty owners to consider opting out of a rent al assistance contract, thereby increasing the opt-out risk.

PAGE 172

172 A recommendation is also for governments to increase the accessibil ity to property-level data. Federal, state and local governments that administer housing programs collect massive amounts of data as part of the initial loan or contract origination pro cess and the subsequent program compliance and contract renewal process. It is suggested that th ey make data publicly available in an accessible format for the purpose of research and preservation. These data should be at the property-level and are to be comprehe nsive in terms of the variables that are reported (including tenant data and financial data). The information should be current and updated regularly. Comprehensive historical data relate d to terminated subsidies and past property characteristics would also be essential, allowing fo r the research of histor ical patterns of fail-outs and opt-outs. Recommendations for Future Research It is recommended that the research is e xpanded to other geographical areas. Future research could focus on all metropolitan areas in Fl orida, or could encompass all counties in the state. Expanding the research to incorporate other juri sdictions would increase the primary data collection effort, but it would also provide insight into the level of risk of a larger population of properties with project-based re ntal assistance contracts. The research could even expand beyond Florida. Rental assistance cont racts could be compared across metropolitan areas in multiple states or across entire states. The limitation is that many other areas outside of Florida do not have a database comparable to the Assisted Housing Inventory, which contains data that complement the HUD public datasets and that ar e essential to the res earch. Therefore, any comparative research on areas outside of Florid a should focus on the jurisdictions that have extensive public data sources. A related recommendation is to expand th e research to othe r funding programs. Properties subsidized by other programs are also at risk of fail-out due to large capital needs and

PAGE 173

173 deterioration or at risk of opt -out due to prepayment of mortgages or termination of use restrictions. All subsid ized properties contribute to the mu ch needed affordable housing stock, whether they provide deep or shallow subsidies. Therefore, the preservation of all types of subsidized properties is important. Expanding th e research to other funding programs increases the complexity of the analysis, especially if many properties have multiple funding layers with multiple options for the termination of affordability. It also requires substantial property data collection, because public data sources are limite d for programs under funding sources other than HUD; even the public HUD datasets do not report on several critical data fields and various housing programs. Another recommendation for future research is to pursue the collection of historical data on Section 8 project-based rental assistance contracts that have been terminated. These data would have to include detailed property-level in formation such as address, type of ownership, target population, REAC score, pr oject rent to FMR ratio, origin al contract effective date, contract renewal dates and opt-out date. Data on the financial and physical condition of the property at the time of opt-out would also be va luable. The historical data would allow for the research of properties that already made a decisi on not to renew the contract. The analysis could provide insight into the indicator s of actual loss. Abt Associates Inc. in a study for HUD (Finkel et al. 2006) conducted analysis of the Section 8 opt-outs and opt-ins, and compared property characteristics. Abt performed the same analysis fo r Florida under contract with the University of Florida (Finkel and Lam 2008). However, under th e contractual agreement with HUD, Abt could not share the raw historical data. Therefore, to acquire the raw data would allow for additional research about the opt-outs. The historical data would have to be obtained from HUD.

PAGE 174

174 Alternatively, case studies can be conducted and property data can be collected from owners or property management firms. Validation of the risk assessment method applied in this research is also a recommendation. First, it is recommended that the outcomes presented in this study are compared to the results of other risk assessment tools that have analyzed the assisted housing stock in Florida. This comparative analysis can assess if the various methods identify the same properties, the same types of properties in terms of characteristics, and the same magnitude of properties at risk. Secon d, if detailed historical data can be obtained about the properties that already opted out of a rental a ssistance contract, the risk assessm ent method could be tested to assess if the properties would have been identified at higher risk of loss.

PAGE 175

175 APPENDIX A INPUT VARIABLES AND SIMULATION RESULTS

PAGE 176

176 Table A-1. Input variable s and simulation results Input Variable Model/ Scenario Probability Graph Mean Min 5% 25% 50% 75% 95% Max Vacancy Loss and Bad Debt Allowance All Uniform 7.50% 5.00% 5.25% 6.25% 7.50% 8.75% 9.75% 10.00% Operating Expense Ratio: Section 236 All Uniform 73.80% 65.71% 66.50% 69.75% 73.79% 77.84% 81.09% 81.89% Operating Expense Ratio: Section 8 Family (1965-77) All Uniform 62.10% 55.60% 56.24% 58.84% 62.09% 65.34% 67.95% 68.59% Operating Expense Ratio: Section 8 Family (1978-06) All Uniform 61.35% 54.11% 54.82% 57.72% 61.34% 64.97% 67.87% 68.59% Operating Expense Ratio: Section 8 Elderly/Persons w Disab. (1965-77) All Uniform 55.20% 50.71% 51.15% 52.95% 55.19% 57.45% 59.25% 59.69% Operating Expense Ratio: Section 8 Elderly/Persons w Disab. (1978-06) All Uniform 53.75% 50.70% 51.00% 52.22% 53.75% 55.27% 56.49% 56.80% Replacement Reserve All Triangular 29.67 1.24 9.07 20.33 28.93 38.78 51.88 61.50 Cathedral Terrace S. 236 Rent 1 br Fail-out Uniform 594.50 518.07 525.54 556.17 594.49 632.72 663.25 670.94 Fannie E. Taylor Home For The Aged S. 236 Rent 1 br Fail-out Uniform 597.00 550.01 554.66 573.43 596.92 620.46 639.27 643.98 Florida Christian Home S. 236 Rent 0 br Fail-out Uniform 538.50 505.01 508.31 521.72 53 8.48 555.20 568.64 571.97

PAGE 177

177 Florida Christian Home S. 236 Rent 1 br Fail-out Uniform 596.00 563.05 566.26 579.45 595.95 612.46 625.70 628.95 Miami Beach Marian Towers S. 236 Rent 1 br Fail-out Uniform 508.00 490.02 491.78 498.97 507.97 516.99 524.18 525.97 Mt Carmel Gardens S. 236 Rent 0 br Fail-out Uniform 420.50 408.00 409.23 414.23 420.48 426.74 431.75 432.98 Mt Carmel Gardens S. 236 Rent 1 br Fail-out Uniform 511.00 494.01 495.68 502.50 510.97 519.47 526.27 527.97 Pablo Towers S. 236 Rent 1 br Fail-out Uniform 667.50 631.06 634.58 649.20 667.47 685.73 700.28 703.96 Riverside Presbyterian Apartments S. 236 Rent 0 br Fail-out Uniform 515.50 476.07 479.94 495.74 515.48 535.25 551.05 554.96 Riv erside Presbyterian Apartments S. 236 Rent 1 br Fail-out Uniform 669.50 618.09 623.09 643.68 669.45 695.21 715.83 720.98 The Towers Of Jacksonville S. 236 Rent 1 br Fail-out Uniform 661.50 555.08 565.47 608.09 661.48 714.66 757.29 767.79 Town Park Plaza South, Inc. S. 236 Rent 1 br Fail-out Uniform 457.00 418.06 421.87 437.43 456.98 476.46 492.09 495.94 Town Park Plaza South, Inc. S. 236 Rent 2 br Fail-out Uniform 474.00 435.03 438.86 454.44 473.96 493.46 509.04 512.98 Town Park Plaza South, Inc. S. 236 Rent 3 br Fail-out Uniform 510.00 468.07 472.16 488.98 509.99 530.92 54 7.72 551.95

PAGE 178

178 Town Park Plaza South, Inc. S. 236 Rent 4 br Fail-out Uniform 513.50 471.07 475.20 492.19 513.48 534.74 551.71 555.94 Town Park Village I Market Rent 3 br Fail-out Uniform 631.50 621.01 622.05 626.23 631.49 636.74 640.94 641.98 Caroline Arms Apartments Market Rent 2 br Fail-out; Opt-out @ Assisted Uniform 550.00 516.04 519.39 532.99 549.99 566.94 580.58 584.00 Caroline Arms Apartments Market Rent 3 br Fail-out; Opt-out @ Assisted Uniform 563.50 529.06 532.43 546.25 563.48 580.69 594.52 597.98 Caroline Arms Apartments Market Rent 4 br Fail-out; Opt-out @ Assisted Uniform 580.00 545.03 548.46 562.45 580.00 597.47 611.49 614.94 Fieldcrest Apartments Market Rent 2 br Fail-out; Opt-out @ Assisted Uniform 834.00 789.05 79 3.46 811.43 833.96 856.41 874.42 879.00 Monaco Arms Apartments II Market Rent 1 br Fail-out; Opt-out @ Assisted Uniform 497.00 480.01 481.67 488.47 497.00 505.50 512.29 513.97 Monaco Arms Apartments II Market Rent 2 br Fail-out; Opt-out @ Assisted Uniform 554.00 535.02 536.88 544.49 554.00 563.49 571.08 572.98 Monaco Arms I Market Rent 1 br Fail-out; Opt-out @ Assisted Uniform 498.00 480.01 481.79 488.97 497.97 506.97 514.20 516.00 Monaco Arms I Market Rent 2 br Fail-out; Opt-out @ Assisted Uniform 550.50 535.00 536.54 542.73 550.49 55 8.24 564.42 566.00 Duval Market Rent 0 br Opt-out @ Market Lognormal 628.11 427.55 473.43 529.84 592.28 684.99 899.54 1,862.1 0

