|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 THE VALUATION IMPLICA T I ONS OF UNREALIZED GAINS AND LOSSES ON NON AGENCY SECURITIES By SUKYOON JUNG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FO R THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 2
2 201 2 Sukyoon Jung
3 I dedicate this to my parents, my wife, and my daughter.
4 ACKNOWLEDGMENTS Above all, I am so grateful to God for everything. My thanks then go to my advis or Professor Stephen Asare, for his constant encouragement and mental support during ve helped me face academic as well as nonacademic challenges. I also th ank my other commi ttee members Professor Marcus Kirk, Professor Mark Flann ery, and Professor Gary McGill for their time and for their thoughtful comments on my dissertation. I thank all the other professors in our department for their help throughout my graduate study. Nex t, I thank my friends Mike Donohoe, Stephen Brown, Ping Wang, and Song Xue for the pleasant time I spent with them. Last, but of course not least, I thank my family for their unconditional support. I acknowledge the constant emotional support and encourage ment I received from my mother, father, brother, and wife. Without them I could not have completed my doctoral work.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LI ST OF TABLES ................................ ................................ ................................ ............ 7 ABSTRACT ................................ ................................ ................................ ..................... 8 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ...... 9 Motivation and Researc h Question ................................ ................................ ........... 9 Preview of Results ................................ ................................ ................................ .. 10 Contributions ................................ ................................ ................................ ........... 11 Organization ................................ ................................ ................................ ........... 1 2 2 BACKGROUND AND RELAT ED LITERATURE ................................ ..................... 13 Fair Value Accounting ................................ ................................ ............................. 13 Existing Accou nting Research on Fair Value ................................ .......................... 14 3 NON AGENCY SECURITIES ................................ ................................ ................. 17 4 HYPOTHESES DEVELOPME NT ................................ ................................ ........... 20 Valu e Relevance of Unrealized Gains and Losses on Non agency Securities ....... 20 Predict ive Ability of Unrealized Gains and Losses on Non agency Securities for Future Earnings ................................ ................................ ................................ ... 23 Valuation Implication s of Unrealized Gains and Losses on Non agency Securities during the Financial Crisis ................................ ................................ ... 27 5 S AMPLE AND DESCRIPTIV E STATIST ICS ................................ .......................... 32 6 EMPIRICAL RESULTS ................................ ................................ ........................... 39 Value Relevance of Unrealized Gains and Losses on Non agency Securities ....... 39 Predictive Ability of Unrealized Gains and Losses on Non agency Securities for Future Earnings ................................ ................................ ................................ ... 40 Valuation Implications of Unrealized Gains and Losses on Non agen cy Securities during the Crisis ................................ ................................ .................. 41 7 SENSITIVITY OF RESULTS ................................ ................................ .................. 50 8 CONCLUDING REMARKS ................................ ................................ ..................... 51 LIST OF REFERENCES ................................ ................................ ............................... 53
6 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 58
7 LIST OF TABLES Table page 5 1 Des criptive Statistics for the Full Sample ................................ ........................... 36 5 2 Spearman and Pearson Correlations ................................ ................................ 37 5 3 The Mean and Median Ratios of Amortized Costs to Fair Values of Non agency Securities from 2001 to 2009 ................................ ................................ 38 6 1 Value R elevance of Unrealized Gains and Losses on Non agency Securities ... 43 6 2 Test of Value R elevance Equality ................................ ................................ ....... 44 6 3 Predictive Ability of Unrealized Gains and Losses on Non agency Securities for Future Interest Income ................................ ................................ .................. 45 6 4 Predictive Ability of Unrealized Gains and Losses on Non agency Securities for Future Realized Gains and Losses ................................ ............................... 46 6 5 Value Relevance of Unrealized Gains and Losses on N on agency Securities during the Crisis ................................ ................................ ................................ .. 47 6 6 Predictive Ability of Unrealized Gains and Losses on Non agency Securities for Future Interest Income during the Crisis ................................ ....................... 48 6 7 Predictive Ability o f Unrealized Gains and Losses on Non agency Securities for Future Realized Gaines and Losses during the Crisis ................................ ... 49
8 Abstract of Dissertation Presented to the Grad uate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE VALUATION IMPLICA T I ONS OF UNREALIZED GAINS AND LOSSES ON NON AGENCY SECURITIES By Sukyoon Jung May 2012 Chair: Stephen K. Asare Major: Business Administration In this paper I examine the value relevance and future earnings predictive ability of unrealized gains and losses on non agency securities and how these valuation properties have changed in the recent crisis period. I find value relevance of these unrealized changes, but only in the crisis period, which is consistent with the fire sale expectation story. A stronger relation between the unrealized changes and future realized gains/losses in the crisis period p rovides economic rationale to the value relevance during the crisis period. I also find that fair value revaluations for non agency securities are positively associated with interest income in future periods. In the crisis period, however, this association is weaker. Such results suggest the potential market instability, although fair value information satisfies the value relevance criterion.
9 CHAPTER 1 INTRODUCTION Motiv ation and Research Question Non agency securities, which include non agency mortgage backed securities (MBS) and other asset backed securities (ABS), are generally difficult to value due to limited market participants, infrequent transactions, or complex s tructures (Ashcraft and Sch uermann 2008; Gorton 2008). The difficulty of valuing non agency securities has been amplified during the financial crisis, rendering fair value estimates for those securities less reliable. Emphasizing the reduced reliability of fair value estimates for non agency securities, the banking industry has questioned the usefulness of fair value estimates for these se curities during the crisis (SEC 2008; Leone 2008). 1 In this paper I examine the usefulness of fair value information for non agency securities from two perspectives : (1) the value relevance and (2) the future earnings predictive ability of unrealized changes on non agency securities. 2 Value relevance tests provide evidence of fair value esti mates for these complex financial instruments. Future earnings predictive ability tests provide direct evidence of the relation between fair value revaluations for non agency securities and future earnings realizations from those securities. Such evidence potentially helps standard setters evaluate the usefulness of fair value information for complex financial 1 They also expressed concern that fair value measurement yielded an adverse feedback effect which had caused further reductions in the market values of illiquid assets and possibly even systematic risk. See SEC (2008), Shaffer (2010) and Badertscher et al. (2010) for a discussion of this concern. 2 Unrealized changes, unrealized gains and losses and fair value revaluation are used interchangeably throughout this article.
10 instruments based on one of qualitative characteristics of accounting information as stated in the Statement of Financial Accounting Concepts No. 2. 3 I next examine whether these valuation implications of unrealized changes on non agency securities may have changed during the recent financial crisis. Although the effect of market instability on valuation implications of fair values is not documented i n the literature, market instability may render fair values difficult to measure and thus unreliable, as appears to have occurred for non agency securities during the credit crunch. Thus, the usefulness of fair value estimates for complex financial instrum ents to financial statement users may hav e decreased (Barth and Landsman 2010). Preview of Results Using a sample of bank holding company (BHC) data between 2001 and 2009, I find that unrealized gains and losses on non agency securities are not value rele vant. The result implies that investors do not consider these revaluations informative because of the (assumed) lack of reliability of fair value estimates for these securities. The unrealized gains/losses on non agency securities, however, gain value rele vance in the crisis period although fair values for these securities are allegedly not reliable due to market illiquidity. A possible explanation for this is that investors may increase their assessments of fair value for such illiquid securities because o liquidation of those securities during this period. A stronger relation between the unrealized changes and future realized gains/losses in the crisis period provides an economic rationale for value relevance during the crisis period. 3 Statement of Financial Accounting Concepts No. 2 states that predictive value is an important consideration in distinguishing relevant from irrelevant accounting information (FASB 1980).
