Three Essays in the Economics of Marine Resource Management


Material Information

Three Essays in the Economics of Marine Resource Management
Physical Description:
1 online resource (122 p.)
Alvarez, Sergio
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Food and Resource Economics
Committee Chair:
Larkin, Sherry L
Committee Members:
Adams, Charles M
Weldon, Richard N
Valderrama, Diego
Allen, Micheal S


Subjects / Keywords:
economics -- ecosystem -- fisheries -- management -- oil -- portfolio -- resources -- spills -- valuation
Food and Resource Economics -- Dissertations, Academic -- UF
Food and Resource Economics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


During the past 200 years the world’s oceans have experienced drastic changes as a result of increasing pressure from a growing population whose appetites for seafood, recreation, energy sources and other marine resources seem to have no bounds. The oceans, once thought of as an indestructible and inexhaustible resource, are now under great pressure from overfishing, pollution, and modification of marine and coastal environments. We are now learning to cope with the resulting disruption of ecological processes while trying to maintain the flows of marine resources that we have grown so accustomed to. This dissertation focuses on the role that approaches based on economics can play in the management of marine resources. We focus on two particular problems facing the earth’s oceans and coastal areas. The first of these problems is water pollution and its effect on recreational use of marine resources. The second problem is the study and management of marine resources from a holistic perspective on an ecosystem-wide scale. Both of these problems have become central in today’s marine resource management, as a burgeoning population increases its standards of living and demands an ever increasing amount of seafood, energy, and recreational opportunities. This dissertation is an attempt to better understand the effects of this increased demand and propose some ways in which society’s management of marine resources can be improved. The first essay deals with the effects of pollution of the marine environment from the worst oil spill in U.S. history—the Deepwater Horizon—on recreational use of marine resources, specifically recreational fishing. The main result from this essay is the estimation of monetary compensation measures which could be used in damage assessment and litigation.  The second essay reviews the use of portfolio theory for natural resource management and illustrates the approach by developing portfolio selection models that could be used to set allowable catch policies in marine ecosystems.  The third essay summarizes the development of a simple ecosystem model in Ecopath with Ecosim, one of the leading platforms for development of such models, and its use for conducting analyses of an economic nature.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
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 Sergio Alvarez.
Thesis (Ph.D.)--University of Florida, 2013.
Adviser: Larkin, Sherry L.
Electronic Access:

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
lcc - LD1780 2013
System ID:

This item is only available as the following downloads:

Full Text




2 2013 Sergio Alvarez


3 To Cassie


4 ACKNOWLEDGMENTS The research described in this manuscript would have never been undertaken without the encouragement, support and guidance of Professor Sherry Larkin, who has advised me throughout many years of study. Professor Larkin has been heavi ly involved in all stages of development of the three essays in this dissertation, and her feedback on previous drafts has allowed me to drastically improve the analysis and the writing within Her invitation to join an interdisciplinary Sea Grant project on ecosystem modeling opened the door for the research that is documented in the last two essays. Similarly, an invitation to join her group conducting research on the impacts of the Deepwater Horizon oil spill gave me the opportunity to develop the models that are described in the first essay. Being part of those research groups has been an incredible learning opportunity Also it is due to her encouragement and support that I have had the opportunity to share this research with a diverse set of audiences, an experience which has been valuable both for the im provement of the essays and for my personal and professional development. My graduate committee, composed of Professors Charles Adams, Miche a l Allen, Diego Valderrama, and Richard Weldon, has also been a source of good ideas for the improvement of this manuscript and the methods and applications described herein. I am indebted to Professor Weldon for guiding my literature review on portfolio selection and its uses outside of finance. I also owe much gratitude to Professor Allen, with whom Ive learned much about fisheries science both in the classroom and as a member of his Sea Grant ecosystem research group. Similarly, Professor Adams helped me understand the wider implications of my research and has shown me the importance of going beyond academic exercise. Likewise, Professor Valderrama helped me think


5 through my numerical models and has offered great insight on how to make these models more realistic. I have been very fortunate to count with such knowledgeable researchers for guidance. The essay on the ef fects of the Deepwater Horizon spill also benefited from the advice of Professors John Whitehead and Tim Haab, who were instrumental in the beginning of the project and have me given insightful advice on key areas of the development of the essay I am also indebted to the audiences in the Stated Preferences session of the Agricultural and Applied Economics Associations 2012 meeting and the Graduate Student Organization weekly seminar at the Food and Resource Economics Department of the University of Flor ida for excellent feedback. I must, however, take credit for all errors and omissions in this document.