PAGE 179

179 Duval Market Rent 1 br Opt-out @ Market Lognormal 628.10 456.45 501.72 553.13 605.45 677.15 829.08 1,447.2 4 Duval Market Rent 2 br Opt-out @ Market Lognormal 789.47 521.48 619.29 698.45 770.14 858.84 1,025.2 3 1,419.6 3 Duval Market Rent 3 br Opt-out @ Market Lognormal 976.05 693.77 772.08 859.19 944.20 1,056.8 4 1,285.1 0 2,053.0 3 Duval Market Rent 4 br Opt-out @ Market Lognormal 1,205.1 0 824.08 921.60 1,034.1 0 1,151.5 2 1,314.4 8 1,667.3 6 3,018.2 4 Miami-Dade Market Rent 0 br Opt-out @ Market Lognormal 850.67 552.20 602.87 680.67 777.30 933.26 1,338.6 7 2,997.2 2 Miami-Dade Market R ent 1 br Opt-out @ Market Lognormal 954.34 673.71 750.21 834.36 919.04 1,033.2 8 1,274.9 8 2,447.0 0 Miami-Dade Market Rent 2 br Opt-out @ Market Lognormal 1,293.6 3 891.11 980.06 1,095.8 2 1,223.2 7 1,409.6 4 1,839.0 9 3,261.2 9 Miami-Dade Market Rent 3 br Opt-out @ Market Lognormal 1,701.1 6 1,140.8 7 1,239.3 3 1,382.7 7 1,561.3 9 1,853.1 4 2,623.1 2 6,682.3 7 Miami-Dade Market Rent 4 br Opt-out @ Market Lognormal 2,160.5 9 1,244.4 9 1,454.7 1 1,703.3 4 1, 987.0 4 2,413.5 3 3,429.0 7 8,395.8 7

PAGE 180

180 APPENDIX B OUTPUT RESULTS BY PROPERTY, OUTP UT, VARIABLE AND MODEL

PAGE 181

181 Table B-1. Output results by prope rty, output variable and model Output Variable NOI or DCR Model/Scenario Graph Mean Min 5% 25% 50% 75% 95% Max Duval / NOI 1 NOI Fail out 108.46 41.73 61.79 87.13 108.71 129.59 152.92 173.13 Duval / NOI 1 NOI Fail out 173.38 119.40 143.99 161.38 172.52 186.03 202.48 219.04 Duval / NOI 1 NOI Fail out 87.76 23.10 47.45 70.36 88.33 105.93 127.90 143.82 Duval / NOI 1 NOI Fail out 109.43 40.18 63.67 87.70 109.03 129.99 153.94 173.68 Duval / NOI 1 NOI Fail out 145.30 64.86 90.08 118.45 144.98 171.04 198.97 218.66 Duval / NOI 1 NOI Fail out 121.64 47.41 72.69 98.72 121.85 144.68 170.07 190.29 Duval / NOI 1 NOI Fail out 131.69 53.71 79.88 106.77 131.39 155.65 181.97 203.77 Duval / NOI 1 NOI Fail out 190.95 127.04 149.48 172.22 190.81 209.59 232.71 258.32 Duval / NOI 1 NOI Fail out 180.19 117.88 139.98 162.31 179.92 198.11 220.29 245.59 Duval / NOI 1 NOI Fail out 286.71 208.64 233.14 258.82 286.69 312.20 341.58 371.61 Duval / NOI 1 NOI Fail out 290.70 230.97 261.25 277.59 290.37 304.12 320.47 344.05

PAGE 182

182 Duval / NOI 1 NOI Fail out 233.51 163.30 187.18 210.69 232.77 255.06 280.78 308.66 Duval / NOI 1 NOI Fail out 394.14 325.23 359.91 378.07 393.64 410.13 428.88 455.92 Duval / NOI 1 NOI Fail out 200.53 135.20 158.12 180.77 200.37 219.73 243.89 269.65 Miami-Dade / NOI 1 NOI Fail out 204.11 125.05 156.44 183.38 203.13 227.00 248.98 270.01 Miami-Dade / NOI 1 NOI Fail out 120.42 46.47 71.36 97.43 120.83 143.02 168.13 189.39 Miami-Dade / NOI 1 NOI Fail out 116.04 43.41 68.36 94.33 116.12 138.04 162.81 182.44 Miami-Dade / NOI 1 NOI Fail out 83.80 20.91 44.25 66.86 84.26 100.85 123.07 140.74 Miami-Dade / NOI 1 NOI Fail out 245.42 189.72 218.32 233.42 245.19 257.96 273.09 295.08 Miami-Dade / NOI 1 NOI Fail out 172.00 117.62 142.71 160.34 171.39 184.52 201.38 218.98 Miami-Dade / NOI 1 NOI Fail out 259.80 185.70 209.42 234.55 259.75 283.42 310.85 339.77 Miami-Dade / NOI 1 NOI Fail out 156.87 98.00 119.39 140.59 156.80 173.06 194.13 218.00

PAGE 183

183 Miami-Dade / NOI 1 NOI Fail out 362.64 296.53 329.80 347.46 362.29 377.98 395.79 421.86 Miami-Dade / NOI 1 NOI Fail out 237.49 182.49 210.75 225.69 237.37 249.92 265.03 286.51 Miami-Dade / NOI 1 NOI Fail out 263.39 206.09 235.32 250.83 263.07 276.34 291.57 314.52 Miami-Dade / NOI 1 NOI Fail out 277.47 218.92 248.69 264.65 277.20 290.61 306.44 329.75 Miami-Dade / NOI 1 NOI Fail out 236.61 181.69 209.87 224.84 236.51 249.05 264.11 285.55 Miami-Dade / NOI 1 NOI Fail out 255.98 199.34 228.31 243.65 255.84 268.62 283.93 306.50 Miami-Dade / NOI 1 NOI Fail out 270.92 212.95 242.44 258.20 270.73 283.79 299.43 322.66 Miami-Dade / NOI 1 NOI Fail out 259.28 202.35 231.43 246.93 259.12 272.12 287.28 310.07 Miami-Dade / NOI 1 NOI Fail out 314.60 252.75 284.17 300.82 314.15 328.88 345.38 369.90 Miami-Dade / NOI 1 NOI Fail out 245.75 160.07 192.87 220.15 244.47 272.37 296.77 320.74 Miami-Dade / NOI 1 NOI Fail out 270.76 180.14 213.64 242.43 269.42 299.99 325.60 350.60

PAGE 184

184 Miami-Dade / NOI 1 NOI Fail out 261.99 173.10 206.13 234.61 260.42 290.19 315.45 340.13 Miami-Dade / NOI 1 NOI Fail out 271.37 180.63 214.18 242.98 270.02 300.68 326.31 351.33 Miami-Dade / NOI 1 NOI Fail out 371.22 260.75 299.34 333.62 369.60 408.95 443.63 470.50 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 237.33 154.87 185.60 212.05 236.70 263.34 288.61 313.11 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 203.54 126.20 156.75 182.49 202.59 226.50 248.04 270.37 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 215.75 148.39 171.91 194.41 214.96 235.87 260.57 287.73 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 203.43 137.67 160.67 183.38 203.13 222.89 247.13 273.08 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 210.56 143.68 165.97 190.16 210.18 230.49 255.13 280.33 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 224.18 155.17 179.16 202.17 223.77 245.26 270.09 297.40 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 224.90 155.97 179.59 202.74 224.49 245.87 271.08 298.48 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 149.75 91.93 113.11 134.15 149.64 165.31 186.16 209.57