11 I also find that fair value revaluations for non agency securities are positively associated with one and two year ahead future interest income. This suggests that unrealized gains/losses o n non agency securities reflect changes in values associated with p repayment and credit risk in spite of the less reliable fair values. The relation between the fair value revaluations and future interest income in the crisis period, abili ty for future earnings during times of market instability, although fair value information satisfies the value relevance criterion. Contributions M y study makes three contributions to the existing literature. First, I examine the valuation implications of non agency mortgage backed and other a sset backed securities which have not been previously examined in the literature. This analysis is important to standard setters as they consider expanding fair value accounting to loans and other financial instrument s. Particularly for bank loans that share similar characteristics with mortgaged related securities, my results suggest that the fair value mandate may introduce measurement error that ultimately compromises the usefulness of fair values in evaluating bank Second, my predictive approach allows me to evaluate claims made by the banking industry directly, thus offering an alternative perspective for standard setters when evaluating the relevance of fair va lue. I provide evidence that fair value revaluations for non agency securities are useful in predicting future interest income although investors do not generally consider them informative. However, the usefulness of these revaluations is reduced during ti mes of market instability such as the financial crisis.
12 Third, I investigate the valuation implications using a longer period that includes observations from before and during the crisis period. This is in contrast to recent studies, which examine the va lue relevance of Level 3 assets (many of which are presumably non agency securities) during only the first three quarters of 2008 (Song et al. 2011; Kolev 2010). More importantly, rather than rely solely on fair values I also consider amortized cost. Focus ing on the valuation implications of value differences (fair value less amortized cost) allows me to explore the incremental usefulness of fair values to amortized costs, particularly in predictive tests. My analysis thus complements and extends recent stu dies in the growing literature on fair values. Organization The remainder of the paper is organized as follows. Chapter 2 provides a brief background and related research concerning fair value measurement. Chapter 3 describes non agency securities owned b y BHCs, and Chapter 4 develops hypotheses and outlines the research design for the value relevance and predictive ability tests. Sample selection, empirical results, and sensitivity of results are described in Chapter 5, 6, and 7, respectively Finally, Ch apter 8 offers concluding remarks.
13 CHAPTER 2 BACKGROUND AND RELAT ED LITERATURE Fair Value Accounting FAS 157 (also known as Accounting Standards Codification 820 in the updated FASB Codification) defines fair value as the price that would be received to sell an asset or paid to transfer a liability in an orderly transaction between market p articipants at the measurement date (FASB 2006). 1 that an orderly transaction is neither a forced liquidation n or a distressed sale. The standard setters (FASB and IASB) consider fair value measurement as a possible measurement basis in many situations (Barth 2006). The fundamental case in favor of fair value accounting is that fair value incorporates current infor mation about future cash flows and current risk adjusted discount rates into the financial statements. Incorporating timely information enhances the relevance of information in financial with potential advantages to investors, managers, and other parties in making decisions (FASB 2000 ; Ryan 2007 ) When quoted prices in active markets are available, there are few conceptual objections against fair value accounting. However, when financial i nstruments are not actively traded, firms would have to either estimate their fair value or use the quoted 1 FAS 157 also outlines a hierarchy of inputs to derive the fair value of an asset or liability. Level 1 inputs are quoted prices in active markets for identical assets. If Level 1 inputs are not available, models are used to determine fair value, which is sometimes call models use observable inputs (Level 2), which includes quoted prices for similar assets and other relevant market data (such as interest rate yield curves or spreads between related interest rates). Level 3 inputs are unobservable inputs, typically model assumptions, and can be used if observable inputs are not available.
14 price from the illiquid market. In either case fair value estimates may contain significant bias and error, thereby lacking reliability. Particularly in the context of the financial crisis, the banking industry complains that fair values of mortgage related assets are poor indicators of their long run value as the market price of them has tumbled (Hughes and Tett 2008). Thus, fair values are of limited usefulness to bank investors to adequately Existing Accounting Research on Fair Value Empirical accounting research has taken a value relevance approach to evaluate the usefulness of fair value i nformation to investors. Such research generally analyzes the association of fair value estimates with stock prices and returns. A significant and predicted association of fair value estimates with stock price or returns implies that the estimates are rele vant and sufficiently reliable to be impounded in firm value (Barth et al. 2001). Much of the accounting research assessing the value relevance of fair value information focuses on banks, since banks are largely comprised of financial assets and liabilitie s. In general, prior studies provide substantial evidence that fair value estimates for financial instruments are relevant to investors and reliable enough to be reflected in share price or returns. This conclusion h old s for pensions and other post r etirement liabilities (Landsman 1987; Barth 1991), bank loans and core deposits (Barth et al. 1996; Eccer et al. 1996; Nelson 1996), derivatives (Schrand 1997; Venketachalam 1996), and investment securities held by closed e nd mutual funds (Carroll et al. 2 003). values are incrementally associated with bank share prices after controlling for e
15 value after controlling for future profitability. The results from a returns specification, which may implicitly control for correlated omitted variables, are also mi xed. Barth (1994) finds that unrealized gains and losses on investment securities do not possess explanatory power in explaining contemporaneous stock returns. The ambiguous finding for securities gains and losses is typically attributed to 1) measurement errors in the estimated unrealized gains and losses and 2) the omission of correlated unrealized gains and losses on other assets and liabilities. Barth et al. (1995) lend support to the measurement error explanation by showing that fair value based measur es of net income are more volatile than historical cost based measures, but the incremental volatility is not reflected in bank share prices. Ahmed and Takeda (1995) provide support for the second explanation by showing that, after controlling for interest rate sensitivity of other assets and liabilities, unrealized gains and losses on investment securities become positively related to stock returns. Another factor that might contribute to the mixed findings above is that prior research on the value relevan ce of fair value estimates for investment securities often used an indirect classification to distinguish securities with high versus low measurement error (i.e., Barth 1994; Khurana and Kim 2003). For example, Barth (1994) splits her sample by the proport ion of U.S. Treasury securities to explore the measurement error explanation because U.S. Treasury securities are presumably actively traded. Such indirect identification likely leads to imperfect partitioning because banks hold a wide range of investment securities, among which U.S. Treasuries account for, on average, less than 5% o f investment securities (Penman 2007).
16 Recently, Song et al. (2010) and Kolev (2010) document the value relevance of the FAS 157 fair value estimates for samples of firms in th e banking industry during the first three quarters of 2008. While the estimated value relevance parameters differ across studies (due to different samples and specifications), these studies find that investors price a reported $1 of Level 3 assets signific antly below a reported $1 of Level 1 assets. They suggest that investors apply larger discount factors to the reported Level 3 fair values because they are subject to more model risk (or noise) and larger information asymmetry. While much literature discus ses the value relevance of fair value information (some of which is summarized above), research on the relation between fair values and future performance measures is limited. The few such studies include Aboody et al. (1999), Park et al. (1999), and Evans et al. (2011). Aboody et al. (1999) provide evidence that U.K asset revaluations are associated with future performance as well as share prices, indicating that the asset revaluations are not unreliably measured. Park et al. (1999) document that the unre alized changes from the available for sale securities explain one year ahead bank earnings, while those from the held to maturity securities do not. Recently, Evans (2011) et al. show that fair values from the investment securities have predictive ability for future realized income and the value relevance of fair values varies with the ability of fair values to predict reported income.
17 CHAPTER 3 NON AGENCY SECURITIES Bank holding companies are required to classify their investment securities into trading, held to maturit y (HTM) and available for sale (AFS) portfolios. Trading portfolios are carried at fair value, with realized and unrealized gains and losses reported in the income statement as part of trading revenue. Under FASB ASC 320 (formerly FAS 115) securities in the HTM portfolio are ac counted for at amortized cost, with fair values disclosed but not recognized. Securities that do not qualify for the HTM portfolio (i.e., no intent and ability to hold the securities until they mature) are to be classified as AFS. For A F S se cu ri ti es FASB ASC 320 requires formal balance sheet recognition at fair value, with unrealized gains and losses recognized in the owners' equity section 1 My study focuses on non agency securities classified as AFS and HTM investment securities. Bec ause of the restrictive rules on when a n asset could be considered HTM most BHCs in my sample carry non agency MBS and ABS as AFS. 2 I exclude non agency securities in trading portfolios for two reasons. First, BHCs do not provide amortized costs for trad ing portfolios, preventing me from computing unrealized gains/losses on trading securities. Second, fair value accounting for trading securities is supported even by the banking industry. The banking industry agrees that fair value accounting is appropriat e for assets that are held for trading purposes and provides useful information for financial statements users (I nternational Banking Federation 1 If the fair value option is applied to AFS and HTM securities for existing securities, then those securities would be classified as tra ding as prescribed by FASB ASC 825 10 (formerly FAS 115) 2 AFS (HTM) category contains approximately 90% (10%) of bank holdings of non agency securities in my sample.