6 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURES .......................................................................................................... 9 ABSTRACT ................................................................................................................... 10 CHAPTER 1 GENERAL INTRODUCTION .................................................................................. 12 2 A REVEALED PREFERENCE APPROACH TO VALUING NON MARKET RECREATIONAL LOSSES FROM OIL SPILLS ..................................................... 17 Introduction ............................................................................................................. 17 A RUM to Measure Oil Spill Impacts ....................................................................... 20 Data Sources, Variables and Model Specification .................................................. 24 Results .................................................................................................................... 31 Site Quality Measure ........................................................................................ 31 Choice Models .................................................................................................. 32 Shore fishing mode .................................................................................... 33 For hire fishing mode ................................................................................. 3 4 Privat e and rental boat fishing mode .......................................................... 35 WTP to avoid spilled oil .............................................................................. 36 Discussion .............................................................................................................. 37 3 USING PORTFOLIOS TO MANAGE NATURAL RESOURCES ............................. 49 Introduction ............................................................................................................. 49 Portfolios for Natural Resource Management ......................................................... 53 Four Questions for Portfolio Development .............................................................. 55 Who is the Portfolio Manager? ......................................................................... 55 What Constitutes an Asset? ............................................................................. 56 How can we Measure Returns? ....................................................................... 58 What Constraints do DecisionMakers Face? .................................................. 60 Time Series Properties of Returns in Natural Resource Portfolios ......................... 61 Sustainability Constraints Using Historical Catch ................................................... 64 Portfolios of Total Allowable Catches Using Biomass ............................................. 67 The Colombian Pacific Coast ........................................................................... 67 CPC Results ..................................................................................................... 68 Portfolios of Total Allowable Catches Using Revenues .......................................... 70 The Florida Gulf Coast ..................................................................................... 70 FGC Results ..................................................................................................... 72


7 Discussion .............................................................................................................. 73 4 MODELING IMPACTS OF HUMAN ACTIVITIES ON MARINE ECOSYSTEMS: A STYLIZED APPROACH USING ECOPATH WITH ECOSIM .............................. 88 Introduction ............................................................................................................. 88 Ecopath: A Simple Model of a Tropical Ecosystem ................................................ 89 Mass Balance ................................................................................................... 90 Economic Input to Mass Balance Stage ........................................................... 91 Financial information on cost structure ...................................................... 91 Market and nonmarket prices ................................................................... 92 Economic Output of Mass Balancing ................................................................ 93 Ecosim: Time Dynamic Simulation and Policy Optimization ................................... 94 Fishing Effort Forcing Function ......................................................................... 94 Bionomic Dynamics: Fleets as Dynamic Predators .......................................... 95 Fishing Policy Search ....................................................................................... 97 Maximizing a simple objective function ...................................................... 99 Alternative: prevent cost > earnings ........................................................... 99 Maximizing social utility of a risk averse portfolio ..................................... 100 Discussion ............................................................................................................ 101 5 CONCLUSIONS ................................................................................................... 110 APPENDIX A COMPONENTS OF A VESSELS TOTAL OPERATION COSTS ......................... 112 B TOTAL ECONOMIC VALUE OF AN ECOSYSTEM ............................................. 113 LIST OF REFERENCES ............................................................................................. 115 BIOGRAPHICAL SKETCH .......................................................................................... 122