PAGE 185

185 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 212.64 145.52 168.79 191.89 212.22 232.75 257.06 283.98 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 255.85 168.18 201.11 229.13 254.38 283.50 308.45 332.80 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 255.68 168.04 200.98 228.99 254.22 283.33 308.26 332.61 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 204.60 127.05 157.61 183.43 203.63 227.68 249.28 271.63 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 230.24 147.63 179.25 206.34 228.64 255.51 278.36 302.24 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 224.89 171.01 198.28 213.13 224.95 237.22 251.79 272.87 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 212.71 145.58 168.85 191.95 212.28 232.82 257.14 284.06 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 191.45 127.47 149.92 172.67 191.32 210.12 233.30 258.92 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 248.42 176.01 200.02 224.01 247.91 271.09 297.60 326.30 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 218.58 150.58 174.02 197.14 218.01 239.12 263.93 291.00 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 218.53 150.54 173.98 197.11 217.97 239.08 263.88 290.95

PAGE 186

186 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 226.59 144.70 176.25 203.07 225.06 251.53 274.02 297.88 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 230.94 161.11 184.91 208.48 230.26 252.33 277.93 305.62 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 223.46 154.74 178.32 201.49 222.98 244.33 269.41 296.77 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 243.80 158.51 191.20 218.39 242.65 270.20 294.43 318.42 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 205.98 139.85 162.92 185.79 205.70 225.53 249.85 276.10 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 268.09 192.77 216.67 241.84 267.81 292.26 320.46 349.57 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 216.44 148.76 172.14 195.27 215.91 236.87 261.46 288.48 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 174.72 113.22 135.15 157.48 174.43 192.11 214.15 239.12 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 158.25 99.18 120.61 141.84 158.14 174.62 195.68 219.63 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 181.57 119.05 141.20 163.52 181.32 199.62 221.83 247.22 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 242.90 157.78 190.42 217.58 241.79 269.26 293.35 317.34

PAGE 187

187 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 209.96 143.24 166.43 189.49 209.61 230.05 254.10 280.81 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 222.73 141.60 173.00 199.72 221.35 247.44 269.55 293.27 Duval / NOI 1 NOI Fail out; Opt-out @ Assisted 185.10 122.06 144.31 166.69 184.75 203.39 225.80 251.40 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 266.69 191.58 215.43 240.61 266.44 290.69 318.88 347.92 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 281.54 204.23 228.61 253.98 281.61 306.63 335.72 365.49 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 302.04 221.70 246.12 272.24 301.78 328.60 358.98 389.74 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 253.14 180.03 203.92 228.37 252.93 276.25 303.10 331.89 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 258.83 184.88 208.63 233.66 258.82 282.39 309.73 338.62 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 301.28 204.64 240.02 270.16 299.79 333.34 361.62 387.03 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 176.34 85.93 113.68 144.37 175.67 206.87 237.85 260.32 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 342.56 278.23 310.70 327.91 342.14 357.44 374.65 400.14

PAGE 188

188 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 280.31 213.51 243.60 263.77 279.29 296.72 317.02 337.62 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 378.09 310.61 344.70 362.43 377.71 393.87 412.00 438.57 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 327.21 264.24 296.27 313.02 326.71 341.65 358.62 383.54 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 257.79 183.99 207.76 232.78 257.81 281.28 308.51 337.39 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 270.93 180.28 213.79 242.58 269.59 300.18 325.79 350.80 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 300.18 239.61 270.41 286.69 299.72 313.72 330.48 354.31 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 292.15 197.31 232.20 261.82 290.38 323.08 350.92 376.13 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 296.73 200.98 236.22 266.03 295.22 328.19 356.36 381.60 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 240.96 156.23 188.77 215.82 239.78 267.25 291.04 315.04 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 336.71 233.06 269.89 302.15 335.25 372.20 403.08 429.32 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 402.68 333.01 367.86 386.46 402.15 418.95 437.92 465.15

PAGE 189

189 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 276.99 218.48 248.22 264.16 276.73 290.09 305.91 329.22 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 285.17 225.93 255.94 272.26 284.80 298.55 314.61 338.06 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 238.63 154.36 186.76 213.78 237.34 264.74 288.24 312.25 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 271.29 213.29 242.79 258.57 271.11 284.16 299.81 323.06 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 305.61 244.56 275.57 291.94 305.09 319.51 336.09 360.18 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 196.65 145.28 171.04 185.10 196.98 208.14 222.33 242.34 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 270.58 212.64 242.12 257.87 270.38 283.46 299.07 322.29 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 263.06 205.79 235.00 250.49 262.72 275.98 291.22 314.15 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 252.69 196.34 225.20 240.43 252.46 265.31 280.60 302.95 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 263.39 206.09 235.32 250.83 263.07 276.34 291.57 314.52 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 227.45 173.35 200.83 215.74 227.46 239.76 254.50 275.65

PAGE 190

190 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 265.67 208.17 237.48 253.08 265.38 278.60 293.95 316.98 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 241.21 185.88 214.26 229.30 241.02 253.63 268.89 290.53 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 220.18 166.72 193.74 208.49 220.35 232.43 246.81 267.78 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 292.65 232.75 263.12 279.50 292.27 306.08 322.52 346.15 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 312.94 251.24 282.58 299.22 312.40 327.13 343.66 368.10 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 248.72 192.72 221.44 236.58 248.52 261.36 276.45 298.65 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 296.33 236.11 266.75 283.01 295.90 309.78 326.38 350.14 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 222.49 168.82 195.92 210.76 222.64 234.74 249.25 270.28 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 346.42 281.74 314.37 331.64 345.90 361.34 378.72 404.31 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 303.58 206.48 241.96 272.27 302.18 335.93 364.26 389.77 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 262.28 173.34 206.37 234.87 260.72 290.54 315.78 340.48

PAGE 191

191 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 272.05 213.98 243.50 259.32 271.88 284.94 300.60 323.88 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 252.76 165.70 198.60 226.38 251.27 280.21 304.93 329.12 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 209.24 130.77 161.58 187.58 208.29 232.76 254.68 277.17 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 264.76 207.34 236.61 252.17 264.46 277.70 293.00 316.00 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 245.25 159.68 192.45 219.71 244.01 271.81 296.18 320.16 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 245.42 189.72 218.32 233.42 245.19 257.96 273.09 295.08 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 302.89 242.08 272.99 289.34 302.37 316.65 333.28 357.24 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 322.12 221.36 257.89 288.53 320.85 356.42 385.67 411.90 Miami-Dade / NOI 1 NOI Fail out; Opt-out @ Assisted 242.43 186.99 215.46 230.48 242.20 254.88 270.10 291.84 Duval / NOI 2 NOI Opt out @ Market 232.76 111.87 172.03 202.27 228.77 259.16 302.34 501.45 Duval / NOI 2 NOI Opt out @ Market 223.29 103.06 159.82 191.38 219.97 250.74 295.19 478.55

PAGE 192

192 Duval / NOI 2 NOI Opt out @ Market 271.43 163.59 207.88 240.83 269.49 299.92 342.51 411.83 Duval / NOI 2 NOI Opt out @ Market 257.92 151.24 192.90 226.11 254.87 287.35 332.20 401.00 Duval / NOI 2 NOI Opt out @ Market 230.94 124.35 165.64 199.50 225.36 258.41 310.77 406.16 Duval / NOI 2 NOI Opt out @ Market 228.99 122.13 164.37 198.12 223.51 255.51 306.86 395.03 Duval / NOI 2 NOI Opt out @ Market 275.46 169.03 207.32 242.06 272.35 305.31 354.86 447.04 Duval / NOI 2 NOI Opt out @ Market 221.34 113.41 159.46 191.96 215.88 246.47 293.57 406.61 Duval / NOI 2 NOI Opt out @ Market 307.46 181.06 233.30 271.91 305.42 336.54 399.19 546.87 Duval / NOI 2 NOI Opt out @ Market 239.92 119.64 176.77 207.33 236.76 268.24 314.93 477.85 Duval / NOI 2 NOI Opt out @ Market 278.79 157.85 205.29 240.51 273.75 310.89 370.80 480.11 Duval / NOI 2 NOI Opt out @ Market 253.29 133.12 188.67 218.74 248.70 282.56 331.15 471.23 Duval / NOI 2 NOI Opt out @ Market 252.82 133.62 172.95 210.98 246.90 284.67 360.09 529.35