18 2008). Thus, the usefulness of fair value application to trading securities is hardly a controversial issue. S chedule HC B in the regulatory call report database (FR Y 9C forms) provides fair values and amortized costs for various types of investment securities held by bank holding companies. Among various types of investment securities, I focus on non agency secu rities which include bonds typically issued by homebuilders or financial institutions through subsidiaries and are backed by pools of loans (e.g., mortgage, credit cards, auto loans). Unlike agency securities (i.e., agency MBS), there is no government gua rantee for these securities. Thus, credit risk, as well as prepayment and interest rate risk, resides in the non agency security market. Many non agency asset backed securities are not liquid, and their price s are not transparent This is partly because no n agency securities are not as standardized as agency mortgage backed securities, and investors have to evaluate the different structures, maturity profiles, credit enhancements, and other features of an asset backed security before trading it (Sabarwal 20 05) During the financial crisis secondary markets for trading non agency MBS and ABS have been extremely illiquid due to an increase in information asymmet r y about the quality of the underlying assets (Ashcraft and Schuermann 2008). Although the inform ation on the trading activity of these securities is not publicly available, there is some evidence of the dissolution of the non agency security market. For 2008, non agency residential MBS issuance, which includes jumbo mortgages that exceed government s risk mortgages that do not meet agency underwriting guidelines (i.e., sub prime mortgages),
19 fell to an all time low of $ 25.3 billion, a decline of 94.4% from the $451.2 billion issued in 2007 Non agency commercial MBS (CMBS) issuance totaled $6.4 billion in 2008, down dramatically from the $228.2 billion issued in 2007. No new CMBS were issued in the fourth quarter of 2008 and only $0.1 billion was issued in the third quarter compared to the $28.4 billion issued in the fourth quarter of 2007. Total ABS issuance for 2008 was $137.2 b illion, a decline of 73% from $509.7 billion raised in all of 2007. In fact, the fourth quarter of 2008 marked the first time that four of the major sectors (home e quity, credit card, student loan, and equipment leases) had no issuances. 3 As markets became inactive and transaction prices were no longer available for non agency securities, there was vast uncertainty over how these securities should be valued (Gorton 2008), contributing to the valuation challenge. The banking industry claimed that the unusual market condition recently experienced led fair values for mortgage or other asset backed securities to understate the intrinsic values and to be more indi cative of distressed sales (ABA 2008;Krumwiede et al. 2008; Ryan 2008). This kind of assertion has been also made by the Bank of England (2008) and the Bank of Financial Stability Forum (2008), among others. In contrast, some argue that banks tended to overvalue the illiquid securities in their books by classifying more of t hem as Level 3 assets (Laux and Leuz 2010). 3 All statistics in this paragraph are based on the 2009 March Report prepared by Securities Industry and Financial Markets Association
20 CHAPTER 4 HYPOTHESES DEVELOPME NT Valu e Relevance of Unrealized Gains and Losses on Non agency Securities I begin my analysis by examining how in vestors value unrealized gains/losses from non agency securities owned by BHCs. The fair value of investment securities incorporates timely information about future cash flows and ris k adjusted discount rates (FASB 2000). Accordingly, unrealized gains/loss es, which are fair values less amortized costs, contain current information about changes in value of investment securities. Current information about changes in value of assets helps investors assess the value of the firm to make informed investment choic es. Despite this conceptual appeal of fair value accounting, critics of fair value accounting argue that the reduced reliability of fair value estimates in the absence of liquid markets would reduce the usefulness of fair value information. This argument i s relevant particularly for non agency securities whose fair values are less observable in active markets. Thus, fair value estimates for these securities likely require more estimation than estimates for other investment securities. As the fair value est imates rely on managerial estimation, managers can opportunistically make adjustments or choose unobservable inputs in the valuation process (Ronen 2008). Such managerial judgment and discretion could compromise the reliability of fair value estimates for non agency securities. Due to the possible lack of reliability in measuring fair value estimates for these securities, market investors may not perceive unrealized changes on these debt securities as sufficiently reliable to be used in v aluation Wh ether unrealized changes in non agency securities are used by market investors (thus, value relevant) or are not
21 perceived as useful information is an empirical question. Formally, I test the following hypothesis on non agency securities: H1 a : Unrealized g ains/losses on non agency securities are value relevant. As discussed above, fair values for non agency securities are more likely to be estimated using valuation models instead of observable market prices from active markets. Accordingly, fair value measu rements for non agency securities are likely to be less precise than fair value measurements for other investment securities. Prior research suggests that the value relevance of fair value measurements varies with the reliability of information. For examp le, Petroni and Wahlen (1995) find that fair values for equities and U.S. Treasury securities are value relevant, but fair values of municipal and corporate bonds are not, suggesting that securities actively traded in the market are more reliably associate d with the market value of equity. Barth (1994) divides her sample by the proportion of U.S. Treasury securities held by banks. She finds that unrealized gains and losses are more strongly associated with stock prices for banks that hold a high proportion of U.S. Treasury securities. Recently, Kolev (2010) and Song et al. (2010) document that fair value estimates based on Level 1 input are more strongly associated with stock price than estimates based on Level 3 input. Based on finding s in prior research, I predict that market investors give a lower valuation to unrealized changes of non agency securities than to those of other investment securities. My prediction is formally stated as Hypothesis 1b : H 1b : The value relevance of unrealized gains/losses on non agency securities is less than the value relevance of unrealized gains/losses on other investment securities. To test H1, I estimate the equation (1):
22 where P is the share price of the bank measured at the end of February of Yea r t+1 1 BVE is the book value of equity minus unrealized gains/losses from AFS investment securities. NI is earnings before extraordinary items. URGL TOTAL are unrealized gains/losses on all investment securities (excluding non agency MBS and ABS) classifie d as HTM and AFS. URGL NA_ A BS are unrealized gains/losses on non agency MBS and ABS. 2 Following Kolev (2010), I also include the log transformed total assets for the bank (SIZE) and the percentage change in total assets (GROWTH). These factors have been sho wn in the literature to affect the relationship between the price and book value of equity (Eccher et al. 1996; Khuranna and Kim 2003; Liu and Ohlson 2000; Nelson 1996; Nissim 2007). Yearly intercepts are included to account for macroeconomic factors. All variables except for SIZE and GROWTH are on a per share basis to reduce the scale effects in the regression model similar to extant research (Barth and Clinch 2009; Kolev 2010; Song et al. 2010). i represents the bank subscript while t represents the year subscript. Eq. (1) is based on a valuation model developed by Ohlson (1991) and further extended in Ohlson (1995) and Feltham and Ohlson (1995), which has been extensively 1 The 4 th quarterly filing date for top tier (lower tier) BHCs is 45 (50) calendar days after December 31. 2 Specifically, non agency MBS include all mortgage pass through securities not guaranteed by the U.S. governmen t, collateralized mortgage obligations (CMOs), real estate mortgage investment conduits (REMICs), CMO and REMIC residuals, and stripped mortgage backed securities issued by non U.S Government issuers for which the collateral does not consist of GNMA (Ginni e Mae) pass throughs, FNMA (Fannie Mae) pass throughs, FHLMC (Freddie Mac) participation certificates. ABS include all asset back ed securities collateralized by credit card receivables, home equity loans, automobile loans, commercial and industrial loans, and o ther consumer loans.