8 LIST OF TABLES Table page 2 1 Model Variables .................................................................................................. 41 2 2 Negative binomial predicted catch model of fish caught and kept ...................... 41 2 3 Models of shore fishing site choice ..................................................................... 42 2 4 Models of for hire fishing site choice .................................................................. 43 2 5 Models of private and rental boat fishing choice ................................................. 44 2 6 Per trip willingness to pay for prevention of the Deepwater Horizon spill ........... 45 2 7 MRIP participation estimates by state and mode (2010) .................................... 45 2 8 Total loss estimates (in Millions USD) ................................................................ 45 3 1 Empirical portfolios in the natural resource manag ement literature .................... 77 3 2 Colombi an Pacific Coast species groups ........................................................... 78 3 3 Measure of inefficiency in the Colombian Pacific Coast ..................................... 78 3 4 Fl orida Gulf Coast species groups ...................................................................... 79 3 5 Measure of inefficiency in the Florida Gulf Coast ............................................... 80 4 1 Basic input parameters for the hypothetical model ........................................... 104 4 2 Diet composition matrix .................................................................................... 104 4 3 Landings in the hypothetical ecosyst em (t/km2) ................................................ 104 4 4 Discards in the hypothetical ecosystem (t/km2) ................................................ 105 4 5 Cost structure of the fishing fleets in the hypothetical ecosystem ..................... 105 4 6 Market prices in the hypothetical ecosystem (hypothetical currency) ............... 105 4 7 Non market prices in the hypothetical ecosystem (hypothetical currency) ....... 105 4 8 Economic output from the mass balance procedure in Ecopath ....................... 106 4 9 Employment by fleet in the hypothetical ecosystem ......................................... 106


9 LIST OF FIGURES Figure page 2 1 Federal fishery closure in the Gulf of Mexico ...................................................... 46 2 2 Density histograms of important variables .......................................................... 46 2 3 Ten coastal regions in the Southeastern U.S. .................................................... 47 2 4 Conditional logit estimates of per trip willingness to pay .................................... 47 2 5 Mixed logit estimates of per trip willingness to pay ............................................ 48 3 1 Possibl e answers to the four questions for portfolio development ...................... 80 3 2 Historical returns from Gag grouper (Mycteroperca microlepis) ........................ 81 3 3 Map of Colombia. ............................................................................................... 82 3 4 Efficient frontier of biomass harvest in the CPC ................................................. 83 3 5 Transition map of prescribed harvest policies in the CPC .................................. 84 3 6 Efficient frontier of fishing revenues in the FGC ................................................. 85 3 7 Transition map of fishing revenues in the FGC ................................................... 86 3 8 Transition map of prescribed harvest policies in the FGC .................................. 87 4 1 Flow chart of the hypothetical ecosystem ......................................................... 106 4 2 Expected biomass trends for economically important species .......................... 107 4 3 Biomass trends in the hypothetic al ecosystem under openaccess .................. 107 4 4 Iteration trends of the different objectives under profit maximization ................ 107 4 5 Recommended effort trends to allow profit maximization ................................. 108 4 6 Biomass trends under profit maximization ....................................................... 108 4 7 Iteration tre nds when maximizing for risk averse social utili ty .......................... 108 4 8 Recommended effort trends to maximiz e risk averse utility .............................. 109 4 9 Biomass tren ds under risk averse social utility maximization ........................... 109


10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THREE ESSAYS IN THE ECONOMICS OF MARINE RESOURCE MANAGEMENT By Sergio Alvarez May 2013 Chair: Sherry L. Larkin M ajor: Food and Resource Economics During the past 200 years the worlds oceans have experienced drastic changes as a result of increasing pr essure from a growing population whose appetites for seafood, recreation, energy sources and other marine resources seem to have no bounds. The oceans, once thought of as an indestructible and inexhaustible resource, are now under great pressure from overf ishing, pollution, and modification of marine and coastal environments. Society is learning to cope with the resulting disruption of ecological processes while trying to maintain t he flows of marine resources on which it depends This dissertation focuses on the ro le that approaches based on economic inquiry can play in the management of marine resources. We focus on two particular problems facing the earths oceans and coastal areas. The first of these problems is water pollution and its effect on recreati onal use of marine resources. The second problem is the study and management of marine resources from a holistic perspective on an ecosystem wide scale. Both of these problems have become central in todays marine resource management, as a burgeoning popul ation increases its standards of living and demands an ever increasing amount of seafood, energy, and recreational opportunities.


11 This dissertation is an attempt to better understand the effects of this i ncreased demand and propose ways in which societys management of marine resources can be improved. The first essay deals with the effects of pollution of the marine environment from the worst oil spill in U.S. history the Deepwater Horizonon recreational use of marine resources, specifically recreational fishing. The main result from this essay is the estimation of monetary compensation measures which could be used in damage assessment and litigation. The second essay reviews the use of portfolio theory for natural resource management and illustrates the approach by developing portfolio selection models that could be used to set allowable catch policies in marine ecosystems. The third essay summarizes the development of a simple ecosystem model in Ecopath with Ecosim one of the leading platforms for deve lopment of these models, and its use for conducting analyses of an economic nature.