PAGE 193

193 Duval / NOI 2 NOI Opt out @ Market 239.03 120.63 176.25 204.28 232.03 263.31 325.24 577.60 Duval / NOI 2 NOI Opt out @ Market 275.79 169.76 206.77 241.52 271.84 305.86 357.57 447.86 Duval / NOI 2 NOI Opt out @ Market 278.97 173.38 207.40 243.05 274.68 309.30 368.69 491.62 Duval / NOI 2 NOI Opt out @ Market 281.24 174.89 208.50 244.09 277.34 312.49 371.41 494.65 Duval / NOI 2 NOI Opt out @ Market 289.11 178.62 215.12 250.49 283.52 319.56 387.80 549.63 Duval / NOI 2 NOI Opt out @ Market 289.11 178.62 215.12 250.49 283.52 319.56 387.80 549.63 Duval / NOI 2 NOI Opt out @ Market 267.67 146.67 199.24 232.22 262.00 297.74 351.06 499.73 Duval / NOI 2 NOI Opt out @ Market 276.21 169.88 207.74 242.37 273.09 305.56 356.12 448.03 Duval / NOI 2 NOI Opt out @ Market 247.14 122.88 168.78 209.01 239.72 278.63 346.83 498.92 Duval / NOI 2 NOI Opt out @ Market 247.26 127.16 183.83 213.32 243.07 275.21 325.07 469.98 Duval / NOI 2 NOI Opt out @ Market 241.11 126.76 167.98 206.73 235.12 271.55 333.93 464.38

PAGE 194

194 Duval / NOI 2 NOI Opt out @ Market 293.04 180.49 218.35 254.01 286.76 323.73 397.45 577.13 Duval / NOI 2 NOI Opt out @ Market 247.14 122.88 168.78 209.01 239.72 278.63 346.83 498.92 Duval / NOI 2 NOI Opt out @ Market 215.63 106.91 156.78 188.26 210.41 239.32 286.04 421.11 Duval / NOI 2 NOI Opt out @ Market 190.42 78.20 131.71 158.23 182.89 213.30 272.90 485.22 Duval / NOI 2 NOI Opt out @ Market 239.39 131.19 180.86 211.79 236.28 264.66 305.83 417.18 Duval / NOI 2 NOI Opt out @ Market 282.02 158.98 206.23 241.48 273.13 314.87 383.77 572.98 Duval / NOI 2 NOI Opt out @ Market 307.46 181.06 233.30 271.91 305.42 336.54 399.19 546.87 Duval / NOI 2 NOI Opt out @ Market 252.82 133.62 172.95 210.98 246.90 284.67 360.09 529.35 Duval / NOI 2 NOI Opt out @ Market 232.09 128.44 173.72 203.85 228.70 258.09 295.32 358.08 Miami-Dade / NOI 2 NOI Opt out @ Market 446.23 273.92 328.49 383.96 430.16 491.13 623.01 1151.53 Miami-Dade / NOI 2 NOI Opt out @ Market 440.06 264.36 318.65 372.76 419.08 486.81 623.29 950.26

PAGE 195

195 Miami-Dade / NOI 2 NOI Opt out @ Market 518.83 315.05 376.88 443.38 496.79 571.27 733.32 1608.08 Miami-Dade / NOI 2 NOI Opt out @ Market 406.40 246.07 300.07 350.53 391.82 447.39 550.35 772.79 Miami-Dade / NOI 2 NOI Opt out @ Market 406.05 245.61 299.72 350.04 391.18 447.15 551.71 769.32 Miami-Dade / NOI 2 NOI Opt out @ Market 438.70 248.56 314.93 371.29 424.01 490.79 615.55 1054.07 Miami-Dade / NOI 2 NOI Opt out @ Market 289.69 119.02 175.70 227.93 277.34 334.97 459.47 803.90 Miami-Dade / NOI 2 NOI Opt out @ Market 378.44 239.54 280.99 326.48 362.31 411.03 531.92 969.30 Miami-Dade / NOI 2 NOI Opt out @ Market 322.86 169.08 208.30 251.93 294.28 359.76 538.96 1166.86 Miami-Dade / NOI 2 NOI Opt out @ Market 367.30 228.44 280.08 322.73 355.85 399.18 498.56 794.18 Miami-Dade / NOI 2 NOI Opt out @ Market 366.72 227.67 280.01 321.91 354.58 398.77 498.09 784.96 Miami-Dade / NOI 2 NOI Opt out @ Market 421.35 255.24 310.75 362.26 405.50 464.75 577.83 855.02 Miami-Dade / NOI 2 NOI Opt out @ Market 373.69 222.19 267.44 315.82 362.57 419.22 510.93 847.68

PAGE 196

196 Miami-Dade / NOI 2 NOI Opt out @ Market 378.44 239.54 280.99 326.48 362.31 411.03 531.92 969.30 Miami-Dade / NOI 2 NOI Opt out @ Market 444.22 252.15 321.57 377.33 429.62 494.14 610.43 1018.71 Miami-Dade / NOI 2 NOI Opt out @ Market 529.13 296.34 387.98 450.41 516.12 587.43 717.72 1339.29 Miami-Dade / NOI 2 NOI Opt out @ Market 311.54 158.28 214.65 259.87 300.60 352.27 441.23 802.37 Miami-Dade / NOI 2 NOI Opt out @ Market 410.64 243.67 300.32 349.84 396.94 459.61 550.83 921.90 Miami-Dade / NOI 2 NOI Opt out @ Market 370.54 232.68 282.43 324.56 357.89 402.82 502.79 845.03 Miami-Dade / NOI 2 NOI Opt out @ Market 357.40 215.47 269.96 310.94 340.67 391.27 485.10 782.43 Miami-Dade / NOI 2 NOI Opt out @ Market 350.97 207.05 261.31 298.82 332.47 385.56 499.97 918.86 Miami-Dade / NOI 2 NOI Opt out @ Market 311.54 158.28 214.65 259.87 300.60 352.27 441.23 802.37 Miami-Dade / NOI 2 NOI Opt out @ Market 378.44 239.54 280.99 326.48 362.31 411.03 531.92 969.30 Miami-Dade / NOI 2 NOI Opt out @ Market 372.12 234.75 282.73 325.04 357.95 403.41 505.46 869.89

PAGE 197

197 Miami-Dade / NOI 2 NOI Opt out @ Market 340.69 193.59 236.06 278.48 315.38 372.30 520.07 1136.98 Miami-Dade / NOI 2 NOI Opt out @ Market 370.14 232.16 281.81 324.20 358.04 403.20 502.42 838.82 Miami-Dade / NOI 2 NOI Opt out @ Market 426.08 299.64 340.29 379.12 415.16 459.47 552.02 782.59 Miami-Dade / NOI 2 NOI Opt out @ Market 378.44 239.54 280.99 326.48 362.31 411.03 531.92 969.30 Miami-Dade / NOI 2 NOI Opt out @ Market 334.18 184.34 218.24 262.05 305.75 371.27 548.63 1275.08 Miami-Dade / NOI 2 NOI Opt out @ Market 334.18 184.34 218.24 262.05 305.75 371.27 548.63 1275.08 Miami-Dade / NOI 2 NOI Opt out @ Market 371.06 233.37 282.73 325.10 357.62 403.19 504.00 853.32 Miami-Dade / NOI 2 NOI Opt out @ Market 371.06 233.37 282.73 325.10 357.62 403.19 504.00 853.32 Miami-Dade / NOI 2 NOI Opt out @ Market 334.18 184.34 218.24 262.05 305.75 371.27 548.63 1275.08 Miami-Dade / NOI 2 NOI Opt out @ Market 378.44 239.54 280.99 326.48 362.31 411.03 531.92 969.30 Miami-Dade / NOI 2 NOI Opt out @ Market 385.97 251.47 303.06 341.01 374.78 415.72 515.00 819.67