23 employed in the literature. As a refinement, I expand this model to separately evalu ate the value relevance of unrealized gains/losses on non agency securities 3 In Eq. (1) I first test whether the valuation coefficient o f URGL NA_ A BS ( 2 ) is different from zero. A positive and significant 2 is consistent with the conjecture that equity investors find the unrealized gains/losses on non agency securities sufficiently reliable to be reflected in bank value. I then test whether the coeffic ient for URGL NA_ A BS is smaller than the one for URGL TOTAL ( 1 ). I predict that 1 2 Predict ive Ability of Unrealized Gains and Losses on Non agency Securities for Future Earnings Next, I test whether the unrealized gains/losses on non agency securities predict future earnings. The test is motivated by the current debate among bank managers, investors, and capital market regulators about the usefulness of fair values particularly when markets are illiquid. Investors generally support measurements at fai r value as providing the most transparent financial reporting of an investment (Ryan 2008) In contrast, bank managers and capital market regulators question the usefulness of fair values based on illiquid or distressed prices when those prices do not refl ect the to Market Accounting, Todd Bernstein of Wachovia Securities wrote: The price of many of these [securitized mortgage] pools is well below their valu e based on cash flows, meaning the market is pricing in more losses than have actually, or may ever, occur. Mark to market accounting rules 3 The unrealized gains and losses used in my study are not recognized under the narrow concept of income (i.e., net income). However, they could be considered as components of a broad definition of income (i.e., comprehensive inco securities.
24 force banks to take artificial hits to capital without reference to the actual performance of loans in these pools. 4 The market concerns. Prices future financing and inves ting decisions (Aboody et al. 1999). For example, investors could anticipate that banks will be inclined to sell relatively illiquid assets to replenish their capital and thus price such an expectation, even if a decline in fair values is deemed temporary. In this case the valuation parameters in the market based tests could reflect implications for firm value value revaluations to future realized income from inves tment securities provides direct evidence of the association between revaluations and future operating performance and complements findings from the price specification. Under U.S. GAAP, unrealized gains and losses are the differences between the amortized costs and fair values. The amortized cost of the debt securities is equal to the present value of the remaining contractual payments discounted using the historical at p resent value of the remaining contractual payments discounted at the current expected return on similar investments (Ryan 2007). Thus, differences between the fair and amortized costs of debt securities are due to changes in their expected returns. Changes in expected returns, in turn, result from changes in interest rates, prepa yment expectations, credit risk or the pricing of credit risk depending on type of debt securities ( Nissim and Penman 2007). If unrealized gains/losses on non agency securities refl ect 4 See comment letters for Study on Mark to at http://www.sec.gov/comments/4 573/4 573.shtml).
25 changes in the values of investment securities associated with prepayment and credit risk, they will be positively associated with changes in future interest income from those securities. This leads to my second hypothesis in alternative form: H 2 a : Fai r value revaluations of non agency securities are positively related to future interest income from those securities. Potential lack of reliability of fair values from uncertainties inherent in the estimation can reduce the predictive ability (Aboody et al ., 1999). The predictive ability of unrealized gains/losses is also reduced to the extent that revaluations reflect interest rate movement because they represent the aggregate effect of interest rate, prepayment, and/or credit risk. Whether unrealized gain s/losses of non agency securities reflect information associated with future income and whether these estimates are measured reliably enough to be useful in predicting future interest income remain empirical questions. To examine the predictive ability of unrealized gains/losses on non agency securities, I estimate the equation below: where I is interest income from mortgage backed securities and other investment securities (excluding U.S. Treasury securities and U.S. government agency obligations) in Year t+k where k = 1 or 2. 5 BV OTHER are the amortized costs of investment debt securities excluding U.S. Treasury securities, U.S. government agency obligations, and 5 This framework is similar to Evans et al. (2011) except that the current realization of interest income is absent. In empirical analysis I find a high degree of multicollinearity between BV OTHER and the current interest income as indicated by a high variance inflation factor when t he current interest income is included in Eq. (2). Accordingly, inclusion of the current interest income may affect the estimation of BV OTHER although it does not affect the estimation of URGL NA_ABS
26 non agency securities. 6 BV NA_ABS are the amortized costs of non ag ency securities. SIZE is the natural log of a defined. I include SIZE because the composition of the investment securities portfolio varies considerably across the size of BHCs (Penman 2007). All var iables except for SIZE are then deflated by the book value of equity. I control for the amortized costs (BVs) in the estimating equation. Thus, significant positive coefficients o f URGLs suggest that fair value revaluation contains useful information in pr edicting future interest income beyond that provided by the amortized costs. I limit my tests to two future years as more than two year ahead future income data are not available for the crisis period. Unrealized gains and losses also can be materialized as realized gains/losses in future periods. Specifically, unrealized losses could be recognized in the future as other than temporary impairment charges if these are due to credit related events (FASB 1993) Bank managers can also r ealize the gains /losses through sales of securities. Prior research documents that managers selectively sell appreciated or depreciated assets to minimize tax or manage earnings or capital management (Scholes et al. 1990; Moyer 1990; Warfield and Linsmeier 199 2). This leads to Hypothesis 2b in alternative form: H 2 b : Fair value revaluations of non agency securities are positively related to future realized gains/losses from those securities. 6 Note that interest income used in the analysis does no t include interest income from U.S. Treasury securities and U.S. government agency obligations This is why the subscript OTHER (instead of TOTAL) is used in Eq. (2).
27 To examine the relation between unrealized changes of non agency securi ties and future realized gains/losses, I estimate the following regression: where RGL are realized gains/losses from AFS and HTM investment securities in Year t+k where k = 1 or 2 BV TOTAL are amortized costs of investment securi ties excluding non agency securities and all other variables are as defined previously. All variables are then deflated by the book value of equity at Year t A positive and significant coefficient o f URGL NA_ABS ( 2 ) indicates that the unrealized changes o n non agency securities are realized in future periods through sales of the securities or impairment charges. Valuation I mplications of Unrealized Gains and Losses on Non agency Securities during the Financial Crisis While the above equations would allow me to assess the valuation consequences of fair value revaluations from non agency securities, I also want to determine whether the valuations implications of the fair value revaluations may have changed during the financial crisis. From the perspective o f market investors, the usefulness of fair values particularly for illiquid securities may have been reduced due to the market disruption. The bursting of the housing market bubble in the recent financial crisis resulted in the collapse in prices of loans and other financial instruments whose values were tied to housing prices. Particularly, the extraordinary complexity of the instruments tied to mortgage payment or other loan receivables provided a significant impediment to insight into the value and the r eliability of cash flow (Gorton 2008). As a result, many complex financial instruments were difficult to sell and value, which made the o bservable transaction
28 prices of these securities no longer available. Indeed, at the heart of the financial crisis, non agency securities were rarely Level 1 assets and many banks moved them to Level 3 assets (Laux and Leuz 2010). Fair values for such instruments are therefore likely to be more difficult to estimate, which could reduce their combined relevance and reliabil ity. Such issues as the lack of relevance and reliability have been the general tenor of criticism of fair value, 7 but have not been examined in prior literature because of the recentness of the financial crisis. This leads to H ypothesis 3 a: H 3 a : The v alue relevance of unrealized gains/losses on non agency securities decreased in the crisis period. On the other hand, fair value revaluations of non agency securities might become more useful for investors during the financial crisis. The market expects that b anks facing financial difficulty are forced to sell their non agency securities which are traded in extremely illiquid markets to replenish their capital during the crisis. In this case the economic value of non agency securities more likely will equal liq uidation value. Thus, fair value revaluations for non agency securities become highly relevant marks for investors despite the questionable reliability of these fair value estimates 8 Selling pressure on illiquid assets is well articulated in the recent l iterature, particularly in the context of the financial crisis (Bhat et al. 2011; Laux and Leuz 2010). When asset prices decline and liquidity is reduced, banks are forced to sell their 7 For example, to Market Accounting, the Ame rican Bankers Associatio lower quality financial statements that we believe cannot be considered acceptable from the perspective of reliable accounting." 8 While not directly related to fair value measurement issues, prior research finds that investors place more weight on equity book values, which are better estimates of liquidation value of a firm than net income, as the financial health of a firm decreases (Barth e t al. 1998).