12 CHAPTER 1 GENERAL INTRODUCTION The human population has changed dramatically in the past 200 years. Compared to previous population growth rates, we have recently experienced an outright explosion. It took more than 100,000 years since the first Homo sapiens walked the plains of Africa for the worlds population to reach the 1 billion mark around AD 1810. Just over a century later, by 1930, the worlds population had doubled. By the turn of the millennium, more than 6 billion humans populated every region in Earth, perhaps with the exception of Antarctica (Brown et al. 1999). We could argue that this population explosion was made possible by an equally dramatic expansion in mankinds capacity to alter its environment. Advances in several fields of science, for instance, now allow us to live comfortably and in large numbers in places like Florida and the Brazilian Amazon, where tropical diseases and other inconveniences kept our forefathers at bay for generations. These advances have allowed us to dramaticall y d ecrease our mortality rates thus making it possible for our numbers to swell to the levels we see today. One of the first ecosystems to experience mankinds quest to transform the Earth may have been the sea, or more specifically, the waters off coastal S outh Africa. Before the first modern humans migrated out of Africa about 120,000 years ago, a small group of Homo sapiens learned to harvest fish and shellfish near a place called Pinnacle Point close to the Indian O cean. This development turned the coastl ine into an attractive place for settlement and movement, and allowed these early humans to be better suited to survive changes in environmental conditions. Interestingly, use of pigment and development of advanced stone tools also took place in the same l ocation and time period. The harvest of shellfish, a relatively sedentary food source, may have allowed


13 for social organization of higher complexity and lower mobility, which in turn enhanced the development of symbolic behavior or use of pigment for relig ious or artistic motives. This ability to harvest marine resources also played an important role in determining the migration routes that humans would follow to colonize the rest of the world (Marean et al. 2007). Thus it appears that the resources of the sea played a crucial role in the development of modern man, and that the relationship between humans and the oceans is as old as mankind itself. The origins of fishing as a subsistence activity may be as old as mankind, with fishing playing an important role in the development of our species. Recreational fishing, or fishing for fun, is not a new trend either. The earliest evidence of fishing for fun is from Egyptian murals painted more than 4,000 years ago depicting members of the aristocracy fish ing in ma n made ponds (Pitcher and Hollingworth 2002). Furthermore, the fish and seafood trade, or fishing for profit, is also a relatively old activity. In ancient Rome, for instance, trade routes were established to transport catfish from the lower Nile as far as modern day Turkey (Arndt et al. 2003). Also, the global trade in cod ( Gadus morhua) began in medieval times and played an important role in the colonization of the Americas (Kurlansky 1998). Some ancient cultures that were especially dependent on marine r esources even developed informal systems for managing fisheries, as was the case in some islands in tropical Oceania (Johannes 1982). Harvest of marine resources for different motives has taken place in different parts of the globe for thousands of years, and until recently it seemed that mankind could have no lasting impact on the worlds oceans. As the biologist Thomas Huxley


14 noted in 1883, all the great sea fisheries are inexhaustible: that is to say that nothing we do seriously affects the number of fi sh (quoted in Gordon 1954, p.126). The natural fluctuations in harvests, which puzzled fishermen for centuries, were blamed on pollution, migratory patterns, changes in climate and ocean currents, as well as more obscure reasons such as the ringing of church bells and the wickedness of people (Smith 1994, p. 9). Indeed, Gordons (1954) notion that fishery resources are the richest and most indestructible available to man was prevalent in fishery management circles until a relatively short time ago. How ever, changes in human demography and technological developments have impacted the way s in which we interact with marine resources and the effect humanity has on marine ecosy stems. T he human population of the world has grown more in the last century than i t had ever before and more than 60% of all people on Earth live in coastal areas (Brown et al. 1999). Furthermore, developments such as high powered engines, sonar and global positioning system technology, refrigeration and low cost transportation have in creased the efficiency of fishing activities and the demand for seafood. Some of these developments have also expanded the set of resources that we expect to obtain from the oceans. For instance, a large portion of our energy sources come from offshore oil drilling or are transported by sea, both of which result in benthic ecosystem impacts and higher pollution risks for marine ecosystems. The effect of increased global demand f or seafood has become evident in the status of fish stocks around the world. FAO (2010) estimates that the proportion of under exploited or moderately exploited stocks declined from 40% in the mid 1970s to 15% in 2008. Thirty two percent of global stocks are deemed to be over exploited,