PAGE 198

198 Miami-Dade / NOI 2 NOI Opt out @ Market 346.82 201.63 251.51 290.87 324.92 383.12 509.70 1006.76 Miami-Dade / NOI 2 NOI Opt out @ Market 378.44 239.54 280.99 326.48 362.31 411.03 531.92 969.30 Miami-Dade / NOI 2 NOI Opt out @ Market 338.60 190.86 231.56 273.63 312.57 372.59 529.34 1181.17 Miami-Dade / NOI 2 NOI Opt out @ Market 399.23 268.11 310.19 352.72 386.83 431.00 531.19 893.97 Miami-Dade / NOI 2 NOI Opt out @ Market 433.35 231.08 292.49 350.54 410.83 486.60 651.65 1325.17 Miami-Dade / NOI 2 NOI Opt out @ Market 405.24 232.72 284.49 337.19 389.02 454.45 580.13 1100.20 Miami-Dade / NOI 2 NOI Opt out @ Market 340.99 193.98 236.61 279.46 316.01 372.64 519.67 1130.60 Miami-Dade / NOI 2 NOI Opt out @ Market 311.54 158.28 214.65 259.87 300.60 352.27 441.23 802.37 Miami-Dade / NOI 2 NOI Opt out @ Market 302.31 156.08 212.12 255.56 292.51 339.54 412.54 657.60 Miami-Dade / NOI 2 NOI Opt out @ Market 351.88 208.25 262.54 301.38 334.32 387.06 494.46 899.43 Miami-Dade / NOI 2 NOI Opt out @ Market 311.54 158.28 214.65 259.87 300.60 352.27 441.23 802.37

PAGE 199

199 Miami-Dade / NOI 2 NOI Opt out @ Market 334.18 184.34 218.24 262.05 305.75 371.27 548.63 1275.08 Miami-Dade / NOI 2 NOI Opt out @ Market 363.69 223.71 278.50 319.16 350.53 397.07 497.53 737.34 Miami-Dade / NOI 2 NOI Opt out @ Market 505.68 281.71 358.28 419.15 487.42 566.21 717.01 1605.20 Miami-Dade / NOI 2 NOI Opt out @ Market 334.18 184.34 218.24 262.05 305.75 371.27 548.63 1275.08 Duval / DCR 1 DCR Fail out 3.06 1.18 1.74 2.46 3.07 3.66 4.32 4.89 Duval / DCR 1 DCR Fail out 2.06 1.42 1.71 1.92 2.05 2.21 2.40 2.60 Duval / DCR 1 DCR Fail out 2.49 0.65 1.34 1.99 2.50 3.00 3.62 4.07 Duval / DCR 1 DCR Fail out 2.42 0.89 1.41 1.94 2.41 2.87 3.40 3.83 Duval / DCR 1 DCR Fail out 3.10 1.38 1.92 2.53 3.10 3.65 4.25 4.67 Duval / DCR 1 DCR Fail out 1.98 0.77 1.18 1.61 1.98 2.35 2.77 3.10 Duval / DCR 1 DCR Fail out 2.54 1.04 1.54 2.06 2.54 3.00 3.51 3.93

PAGE 200

200 Duval / DCR 1 DCR Fail out 3.25 2.37 2.65 2.94 3.25 3.54 3.88 4.22 Duval / DCR 1 DCR Fail out 1.56 1.24 1.40 1.49 1.56 1.63 1.72 1.85 Duval / DCR 1 DCR Fail out 6.68 4.67 5.36 6.03 6.66 7.30 8.04 8.84 Miami-Dade / DCR 1 DCR Fail out 2.84 1.09 1.68 2.29 2.85 3.37 3.96 4.46 Miami-Dade / DCR 1 DCR Fail out 2.51 0.94 1.48 2.04 2.51 2.99 3.52 3.95 Miami-Dade / DCR 1 DCR Fail out 2.40 0.60 1.27 1.91 2.41 2.88 3.52 4.02 Miami-Dade / DCR 1 DCR Fail out 1.31 1.02 1.17 1.25 1.31 1.38 1.46 1.58 Miami-Dade / DCR 1 DCR Fail out 0.83 0.68 0.76 0.80 0.83 0.87 0.91 0.97 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 4.19 2.87 3.33 3.78 4.19 4.59 5.07 5.60 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 1.41 0.92 1.10 1.26 1.40 1.56 1.69 1.83 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 1.47 0.91 1.13 1.31 1.46 1.63 1.79 1.95

PAGE 201

201 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 3.89 2.49 3.03 3.48 3.86 4.32 4.70 5.10 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 3.96 2.72 3.15 3.57 3.95 4.33 4.78 5.27 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 5.94 4.09 4.73 5.36 5.93 6.50 7.18 7.91 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 1.09 0.71 0.86 0.98 1.09 1.21 1.32 1.43 Duval / DCR 1 DCR Fail out; Opt-out @ Assisted 3.93 2.68 3.12 3.55 3.92 4.31 4.76 5.26 Miami-Dade / DCR 1 DCR Fail out; Opt-out @ Assisted 3.25 2.38 2.65 2.93 3.24 3.53 3.86 4.19 Miami-Dade / DCR 1 DCR Fail out; Opt-out @ Assisted 4.81 2.35 3.10 3.94 4.79 5.65 6.49 7.10 Miami-Dade / DCR 1 DCR Fail out; Opt-out @ Assisted 1.18 0.96 1.07 1.13 1.18 1.23 1.29 1.38 Miami-Dade / DCR 1 DCR Fail out; Opt-out @ Assisted 0.90 0.73 0.81 0.86 0.90 0.94 0.99 1.05 Miami-Dade / DCR 1 DCR Fail out; Opt-out @ Assisted 4.14 2.80 3.29 3.71 4.12 4.58 4.97 5.32 Duval / DCR 2 DCR Opt out @ Market 6.06 3.57 4.60 5.36 6.02 6.64 7.87 10.79

PAGE 202

202 Duval / DCR 2 DCR Opt out @ Market 1.32 0.66 0.97 1.14 1.30 1.47 1.73 2.62 Duval / DCR 2 DCR Opt out @ Market 1.81 0.95 1.35 1.57 1.78 2.02 2.37 3.38 Duval / DCR 2 DCR Opt out @ Market 4.27 2.26 2.92 3.56 4.17 4.81 6.08 8.94 Duval / DCR 2 DCR Opt out @ Market 5.23 3.23 3.89 4.53 5.13 5.78 7.02 9.95 Duval / DCR 2 DCR Opt out @ Market 7.86 4.86 5.85 6.81 7.71 8.69 10.55 14.95 Duval / DCR 2 DCR Opt out @ Market 1.27 0.71 0.93 1.09 1.23 1.42 1.73 2.58 Duval / DCR 2 DCR Opt out @ Market 5.76 3.39 4.37 5.09 5.72 6.30 7.47 10.24 Miami-Dade / DCR 2 DCR Opt out @ Market 5.58 3.39 4.05 4.77 5.34 6.14 7.88 17.29 Miami-Dade / DCR 2 DCR Opt out @ Market 7.91 3.25 4.80 6.22 7.57 9.14 12.54 21.94 Miami-Dade / DCR 2 DCR Opt out @ Market 1.30 0.82 0.97 1.12 1.25 1.41 1.83 3.34 Miami-Dade / DCR 2 DCR Opt out @ Market 1.01 0.63 0.77 0.89 0.97 1.10 1.37 2.16

PAGE 203

203 Miami-Dade / DCR 2 DCR Opt out @ Market 7.38 4.13 5.41 6.28 7.20 8.19 10.01 18.67 Note: The amounts noted in th e graphs are per property.

PAGE 204

204 REFERENCES Achtenberg, Em ily P. 2002. Stemming the Tide. A Handbook on Preserving Subsidized Multifamily Housing. New York: Local Initiativ es Support Corporation. http://www.lisc.org/conten t/publications/detail/893/ (accessed October 15, 2007). Achtenberg, Em ily, Edward J. Daly, Bart Ll oyd, Vincent F. ODonnell, and David Whiston. 2005. Recapitalizing Affordable Housing: A Handbook for Nonprofit Owners. New York: Local Initiatives Support Corporation. http://www.lisc.org/conten t/publications/detail/897/ (accessed October 25, 2007). Affordable Housing Study Commiss ion. 2005. Final report 2005. http://www.floridahousing.org/ahsc/AnnualReports.htm (accessed April 17, 2007). Agresti, Alan and Barbara Finlay. 1997. Statistical Methods for the Social Sciences Upper Saddle River, NJ: Prentice Hall, Inc. Albright, S. Christian, Wayne L. Winston, and Christopher J. Zappe. 2006. Data Analysis & Decision Making with Microsoft Excel, Third Edition Mason, OH: Thomson SouthWestern. Archer, Wayne. 2005. Monte Carlo for mortgage analysis. Class handout 10/31/05. University of Florida. Baroni, Michel, Fabrice Barthlmy, and Mahdi Mokrane. 2005. Monte Carlo simulations versus DCP in real estate portfolio valuation. http://www.u-cergy.fr/IMG/2005-15Baroni.pdf (accessed Septem ber 26, 2008). Belsky, Eric S. and Rachel Bogardus Drew. 2006. Taking stock of the nations rental housing challenges and a half century of public polic y responses. Prepared for Revisiting Rental Housing: A National Policy Summit. Join t Center for Housing Studies of Harvard University. http://www.jchs.harvard.edu/publications/ren tal/revisiting_rental _sym posium/papers/inde x.htm (accessed March 3, 2008). Bodaken, Michael. 2008. Testimony of Michael B odaken, President, National Housing Trust, before the Subcommittee on Transportati on, HUD, and Related Agencies. Committee on Appropriations. Hearing on the challenges of the project-based Sec tion 8 program. April 23, 2008. http://www.nhtinc.org/Section_8/NHT_tesim ony_pbSec8_042308.pdf (accessed September 23, 2008). 2002. The increasing shortage of affordable re ntal housing in America: Action items for preservation. Housing Facts and Findings 4 (4): 1-2.