29 investments or raise capital due to the interaction of regulatory capi tal requirements that are based on the value of their assets (SEC 2008). In selling their investments, banks could be inclined to sell relatively illiquid assets such as non agency securities at a price below the fundamental value to pre empt the anticipat ed sales of other market participants. I f investors price such an expectation, the valuation parameter on non agency securities would capture such fire sale discounts. This leads to H ypothesis 3 b: H 3 b : The v alue relevance of unrealized gains/losses on non agency securities increased in the crisis period. To test H 3 I estimate the following regression model: where Pre_Crisis is a time dummy equal to 1 for 2001 to 2006 and 2009 9 and all o ther variables are as defined previously. A positive (negative) 2 is consistent with H 3 a (H 3 b). As in H 3 b, investors may increase their assessments of fair value for non agency liquidity crisis. The effect of banks s could lead to an increase in the value relevance of fair value revaluations for non agency securities even if the fair value revaluations do not reflect changes in fundamentals. Therefore, market based tests do not adequ 9 Alternatively, I define Pre_Crisis equal to 1 for 2001 to 2007 and 2009 and confirm that the results are very similar.
30 assets significantly deviated from fundamental values due to the liquidity discounts and were more indicative of distressed sales. An example of substantial liquidity discounts relates to the 2008 price collapse of AAA rated tranches of mortgage backed securities. Some banks w rote down the AAA rated super senior tranches of mortgage linked collateralized debt obligations by as much as 30% (Tett 2008) due to a fall in market prices. The Bank of England (2008) noted that if this change in price had stemmed from deterioration in fundamentals, it would have implied a loss rate of 38% for securit ized subprime mortgages. This, in turn, translates as 76% of the households defaulting and only repaying 50% of the face value of the mortgages. The Bank of England further noted that this seemed unrealistic because none of the AAA rated tranches had yet d efaulted and that there should not be any future defaults at all, even with a continued decline in U S house prices. Such possible divergences of fair values of non agency securities from their intrinsic values due to liquidity pricing lead to my fourth hypothesis: H 4 a : The relation between fair value revaluations of non agency securities and future interest income from those securities is weaker in the crisis period. To test H 4 a, I estimate the equation below: where Pre_Crisis is a time dummy equal to 1 for 2001 to 2006 and 2009, and all other variables are as defined previously. If bank managers are correct that unrealized gains/losses are mainly driven by factors other than fundamentals due to the illiquid market during the fi nancial crisis, I should expect that 2 > 0 or 2 < ( 2 2 ).
31 While the ability of unrealized changes of non agency securities to predict future interest income may be reduced in the crisis period, they can be more closely related to near term realized gai ns/losses during this period. As in the fire sale expectation story in H 3 b, a liquidity crisis may lead banks to sell the non agency securities at a fire sale price or to write down these securities to the fair values. Such possible increase in sales and i mpairment charges in the crisis period can lead to more realizations of unrealized changes on non agency securities in the near future. This leads to H ypothesis 4 b: H 4 b : U nrealized changes of non agency securities in the crisis period are more closely rela ted to future realized gains/losses from those securities than those in the non crisis period. To test H 4 b, I estimate the equation below: where Pre_Crisis is a time dummy equal to 1 for 2001 to 2006 and 2009, and all other varia bles are as defined previously. H 4 b expect that 2 < 0 or 2 > ( 2 2 ).
32 CHAPTER 5 SAMPLE AND DESCRIPTI VE STATISTICS I identify U.S. domestic BHC s regulatory call reports (FR Y 9C) for the fiscal years 2000 to 2010 provided by the Federal Reserve Bank of Chicago. Because lag and forward variables a re required in the analysis, my actual sample period runs from 2001 to 2009. The sample period begins in 2001 primarily because some of the variables (i.e., disaggregate information of investment securities) used were added to the report in the first quart er of 2001. Observations are reduced by bank years with missing total asset, net income, equity capital, and investment securities related interest income. I further identify publicly traded BHCs using the employer identification number and the Center for Research of Security Prices (CRSP) Federal Reserve Bank (FRB) link provided by the FRB of New York. I then require banks to have valid price data in the CRSP database. This initial procedure yields 3,717 bank years. To eliminate possible recording errors, I delete bank year observations with absolute values of unrealized gains/losses greater than the amortized costs and with ratios of amortized costs to fair values less than 0.4 or greater than 1.4 (22 bank years). Finally, I exclude bank year observations with stock prices less than $1, return on equity greater than 1.00, and asset growth greater than .90. These procedures leave me a final sample of 3,640 firm years. Untabulated statistics reveal that my sample has an annual maximum of 472 banks (2002) and an annual minimum of 302 banks (2009). To ensure that estimating expressions are not sensitive to extreme observations, I remove any observations that the Belsley, Kuh, and Welsch (1980) diagnostics indicate
33 are influential observations (studentized resid ual greater than 2). 1 Thus, the actual sample size employed varies with empirical tests. Table 5 1 reports descriptive statistics for the variables used in the multivariate tests. The mean (median) bank year in my sample has a share price (Price) of $23.1 7 ($20.83) and a book value per share (BVE) of $14.86 ($12.96). Annual income b efore extraordinary items (NI) averages $2.04 per share. Average unrealized changes from investment securities excluding non agency MBS and ABS (URGL TOTAL ) are positive due to t he price appreciation of U.S. Treasuries, Agency MBS, and Municipal bonds over the sample period, which accounts for more than 80 % of debt investment securities owned by bank holding companies. However, unrealized changes on non agency securities (URGL NA_A BS ) are negative, reflecting the substantial price depreciation in the crisis period. 2 The mean (median) yield for one year ahead interest income from mortgage backed securities and other debt securities (I t+1 ), excluding U.S. Treasury securities and U.S. government agency ob ligations, is approximately 8.4% (6.8 % year ahead realized gains and losses from investment securities have a mean and median value of 0.001 an d 0.000, respectively. About 80% (20 % ) of bank years bel ong to the pre crisis (crisis) period sample. Table 5 2 presents Pearson and Spearman correlations between selected variables used in the multivariate models. As expected, BE, NI, and URGL TOTAL are positively correlated with Price. URGL NA_ABS are significa ntly Spearman correlated in the positive direction with Price, implying that unrealized changes of non agency securities 1 This elimination procedure is similar to Song et al. (2010). 2 The descriptive statistics (not reported here) for th e sample partitioned by pre crisis and crisis years reveal that URGL AG_MBS are positive during the pre crisis period.
34 are value relevant. However, the Pearson correlation between URGL NA_ABS and Price is not significant. URGL TOTAL is significantly correl ated with I t+1 in a positive direction. In contrast, URGL NA_ABS is insignificantly Spearman correlated with I t+1 in the negative direction, implying that unrealized changes of non agency securities are not useful in predicting future interest income from t hose securities. The Pearson correlations between URGL NA_ABS and I t+1 are significant but negative. One year ahead realized gains/losses are significantly positively correlated with URGL TOTAL and URGL NA_ABS consistent with the predicted relation. Finally, the log transformed total asset (SIZE) and percentage change in total assets (GROWTH) are positively correlated with Price. Table 5 3 provides the mean and median ratios of amortized costs to fair values for non agency securities vs. other investment secu rities and changes in the five year U.S. Government bond rate during 2001 to 2009. Overall, t he mean and median ratios of the amortized cost to fair value are very close to 1 over the sample period. This indicates that the difference between amortized cost s and fair values of debt investment securities is small. These ratios are related to changes in the five year U.S. Government bond rate. When annual changes in five year U.S. Government bond rate are negative (positive), fair values in investment securit ies increase (decrease) due to the denominator effect (i.e., lower (higher) interest rate means higher (lower) bond price), which leads to the ratio being less (greater) than 1. For example, when there were substantial rate decreases in 2001 and 2002, the mean ratios of investment securities were less than 1 on average, while the mean ratios were above 1 when there were increases in the five year bond rate in 2005 and 2006.