PAGE 205

205 Bodaken, Michael and Anne Heitlinge r. 2002. Providing affordable housing. Planning Commissioners Journal 45 (Winter): 1-3. http://www.nhtinc.org/documents/pcjarticle-affordable-housing.pdf (accessed November 3, 2006). Bogdon, Amy S. and Jam es R. Follain. 1996. An exam ination of three sets of indicators of financial risk among multifamily properties. Metropolitan Studies Program Series Occasional Paper No. 180. Bradley, Donald S., Amy Crews Cutts, and James R. Follain. 2001. An examination of mortgage debt characteristics and financia l risk among multifamily properties. Journal of Housing Economics 10: 482-506. Bratt, Rachel G. 1989. Rebuilding a Low-income Housing Policy Philadelphia: Temple University Press. California Housing Partnership Corporation. 2006. About CHPC. http://www.chpc.net/ (accessed January 17, 2007). 2001. The tax credit turns fifteen. Conversi on risk in Californias early tax credit portfolio. http://www.chpc.net/ (accessed December 13, 2006 ). Center for Housing Policy. 2008. Taking stock: Th e role of preservat ion inventories in preserving affordable rental housing. http://www.nhc.org/ housing/insights (accessed January 4, 2009). Center for Urban and Environm ental Solutions. 2006 Charting the Cou rse: Where is South Florida Heading? Fort Lauderdale, FL: Florida Atlantic University. Chicago Rehab Network. 2003. Section 8 preservati on initiative. Opt-out risk assessment. In Affordable Housing Study Commissi on meeting packet, April 5-6, 2006. City of Los Angeles. 2002. General plan of the City of Lo s Angeles. Housing Element. http://cityplanning.lacity.org /cwd/gnlpln/HsgElt/index.htm (accessed March 2, 2007). Community Economic Development Assistance Corporation. 2008. Notes to CEDAC report on Massachusetts developments with subsidized mortgages or HUD project-based rental assistance. September 28, 2004. Revised February 19, 2008. http://www.chapa.org/ (accessed Jan uary 5, 2007). Eggers, Frederick J. and Fouad Moumen. 2008. Trends in housing costs. 1985-2005 and the 30percent-of-income standard. Prepared fo r U.S. Department of Housing and Urban Development, Office of Polic y Development and Research. http://www.huduser.org/publicati ons/affhsg/trends_hsg_cost.htm l (accessed December 18, 2008).

PAGE 206

206 Farley, Megan. 2002. Taking stock. Washington State project-base d Section 8 housing: A risk assessment. Prepared for the Washington Low Income Housing Network. Fitterman, Stan. 2007. Presentation at the Joint Statewide Conference of the Florida Coalition for the Homeless and the Florida Supportive H ousing Coalition, St. Pe tersburg, Florida, October 25, 2007. Finkel, Meryl, Donna DeMarco, Deborah Morse, Sandra Nolden, and Karen Rich. 1999. Status of the HUD-insured (or held) multifamily rent al housing in 1995. Prepared for the U.S. Department of Housing and Urban Devel opment, Office of Policy Development and Research. Finkel, Meryl, Charles Hanson, Richard Hi lton, Ken Lam, and Melissa Vandawalker. 2006. Multifamily properties: Opting in, opting out an d remaining affordable. Prepared for the U.S. Department of Housing and Urban Deve lopment, Office of Policy Development and Research. http://www.huduser.org/in tercept.asp?loc=/Publications/pdf/opting_in.pdf (accessed Septem ber 6, 2006). Finkel, Meryl and Ken Lam. 2008. HUD Multifamily Portfolio in Florida. Prepared for University of Florida. Florida Association of REALTORS. 1996-2008. Florida sales report. Single-family, existing homes. http://media.living.net/stati stics/statisticsfull.htm (accessed March 1, 2009). Follain, James R. 1994. Som e Possible Directio ns for Research on Multifamily Housing. Housing Policy Debate 5 (4): 533-68. French, Nick and Laura Gabrielli. 2005. Disc ounted cash flow: Accounting for uncertainty. Journal of Property Investment & Finance 23 (1): 76-89. Friedman, Lawrence M. 1985. Housing reform: Negative style. In Federal housing policy & programs: Past and present ed. J. Paul Mitchell, 207230. New Brunswick, NJ: Center for Urban Policy Research/Rutgers University. Galster, George, Peter Tatian, and Charlene Wilson. 1999. Alternative measures for the financial condition of the multifamily housing stock. Housing Policy Debate 10 (1): 5973. Goodman, Jack. 2004. Determinants of operating costs of multifamily rental housing. Journal of Housing Economics 13: 226-44. Governors Task Force for Housing Preser vation. 2004. Preserving Wisconsins quality affordable rental housing. http://www.wheda.com/ (accessed May 12, 2006). Green, Rich ard K. and Stephen Malpezzi. 2003. A Primer on U.S. Housing Markets and Housing Policy. Washington, DC: The Urban Institute.

PAGE 207

207 Harvard University Graduate School of Desi gn. 2003. Public housing operating cost study. Final report. Prepared for the U.S. Departme nt of Housing and Urban Development. http://www.gsd.harvard.e du/research/research_ce nters/phocs/docum ents.html (accessed September 1, 2008). Hays, R. Allen. 1995. The Federal Government and Urban H ousing: Ideology and Change in Public Policy Albany, NY: State University of New York Press. Hecht, Bennett L. 2006. Developing Affordable Housing. A Practical Guide for Nonprofit Organizations Hoboken, NJ: John Wiley & Sons, Inc. Hoesli, Martin, Elion Jani, and Andr Bender. 2006. Monte Carlo simulations for real estate valuation. Journal of Property Investment & Finance 24 (2): 102-22. Housing Development Center. 2006. Risk assessm ent and strategy recomme ndations to preserve Oregons Low Income Housing Tax Credit (LIHTC) projects reaching year 15 from 2006-2011. http://www.hdc1.org/Yr15Main.html (accessed Nove mber 12, 2006). Institute of Real Estate Management. 2007. Income/expenses Analysis: Federally Assisted Apartments. Chicago: Institute of Real Estate Management of the National Association of REALTORS. Joint Center for Housing Studies of Harvard Univ ersity. 2008. The state of the nations housing: 2008. http://www.jchs.harvard.edu/son/index.htm (accessed August 21, 2008). 2007. The state of the nations housing: 2007. http://www.jchs.harvard.edu/publ ications/m arkets/son2007/index.htm (accessed September 5, 2007). 2006. The state of the nations housing: 2006. http://www.jchs.harvard.edu/publ ications/m arkets/son2006/index.htm (accessed August 13, 2006). Katz, Bruce and Margery Austin Turner. 2007. Re thinking U.S. rental housing policy. Prepared for Revisiting Rental Housi ng: A National Policy Summit. Joint Center for Housing Studies of Harvard University. http://www.jchs.harvard.edu/publications (access ed May 12, 2008). Kelliher, Charles F. and Lois S. Mahoney. 2000. Using Monte Carlo simulation to improve longterm investment decisions. The Appraisal Journal January: 44-56. Khadduri, Jill and Charles Wilkins. 2007. Designing subsidized rental h ousing programs: What have we learned? Prepared for Revisiting Rental Housing: A Na tional Policy Summit. Joint Center for Housing Studi es of Harvard University. http://www.jchs.harvard.edu/publications (accessed May 12, 2008).

PAGE 208

208 Knight, Frank H. 1965. Risk, Uncertainty and Profit New York: Hart, Schaffner & Marx. Kochera, Andrew, Donald L. Redfoot, and Jeremy Citro. 2001. Section 8 project-based rental assistance: The potential loss of affordable federally subsidized housing stock. Research Report, AARP Public Policy Institute, IB Number 47. http://www.aarp.org/research /housing-m obility/affordab ility/aresearch-import-785IB47.html (accessed October 30, 2005). Koebel, Theodore C. and Cara L. Bailey. 1992. State policies and pr ograms to preserve federally assisted low-income housing. Housing Policy Debate 2 (4): 995-1016. Li, Ling Hin. 2000. Simple computer applications improve the versatility of discounted cash flow analysis. The Appraisal Journal January: 86-92. Listokin, David. 1991. Federal housing policy and pres ervation: Historical evolution, patterns, and implications. Housing Policy Debate 2 (2): 157-85. Local Initiatives Support Corporation. 2005. Po licy update. Slipping away? Assessing the risk facing affordable housing. Florida Update Fall 2005. McClure, Kirk and Michael T. Grube. 2007. Low -Income Housing Tax Credit developments come of age: What should planners do when a development is up for sale? Presented at the Annual Conference of the Association of Collegiate Scho ol of Planning, Milwaukee, Wisconsin, October 20, 2007. Melendez, Edwin, Alex Schwartz, and Al exandra de Montrichard. 2007. Year 15 and preservation of tax-credit housing for low-income households: An assessment of risk. Presented at the Annual Conf erence of the Association of Collegiate School of Planning, Milwaukee, Wisconsin, October 20, 2007. Millennial Housing Commissi on. 2002. Report of the Bipartisan Millennial Housing Commission. http://govinfo.library.unt.edu/mhc/MHCReport.pdf (accessed May 27, 2006). Mitchell, J. Paul. 1985. Federal Housing Policies and Programs, Past and Present. New Brunswick, NJ: Center for Urban Policy Research. Myers, Dowell and Cathy Yang Liu. 2005. The emerging dominance of immigrants in the US housing market 1970-2000. Urban Policy and Research 23 (3): 347-65. Myerson, Roger B. 2005. Probability models for economic decisions. http://home.uchicago.edu/~rmyerson/prbook.htm (August 27, 2008). National Housing Law Project. 1999. Preservation crisis m ounts: HUD and Congress respond. http://www.nhlp.org/html/hlb/499/499preser.htm (April 7, 2006).

PAGE 209

209 National Housing Preservation Task Force. 1988. The preservation of low and moderate income housing in the United States of America. Washington, DC: The National Advisory Council of HUD Management Agents. National Housing Task Force. 1989. A decent place to live: The report of the National Housing Task Force. Washington, DC. National Housing Trust. 2006a. The Mark-toMarket program. Testimony of NHT Vice President Scott Kline in support of extension. http://www.nhtinc.org/pub_pol_section8.asp (accessed January 9, 2007). 2006b. Special report: Expiring Sec tion 8 contracts in Florida. http://www.nhtinc.org/data_states.asp#fl (accessed January 1 2, 2007). 2004. Changes to project-based multifamily units in HUDs inventory between 1995 and 2003. Number of affordable pr oject-based units declines by 300,000. http://www.nhtinc.org/documents/PB_Inventory.pdf (accessed Nove mber 1, 2005). National Low Income Housing Coalition. 2009. 2009 Advocates guide to housing and community development policy. http://www.nlihc.org/doc/AdvocacyGuide2009-web.pdf (accessed April 2, 2009 ). National Low Income Housi ng Preservation Commission. 1988. Preventing the Disappearance of Low-income Housing. To the H ouse Subcommittee on Housing and Community Development and the Senate Subcommittee on Housing and Urban Affairs United States Congress. Washington, DC: U.S. Government Printing Office. Nissen, Bruce and Yue Zhang. 2006. State of wo rking Florida 2006. Res earch Institute for Social and Economic Policy at the Center for Labor Research and Studies, Florida International University. http://www.risep-fiu.org/ (accessed February 14, 2007). On-site Insight, Inc. 200 1. An analysis of the replacement reserve needs of certain properties in the affordable housing stock. 2001 update. http://webarchive.org/web/20060711033544/http://on-site-insight.com/hudemo.htm (accessed January 19, 20 08). Orlebeke, Charles J. 2000. The evolution of low-income housing policy, 1949 to 1999. Housing Policy Debate 11 (2): 489-520. Palisade Corporation. 2008. Guide to using @RISK. Risk analysis and simulation add-in for Microsoft Excel. Version 5.0 May, 2008.

PAGE 210

210 Pedone, Carla I. 1991. Estimating mortgage prepayment s and defaults in older federally assisted rental housing and the possible costs of preventing them. Housing Policy Debate 2 (2): 245-88. Peiser, Richard B. 1991. The fallout from fede ral low-income housing preservation programs: A case study in estimating damages. Housing Policy Debate 10 (2): 371-93. Pyhrr, Stephen A. 1973. A computer simulation m odel to measure the risk in real estate investment. American Real Estate & Urban Economics Association Journal 1 (1): 48-78. Quercia, Roberto G and George C. Galster. 1997. The challenges facing public housing authorities in a brave new world. Housing Policy Debate 8 (3): 535-69. Rawls, Linda. 2006. 90% of worker s cant afford typical home. Palm Beach Post, July 13. http://www.palmbeachpost.com (accessed Febru ary 11, 2007). Recapitalization Advisors, Inc. 2009. Glossary. http://www.recapadvisors.com/glossary.php (accessed January 10, 20 09). 2008. Workout group services: Evaluation. http://www.recapadvisors.com/wg/services/evaluation/ (accessed October 1, 2008). 2002. Existing affordable multifam ily housing in Cook County. Presented February 6, 2002. http://www.recapadvisors.com/pdf/cookcounty.pdf (accessed April 3, 2006). Rice, Douglas. 2009. What to look for in HUDs 2010 budget for low-inco me housing. http://www.cbpp.org/cms/index.cfm?fa=view&id=2803 (accessed May 17, 2009). Rice, Douglas and Barb ara Sard. 2009. Decade of neglect has weakened federal low-income housing programs. New resources re quired to meet growing needs. http://www.cbpp.org/cms/index.cfm?fa=view&id=2691 (accessed March 2, 2009). 2007. The effects of the federal budget s queeze on lo w-income housing assistance. http://www.cbpp.org/cms/index.cfm?fa=view&id=1039 (accessed March 20, 2008). Schnare, Ann B. 1991. The preservation needs of public housing. Housing Policy Debate 2 (2): 289-318. Schum ann, C.P. Salty. 2006. Improving uncertainty in valuation using the discounted cash flow method. Valuation Strategies 10 (1): 4-13. Schwartz, Alex F. 2006. Housing Policy in the United States: An Introduction. New York, NY: Taylor & Francis Group, LLC.

PAGE 211

211 Shimberg Center for Housing Studies. 2008a. P opulation projections. Florida Housing Data Clearinghouse. http://flhousingdata.shi m berg.ufl.edu/a/population (accessed December 3, 2008). 2008b. Assisted Housing Inventory. Fl orida Housing Data Clearinghouse. http://flhousingdata.shimber g.ufl.edu/AHI_introduction.htm l (accessed July 15, 2008). 2007. 2007 rental market study. Prepared for Florida Housing Finance Corporation. http://flhousingdata.shimberg.ufl. edu/shim berg_pub_presentation.html (accessed November 13, 2007). Slade, Barrett A. 2006. Property risk assessment: A simulation approach. The Appraiser Journal 74 (4): 347-57. Smith, Stanley K. 2005. Florida population growth: Past, present and future. Gainesville, FL: Bureau of Economic and Business Research. http://www.bebr.ufl.edu/ (accessed F ebruary 14, 2007). Southern California Association of Governments. 2000. The affordable housing preservation crisis: A risk assessment of the potential loss of HUD-assisted multifamily housing in the Southern California region. http://api.ucla.edu/workshop/HousingPreservation/LinksRe sources/HousingPreservation ReportFina l.htm (accessed May 1, 2006). The John D. and Catherine T. MacArthur Found ation. 2007. Policies that work: New ways to preserve and improve affordable rental hom es. Draft discussion paper prepared for the Window of Opportunity Nationa l Policy Forum, November 7-8. U.S. Census Bureau. 2008. Housing vacancies and homeownership. Annual statistics 2007. http://www.census.gov/hhes/www/hous ing/hvs/annual07/ann07t13.htm l (accessed January 6, 2009). 2007a. M2515. Percent of renter-occupied units spending 30 percent or more of households income on rent and utilities: 2007. Universe: Renter-occupied housing units. 2005-2007 American Community Survey. http://factfinder.census.gov (accessed J anuary 11, 2008). 2007b. B25072. Age of householder by gross rent as a percentage of household income in the past 12 months Universe: Re nter-occupied housing units. 2005-2007 American Community Survey. http://factfinder.census.gov (accessed Jan uary 11, 2008). 2007c. B25070. Gross rent as a percentage of household income in the past 12 months Universe: Renter-occupied housing units 2005-2007 American Community Survey. http://factfinder.census.gov (accessed January 11, 2008).