35 At the end of 2008 when the financial crisis was peaking, the mean ratios for non a gency mortgage and asset backed securities (Investment NA_ABS ) became larger than 1 while the mean ratios for other investment securities remained close to (or below) 1. The mean (median ratio) of 1.17 (1.15) for Investment NA_ABS in 2008 was quite large gi ven the declining interest rate environment and substantial impairment charges on non agency securities made during 2008. Such discernible differences between fair values and amortized costs of non agency securities indicate large unrealized losses on non agency securities. It can be also interpreted under FASB ASC 320 that BHCs considered these large unrealized losses on non agency securities temporary and had the intent and ability to retain the securities for a sufficient period to allow for a recovery i n the market.
36 Table 5 1 Descriptive Statistics for t he Full Sample Price denotes the share price of bank. URGL TOTAL are unrealized gains and losses on investment securities excluding non agency MBS and ABS. URGL NA_ABS are unrealized gains and losses on non agency MBS and ABS. BVE is the book value of equity minus unrealized gains/losses from AFS investment securities. NI is income before extraordinary items. I and RGL are interest income and realized gains/losses on investment securities, respectivel y. URGLs, BVE, and NI are on a per share basis while I and RGL are deflated by book value of equity. SIZE is a log transformed total asset and GROWTH is the percentage change in total assets. Pre_Crisis is a time dummy equal to 1 for 2001 to 2006 and 2009
37 Table 5 2 Spearman and Pearson Correlations This table reports the Pearson (above the diagonal) and Spearman correlations for the full sample. P values are in parentheses. All variables are defined in Table 5 1.
38 Table 5 3 The Mean and Median Rati os of Amortized Costs to Fair Values of Non agency Securities from 2001 to 2009 Year Investment TOTAL Investment NA_ A BS Year T Bond R ate Mean Median Mean Median 2001 0.99 0.99 0.99 0.99 1.60 2002 0.98 0.98 0.99 0.99 0.74 2003 0.99 0.99 0.99 1.00 0.85 2004 0.99 1.00 0.99 1.00 0.46 2005 1.01 1.0 1 1.01 1.01 0.62 2006 1.01 1.01 1.01 1.01 0.70 2007 1.00 1 .00 1.02 1.0 1 0.32 2008 1.00 0.9 9 1.17 1.1 5 1.63 2009 0.99 0.99 1.1 7 1. 09 0.60 total 1.00 1.00 1.03 1.00 Investment TOTAL include s all investment securities excluding non agency MBS and ABS. Investment NA_ABS include s non agency MBS and ABS. The me an and median ratios of amortized costs to fair values for investment securities are computed based on bank year observations whose amortized costs as well as fair values for debt investment securities are non zero. Changes in the five year U.S. Government bond at t year are the difference between the ending t year rate and the beginning t year rate.
39 CHAPTER 6 EMPIRICAL RESULTS Value Relevance of Unrealized Gains and Losses on Non agency Securities Table 6 1 reports the regression results from estimatin g Eq. (1). The coefficients are estimated using OLS while standard errors are corrected by two clusters (banks and years). 1 The coefficient of unrealized gains/losses on investment securities excluding non agency MBS and ABS (URGL TOTAL ) is positive and sig nificant ( 1 =1.15, p value<0.01), indicating that these fair value revaluations are value relevant in the banking industry. The positive and significant coefficient of URGL TOTAL is consistent with previous results based on the banking industry (Barth 1994) and insuran ce industry (Petroni and Wahlen 1995). In addition, the estimated coefficient is not statistically different from 1, implying that investors are assigning dollar for dollar value to these unrealized changes. Central to my research interest, the estimated coefficient of unrealized changes on non agency MBS and ABS (URGL NA_ABS ) is not significant ( 2 =0.18, p value>0.10), not providing support for Hypothesis 1 a which states that unrealized gains/losses of non agency securities are value relevant. This implies that market investors may not perceive fair value revaluations for non agency securities as useful information in valuing bank equity. In other words, fair values for non agency securities do not have explanatory power in explaining the share price of banks beyond amortized costs. A potential explanation is that investors consider fair value esti mates for these securities 1 Petersen (2009) shows that these standard errors, clustering by two dimensions, produce less biased standard errors.
40 to be less reliable because they are less likely to be based on readily observable market prices from liq uid markets (Petroni and Wahlen 1995). In Table 6 2 I report the result of the coefficient comparison test to examine wheth er the valuation coefficient of URGL NA_ABS is smaller than the valuation coefficient of URGL TOTAL The coefficient o f URGL TOTAL is significantly larger than the coefficient of URGL NA_ABS (p value < 0.01). Thus, Hypothesis 1b is supported. Consistent with t he finding in Table 6 1 this suggests that the reliability of fair value estimates affects the value relevance of unrealized changes. Predictive Ability of Unrealized Gains and Losses on Non agency Securities for Future Earnings Hypothesis 2 a posits that unrealized changes for non agency securities are useful in predicting future interest income from those securities. To test this hypothesis, I examine whether unrealized gains/losses on these securities are associated with one or two year ahead interest i ncome. The OLS estimation results of regression Eq. (2) are presented in Table 6 3 The significant and positive coefficient o f URGL NA_ABS as well as o f URGL OTHER indicates that fair value revaluations from these securities are significant predictors of on e and two year ahead interest income. The magnitude of coefficients on the unrealized changes slightly increases for a longer prediction horizon (i.e., 0.106 to 0.147 for URGL NA_ABS ; 0.059 to 0.094 for URGL OTHER ). These results are consistent with the pr ediction in Hypothesis 2 a, suggesting that fair value estimates of non agency securities reflect information about the future interest income despite their lack of reliability compared to other debt securities. Examining future realized gains/losses in Eq (3) yields similar evidence supporting the link between unrealized changes of non agency securities and future
41 realized gains/losses, consistent with Hypothesis 2 b. T he estimate d coefficient of URGL NA_ABS in Table 6 4 is positive and significant in both the one and two year ahead realized gains/losses ( 2 = 0. 130 p value < 0.01 for one year ahead gains/losses; 2 = 0. 100 p value < 0.01 for two year ahead gains/losses). Valuation I mplications of Unrealized Gains and Losses on Non agency Securities during the Crisis This section examines the value releva nce of unrealized gains/losses on non agency securities which has been at the heart of the recent financial crisis due to the amplified valuation uncertainty. Hypothesis 3 a posits that the value relevance of unrealized changes on non agency securities decr eased due to reliability concerns. In contrast, Hypothesis 3 b posits that the value relevance of unrealized changes on non agency securities increased due to the fire sale expectation. Table 6 5 shows that the estimated coefficient of URGL NA_ABS is positi ve and statistically significant ( 2 =0.85, p value= 0.066). As can be seen, however, the interaction of Pre_Crisis and URGL NA_ABS produces an estimated coefficient that is negative and statistically significant ( 2 = 1.37, p value=0.068). This indicates an i ncrease in value relevance for unrealized changes for non agency securities during the crisis period. This is consistent with the fire sale expectation story in Hypothesis 3 b although the low liquidity in the market makes reported fair values less observab le and more subject to measurement errors. Turning to the predictive ability of unrealized gains/losses on non agency securities in the crisis period, Hypothesis 4 a posits that the predictive ability of unrealized gains/losses on non agency securities for future interest income is weaker in the financial crisis period. Table 6 6 presents results from estimating E q (5). The
42 estimated coefficient of unrealized gains and losses of URGL NA_ABS is positive but not significant in predicting one year ahead intere st income. As can be seen, however, the coefficient of the interaction of Pre_Crisis and URGL NA_ABS is positive ( 2 =0.126) and statistically significant (p value<0.01) in predicting one year ahead interest income. The inferences drawn from results using two year ahead interest income are similar to those drawn from results using one year ahead interest income, altho ugh the level of the adjusted R 2 is lower. However, the magnitude of coefficients of the interaction of Pre_Crisis and URGL NA_ABS dramatically increases for two year ahead interest income ( 2 =0.318 with p value < 0.05). These results imply that the pred ictive ability of unrealized changes on non agency the expected returns implied in fair value estimates are disproportionately high. Thus, the unrealized changes on non agency securities may not be a good indicator for future interest income from those securities. In contrast to the predictive ability of the unrealized gains/losses for future interest income, the relation between URGL NA_ABS and future realized gain s/losses becomes stronger during the crisis period as indicated by a significantly negative coefficient of the interaction of Pre_Crisis and URGL NA_ABS in Table 6 7 ( 2 = .203 with p value < 0.05 for one year ahead realized gains/losses; 2 = .321 with p val ue < 0.01 for two year ahead realized gains/losses). This is consistent with Hypothesis 4 b, suggesting that a liquidity crisis may have led banks to liquidate non agency securities at a loss or to write down these securities to the fair values in future pe riods.
43 Table 6 1 Value R elevance of Unrealized Gains/Losses on Non agency Securities Dep. Variable = Price Variables: Estimate Clustered Std Err Significance Year Intercept vary BVE 0.50 0.032 *** NI 3.22 0.135 *** URGL TOT AL 1.15 0.186 *** URGL NA_ABS 0.18 0.366 SIZE 1.56 0.133 *** GROWTH 1.42 0.769 No. Obs. 3,483 Adj. R Square 0.80 ***, **, and represent 1%, 5% and 10% significance, respectively. SIZE is the log transf ormed total assets f or the bank. GROWTH is the percentage change in total assets Other variables are defined in Table 5 1 The models are estimated using ordinary least squares (OLS). To mitigate the effects from extreme outliers and influential points, I eliminate observati ons with absolute values of studentized residuals greater than 2
44 Table 6 2 Test of Value R elevan ce Equality Coefficient Comparison T Stat. One side P value 1 ( URGL TOTAL 2 ( URGL NA_ABS ) 2.31 <0.01 All variables are defined in Table 5 1.
45 Table 6 3 Predictive Ability of Unrealized Gains and Losses on Non agency Securities for Future Interest Income ***, **, and represent 1%, 5% and 10% significance, respectively BV OTHER are the amortized costs of investment debt secur ities excluding U.S. Treasury securities, U.S. government agency obligations, and non agency securities Other variables are previously defined.
46 Table 6 4 Predictive Ability of Unrealized Gains and Losses on Non agency Securities for Future Realized Ga ins and Losses ***, **, and represent 1%, 5% and 10% significance, respectively RGL t+1 and RGL t+2 are one and two year ahead realized gains and losses on investment securities respectively RGL t are the current realized gains and losses from investm ent securities. Other variables are previously defined.
47 Table 6 5. Value Relevance of Unrealized Gains and Losses on Non agency Securities during the Crisis Dep. Variable = Price Variables: Estimate Clustered Std Err Significance Year Interce pt vary Pre_Crisis 3.86 0.462 *** BVE 0.50 0.033 *** NI 3.22 0.135 *** URGL TOTAL 1.13 0.189 *** URGL NA_ABS 0.85 0.464 URGL NA_ABS *Pre_Crisis 1.37 0.752 SIZE 1.57 0.133 *** GROWTH 1.44 0.765 No. Obs. 3,483 Adj. R Square 0.80 ***, **, and represent 1%, 5% and 10% significance, respectively. Pre_Crisis is a time dummy equal to 1 for 2001 to 2006 and 2009, and all other variables are as defined previously.
48 Table 6 6. Predictive Ab ility of Unrealized Gains and Losses on Non agency Securities for Future Interest Income during the Crisis ***, **, and represent 1%, 5% and 10% significance, respectively. Pre_Crisis is a time dummy equal to 1 for 2001 to 2006 and 2009, and all other variables are as defined previously.
49 Table 6 7. Predictive Ability of Unrealized Gains and Losses o n Non agency Securities for Future Realized Gaines and Losses during the Crisis ***, **, and represent 1%, 5% and 10% significance, respectively. Pre_ Crisis is a time dummy equal to 1 for 2001 to 2006 and 2009, and all other variables are as defined previously.
50 CHAPTER 7 SENSITIVITY OF RESUL TS The findings so far are based on the level models. This raises the possibility that the inferences are influ enced by correlated omitted variables. Thus, as a robustnes s check, I use difference model s to re estimate the regressions described in the prior chapter The returns models are somewhat sensitive to specification of the window period in which the returns are collected particularly for the crisis period. Nonet heless, the untabulated findings indicate that my inferences are consistent when based on the change change specifications. I also re estimate Eq (1) and Eq. (3) including proxies for default risk. The proxies include charge offs, Tier 1 capital r atio, a nd leverage The estimated coefficients (not reported) of interest are quantitatively similar to ones reported in Chapter 6
5 1 CHAPTER 8 CONCLUDING REMARKS This paper adds to the literature by examining the valuation implications of unre alized changes o n non agency securities along three dimensions. First, it provides f unrealized changes on non agency securities Consistent with prior literature that the reliability concerns affect the value relevance, my empirical test confirms that unrealized gains and losses on non agency securities whose fair values are less observable and more subjective to measur ement errors are not value relevant. Second, the predictive ability tests provide additional insights into information contained in fair value measurements for complex financial instruments. I provide evidence that fair value revaluations for non agency s ecurities are positively associated with future earnings realizations from those securities although investors do not generally consider them informative. This supports the notion that fair values for these complex instruments contain asset specific inform ation as well as macro information in spite of the reliability concerns. Finally, I examine whether the valuation properties of unrealized changes on non agency securities changed during the crisis. Unrealized gains/losses on these securities gain value r elevance in the crisis period, which is consistent with the fire sale expectation story. A stronger relation between unrealized changes on non agency securities and future realized gain/losses is observed in the crisis period. This provides an economic rat ionale for the observed value relevance during this period. However, after controlling for amortized costs unrealized gains/losses on non agency securities, which are allegedly not reliable, become less useful in predicting future inter est income
52 in the cr isis period predictive ability for future earnings during times of market instability.
53 LIST OF REFERENCES Aboody, D., M. E. Barth, and R. Kasznik. 1999. Revaluations of fixed assets and future firm performance: Evidence from UK. Journal of Accounting and Economics 26 (1):149 178. Ahmed, A. S., and C. Takeda. 1995. Stock market valuation of gains and losses on Journal of Account ing and Economics 20 (September): 207 225. American Bankers Association (ABA) 2008. Letter to SEC ( September 23 ) Ameri can Bankers Association. 2009. Fair Value and Mark to Market Accounting. http://www. a ba. com / Issues/Issues_FairValue.htm Ashcraft,A. and T. Schuermann. 2008. Understanding the Securitization of Subprime Mortgage Credit. Federal Reserve Bank of New York Staff Reports No. 318. Badertscher, B., J. Burks, and P. Easton., 2010. Fai r Value Accounting, Other Than Temporary Impairments, and the Financial Crisis. Working paper. University of Notre Dame. Barth, M. E. 1991. Relative measurement errors among alternative pension asset and liability measures. The Accounting Review 66 ( 3 ) : 433 463. Barth, M. E. 1994. Fair value accounting: Evidence from investment securities and the market valuation of banks. The Accounting Review 69 (1): 1 25. Barth, M.E. 2007 Standard setting measurement issues and the relevance of research. Accounting and Business Research 37(Special issue): 7 15. Barth, M., W. Beaver, and W. Landsman. 1996. Value disclosures under SFAS No. 107. The Accounting Review ( 4 ): 513 537. Barth, M., W. Beaver, and W. Landsman 1998. Relative v aluation roles of equity book value and net income as a function of financial health. Journal of Accounting and Economic s, Volume 25, Issue 1 : 1 34 Barth, M. E., Beaver, W. H., Landsman, W. R. 2001. The relevance of the value relevance literature for fina ncial accounting standard setting: Another view. Journal of Accounting and Economics 31: pp. 77 104. Barth, M. E. and G. Clinch. 1998. Revalued financial tangible, and intangible assets: Associations with share prices and non market based value estimates. Journal of Accounting Research 36 (3): 199 233.