PAGE 212

212 2007d. B25074. Household income by gross rent as a percentage of household income in the past 12 months Universe: Re nter-occupied housing units. 2005-2007 American Community Survey. http://factfinder.census.gov (accessed Jan uary 11, 2008). 2005. Florida, California and Texas to dominate future population growth, Census Bureau reports. Press Release, CB05-52. Washington, DC. http://www.census.gov/Press-Release/www/ releases/archives/population/004704.htm l (accessed October 19, 2006). 1990. 1990 Census of population and housing. Table 16. Population: 1790 to 1990. Washington, DC: U.S. Bureau of the Census. http://www.census.gov/prod/cen1990/cph2/cph-2-1-1.pdf (accessed Novem ber 18, 2006). U.S. Congress, Congressional Budget Office. 1987. The potential loss of a ssisted housing units as certain mortgage-interest subsidy progr ams mature. Staff Working Paper, March 1987. U.S. Department of Housing and Urban Development. 2009. Interim report to Congress on the root causes of the foreclosure crisis. http://www.huduser.org/publications/hsg spec/int_foreclosure_report.htm l (accessed April 30, 2009). 2008a. Multifamily assistance and Secti on 8 contracts database. September 2, 2008. http://www.hud.gov/offices/hsg/mfh/mfdata.cfm (accessed Septem ber 29, 2008). 2008b. Insured multifamily mortgages database. June 30, 2008. http://www.hud.gov/offices/hsg/mfh/mfdata.cfm (accessed Ju ly 31, 2008). 2008c. Terminated multifamily mortgages database. June 30, 2008. http://www.hud.gov/offices/hsg/mfh/mfdata.cfm (accessed Ju ly 31, 2008). 2008d. Annual report for Florida. Obtained by the Shimberg Center for Housing Studies from HUD in Jacksonville. 2008e. Physical Inspection Scor es by State. September 23, 2008. http://www.hud.gov/offices/hsg/mfh/mfdata.cfm (accessed October 18, 2 008). 2007. FY 2007 opt out report. Obtained by th e Shimberg Center for Housing Studies from HUD in Washington, DC. 2006. Picture of subsidized households 2000. http://www.huduser.org/picture2000/index.html (accessed Novem ber 30, 2006). 1999. Section 8 program background info rmation. Content updated March 18, 1999. http://www.hud.gov/offices/hsg/mfh/rfp/s8bkinfo.cfm (accessed October 29, 2005).

PAGE 213

213 1996-2008. U.S. housing market conditions. http://www.huduser.org/periodicals/ushmc.html (accessed March 16, 200 9). UBC Real Estate Division. 2001. Real Estate Investment Analysis and Advanced Income Appraisal Course Manual Vancouver: UBC Real Estate Division. United States General Accounting Office. 2004. Multifamily housing. More accessible HUD data could help efforts to preserve housing for low-income tenants. Urban Land Institute. 2006. Dollars & Cent s of Multifamily Housing: 2006. CD-ROM. Vandell, Kerry D. 1995. FHA restructuring pr oposals: Alternatives and implications. Housing Policy Debate 6 (2): 299-393. Van Horne, James C. and John M. Wachowicz, Jr. 2001. Fundamentals of Financial Management Upper Saddle River, NJ: Prentice Hall. Vermont Housing Finance Agency. 1988. At ri sk analysis, VHFA Section 8 portfolio. Memorandum to At Risk Task Force, Co re Staff and VHFA Board, from Sam Falzone, Senior Management Officer, September 15, 1988. Vidal, Avis C. 1997. Foreword. In Affordable Housing and Urban Redevelopment in the United States ed. Willem van Vliet, ix-xii. Thousa nd Oaks, CA: SAGE Publications, Inc. Von Hoffman, Alexander. 1998. The origins of American housing reform. Joint Center for Housing Studies, Harvard University. http://www.jchs.harvard.edu/publications /comm unitydevelopment/von_hoffman_W982.pdf (accessed May 19, 2006). Vose, David. 1996. Quantitative Risk Analysis: A Guide to Monte Carlo Simulation Modeling Chester, England: John Wiley & Sons Ltd. Wallace, James E. 1995. Financing affo rdable housing in the United States. Housing Policy Debate 6 (4): 785-814. 1992. Assessment of the HUD-insured multifamily housing stock. Modeling the future status of HUD-insured multifamily rental housing. Prepared for the Office of Policy Development and Research, U.S. Department of Housing and Urban Development. Wallace, James E., Meryl Finkel, Janine Sull ivan, and Karen Rich. 1993. Assessment of the HUD-insured multifamily housing stock. Final report. Volume 1. Prepared for U.S. Department of Housing and Urban Devel opment, Office of Policy Development and Research. Weaver, William and Stuart Michelson. 2004. A pedagogi cal tool to assist in teaching real estate investment risk analysis. Journal of Real Estate Practice and Education 7 (1): 4352.

PAGE 214

214 White, Douglas, Jim Martinez, Diep Nguyen, and William ODell. 2009. The State of Floridas Housing 2008. Prepared by the Florida Housing Data Clearinghouse, Shimberg Center for Affordable Housing, University of Florida. http://flhousingdata.shimberg.ufl. edu/shim berg_pub_presentation.html (accessed April 19, 2009). White, Douglas, Marc T. Smith, Dean Gatzlaff, Mary Lois White, Jim Martinez, Diep Nguyen, and William ODell. 2005. The State of Flor idas Housing 2004. Prepared by the Florida Housing Data Clearinghouse, Shimberg Center for Affordable Housing, University of Florida. http://flhousingdata.shimberg.ufl.edu/shimberg_pub_presentation.html (acces sed December 11, 2006). Wilkins, Charles S., Jr. 2002. Background paper: Barriers to acquisition of at-risk properties by preserving entities. Prepared for the Millennial Housing Commission. http://govinfo.library.unt .edu/m hc/papers/bgbta.doc (accessed March 12, 2007). Zilpy LLC. 2009. Rent snapshots by zip code and unity type. http://www.zilpy.com/ (access ed February 15, 2009). Zwick, Paul D. and Margaret H. Carr. 2006. Florida 2060. A populati on distribution scenario for the State of Florida. A research project prepared for 1000 Friends of Florida by the GeoPlan Center at the Un iversity of Florida. http://www.1000fof.org/PUBS/2060/ Florida-2060-Report-F inal.pdf (accessed January 9, 2007).

PAGE 215

BIOGRAPHICAL SKETCH Patricia E. R oset-Zuppa was born and rais ed in the Netherlands. She studied human geography at Utrecht University and specialized in urban planning. As part of her graduate studies in Utrecht, she participated in an intern ational exchange program with the University of Florida for one semester in 1996. She completed her thesis research in Toronto in 1997. Upon graduation, Ms. Roset-Zuppa worked as a researcher at the Canadian Urban Institute in Toronto for several years. In 2000-2002, she pursued a Master of Bu siness Administration with a specialization in real property development at York University in Toronto. During this time, she worked as a real estate credit analyst at Scotiabank for one summer. Upon completion of the MBA degree, Ms. Roset-Zuppa worked at Diam ante Development, a Toronto-based builder, where she was charged with the planning approv als process. Her subseq uent position was as project manager in the international office of th e Canadian Urban Institut e in Toronto, managing capacity building projects for local governments in Cuba and the Philippines. Ms. Roset-Zuppa also worked for Monarch Corporation, a land de veloper and homebuilder in Southern Ontario, where she was responsible for feasibility anal ysis and the due diligence process of land acquisition. While pursuing a Ph.D. in urban and regional planning at the University of Florida (2005-2009), she was a research and policy analyst at the Shimberg Center for Housing Studies. Upon completion of her doctoral studies, Ms. Ro set-Zuppa joined the Canada Mortgage and Housing Corporation as a senior policy analyst at its nationa l office in Ottawa.