54 Barth, M. E. and W. Landsman. 2010. How did Financial Reporting Cont ribute to the Financial Crisis? European Accounting Review 19 ( 3 ) : 399 423. Belsley, D. A., E. Kuh, and R. E. Welsch. 1980. Regression di agnostics: Identifying influential data and sources of collinearity New York, John Wiley & Son. Bhat, G., R. Frankel, and Martin, X. 2011. Panacea, Pandora's box, or placebo: Feedback in bank mortgage backed security hol dings and fair value accounting. Journal of Accounting and Economics 52 (2 3), 153 173. Bigman, D., and Desmond, M. 2009. Mark to messy accounting change. Forbes.com (April 02) Carroll, T. J., T. J. Linsmeier, and K. R. Petroni. 2003. The reliability of fair value versus historical c ost information: Evidence from closed end mutual funds. Journal of Accounting, Auditing & Finance 18 (1): 1 23. Eccher, E. A, K. Ramesh., and S. R. Thiagarajan. 1996. Fair value disclosures by bank holding companies. Journal of Accounting and Economics 22 (1 3): 79 117. Evans, M. E., L.D. Hodder, and P.E. Hopkins. 2011. Do fair values predic t future financial performance? Working Paper Indiana University Financial Ac counting Standards Board (FASB). 1980. Qualitative characteristics of accounting inform ation. Statement of Financial Accounting Concepts No.2 Norwalk, CT. Financial Accounting Standards Board (FASB). 1993. Statement of Financial Accounting Standards No. 115, Accounting for Certain Investments in Debt and Equity Securities Norwalk, CT: FAS B. Financial Accounting Standards Board (FASB). 2000. Special Report: Joint Working Group of Standard Setters Recommendations on Accounting for Financial Instruments and Similar Items Norwa lk, C T : FASB Financial Accounting Standards Board (FASB). 2006. Statement of Financial Accounting Standards No. 157, Fair value measurements Stamford, CT: FASB. Financial Accounting Standards Board (FASB). 2007. Statement of Financial Accounting Standards No. 159, The Fair Value Option for Financial Assets and Financ ial Liabilities Including an amen dment of FASB Statement No. 115. Norwa lk, CT : FASB Gorton, System, Proceedings of t he 2008 Jackson Hole Conference. Federal Reserve Bank of Kansas City.
55 Hughes, J and G. Tett. March 13, 2008. An unforgiving eye: Bankers cry foul over fair value accounting rules. Financial Times International Banking Federation (IBFed). 2008. Accounting for Financial Instruments Conceptual Paper London, Engla nd. Katz, I. and J. Westbrook. 2009. Mark to Market lobby buoys bank profits 20% as FASB may say yes. Bloomberg (March 29) Khurana, I. K., and M. Kim. 2003. Relative value relevance of historical cost vs. fair value: Evidence from bank holding companie s. Journal of Accounting and Public Policy 22 (January/February): 19 42. Kolev, K. 2010. Do Investors perceive marking to model as marking to myth? Early evidence from FAS 157 disclosure. Working paper. New York University. Kutner, M. H., C. J. Nachtshe im, J. Neter, W. Li. 2004. Applied Linear Statistical Models 5th ed. McGraw Hill/Irwin, Boston, MA. Landsman, W., 1986. An empirical investigation of pension fund property rights. The Accounting Review 61(4): 662 691. Laux, C. and Leuz, C., 2009. The cr isis of fair value accounting: making sense of the recent debate. Accounting, Organizations and Society 34: 826 834. Laux, C. and Leuz, C., 2010. Did fair value accounting contribute to the financial crisis? Journal of Economic Perspectives forthcoming. Leone, M. 2008. Bankers: Fair value is like throwing gasoline on a fire. CFO.com (April 14). Liu, J., Ohlson, J. A. 2000. The Feltham Ohlson (1995) model: Empirical implications. Journal of Accounting, Auditing and Finance 15, pp. 321 331 Moyer, S., 19 90. Capital adequacy ratio regulations and account ing choices in commercial banks. Journal of Accounting and Economics 13(2): 123 154. Nelson, K. 1996. Fair value accounting for commercial banks: An empirical analysis of FAS No.107. The Accounting Review 71 (2): 161 182. Nissim, D. 2007. Fair value accounting in the banking industry. Center for Excellence in Accounting and Security Analysis, Columbia University. Occasional Paper Series Ohlson, J. 1995. Earnings, Book Values, and Dividends in Security Va luation. Contemporary Accounting Research (Spring): 661 87.
56 Park, M. S., Park, T., Ro., B. T. 1999. Fair value disclosures for investment securities and bank equity: Evidence from SFAS No. 115. Journal of Accounting, Auditing and Finance 14 : 347 370. Pen man, S. H. 2007. Financial reporting quality: Is fair value a plus or a minus? Accounting and Business Research Special Issue : Interna tional Accounting Policy Forum 33 ( 44 ). Petersen, M. 2009. Estimating standard errors in finance panel data sets: Compari ng approaches. Review of Financial Studies 22 (1): 435 480. Petroni, K., and J. Wahlen. 1995. Fair values of equity and debt securitie s and share prices of property liability insurers. Journal of Risk and Insurance 62 (December): 719 737. Ronen, J. 200 8. To fair value or not to fair value: A broader perspective. Abacus 44, pp. 181 208. Journal of Accounting, Auditing & Finance 14 (Summer): 371 377. Ryan, S.G. 2007 Financial Instruments and Institutions. Second Edition John Wiley & Sons, Inc. Ryan, S. G. 2008. Accounting in and for the subprime crisis. The Accounting Review 83 (6) : 1605 1638. Ryan, S.G. 2008. Fair value account ing: Understanding the issues raised by the credit crunch. White Paper prepared for the Council of Institutional Investors : New York. Sabarwal, T., 2006. Common structures of asset backed securities and their risks. Corporate Ownership and Control 1 Schr and, C. 1997. The association between stock price interest rate sensitivity and disclosures about derivatives instruments. The Accounting Review 72 (January): 87 109. Schulz, S. M., & Hollister, J. 2003 Lobbying FASB: On accounting for investments. The Journal of Applied Business Research 19 : 85 103. Scholes M., G. Wilson, and M. Wolfson. 1990. Tax planning, regulatory capital planning, and financial reporting strategy for commercial banks. Review of Financial Studies 3(4): 625 650.
57 Securities Industr y and Financial Markets Association (SIFMA). Various publications, available at http://www.sifma.org/research/reports.aspx Shaffer, S., 2010, Fair value accounting: villain or innocent victim e xploring the links between fair value accounting, bank regulatory capital and the recent financial crisis, Working paper (Federal Reserve Bank of Boston Quantitative Analysis Unit). Song, C.J., W. Thomas, and H. Yi. 2010. Value relevance of FAS 157 f air v alue hierarchy information and the impact of corporate governance mechanisms. The Accounting Review 85(4), 1375 1410. U.S. Securities and Exchange Commission, 2008. Report and recommendations pursuant to section 133 of the Emergency Economic Stabilizati on Act of 2008: Study on Mark to market accounting. Available at: http://www.sec.gov/news/studies/2008/marktomarket123008.pdf Venkatachalam, M. 1996. Value Journal of Accounting and Economics 38 (2): 387 418. Vickery James I. and Wright Joshua, TBA Trading and Liquidity in the Agency MBS Market (August 2010). FRB of New York Staff Report No. 468. Warfield T.D. and T.J. Linsmeier, 1992. Tax planning, earnings management, and the differential information co nte nt of bank earnings components. The Accounting Review 67(3): 546 562.
58 BIOGRAPHICAL SKETCH Sukyoon Jung was born in 1978 in South Korea. He was a second child for his parents and has one elder brother. He received his B achelor of S cience in statistics from the University of Florida in 2004. He received his M aster of S tatistics from the University of Florida in 2006 He joined the accounting doctoral program at the University of Florida in 2006. He has been a PhD candidate in the Fisher School of Accoun ting since November, 2010.He received his doctorate degree in accounting in May 2012