15 depleted, or recovering from depletion. The global marinecapture fishery production peaked in the mid1990s and has remained stable since. However, the worldwide annual per capita consumption of fis h, estimated to be around 17 kilograms, continues to grow despite the burgeoning population. This growth in consumption is due to the recent dramatic growth of aquaculture. The composition of the global fish catch has also been shifting away from long lived piscivorous ground fish, such as cod, haddock, and hake, toward invertebrates and planktivorous pelagic fish, a process known as fishing down marine food webs (Pauly et al. 1998). The effects of capture fisheries may go beyond the impacts on the target stock or fish popul ation. In fact, there i s a growing body of evidence to support the existence of ecosystem effects of fishing (for a summary, see N ational Research Council 2006, C hapter 2). Brodziak and Links (2002) overview of the recent history of the Georges Bank off t he coast of New England provides a telling example of the ecosystem effects of fishing. For centuries European and American vessels have fished the Georges Bank and have targeted several species of groundfish. However, the number of groundfish vessels incr eas ed dramatically in the late 1970s, wh ich resulted in overfishing of the groundf ish stocks. The eventual collapse of the groundfish stocks which were the dominant biomass group in the ecosystem, opened the way for pelagic species such as herring to beco me the dominant biomass group. This effect of fishing on the ecosystem or ecosystem overfishing is in part responsible for the nonrecovery of groundfish even after fishing effort has been drastically curtailed. The three essays in this dissertation are an effort to better understand the changing relationship between mankind and the oceans. The first essay addresses


16 water pollution and its effect on recreational use of marine resources. Specifically, we study the impacts of the Deepwater Horizon spill, whic h took place in 2010, on the choices of recreational fishers in the Southeastern U.S. and use our analysis to estimate monetary compensation measures that could be used for damage assessment and litigation. The second essay reviews the use of portfolio the ory for natural resource management and illustrates the application of portfolio selection models that could be used to set allowable catch policies in marine ecosystems. The third essay summarizes the development of a simple ecosystem model in Ecopath wit h Ecosim, one of the leading platforms for development of such models, and its use for conducting analyses of an economic nature.


17 CHAPTER 2 A REVEALED PREFERENCE APPROACH TO VALUING NON MARKET RECREATIONAL LOSSES FROM OIL SPILLS Introduction Oil spills and other anthropogenic environmental disasters have economic consequences that transcend business losses and property damages. Such nonmarket losses include those accrued by recreational users. Measurement of the nonmarket economic consequences o f man made disasters is often complicated by the ex post nature of such analysis, as most researchers do not have the necessary foresight to collect data before, during, and after the event (Grigalunas et al. 1986). Researchers are thus forced to rely on c ounterfactuals and hypothetical scenarios to recreate what behavior would have been in an alternate state of the world where the event does not occur. However, continuous data collection efforts such as that of the Marine Recreational Information Program (MRIP) in the US provide opportunities for study of the nonmarket impacts of oil spills and other coastal and marine disasters in an ex ante/ex post fashion. Considerable research has been conducted on the valuation of the nonmarket impacts from oil spills in the past 35 years. The earliest research relied mainly on market data or valuation estimates from similar areas or ecosystems to determine damages. For instance, Grigalunas et al. (1986) used a suite of direct and surrogate market methods to value losses to commercial and recreational interests from the 1978 Amoco Cadiz oil spill off the coast of Brittany, France. Similarly, Mazzotta et al. (1994) estimated losses from the 1986 Amazon Venture oil spill in the Savannah River, Georgia, using benefits tr ansfer and focusing their discussion on restoration of the affected resources. The 1989 Exxon Valdez oil spill in Prince William Sound, Alaska, is


18 perhaps the most widely studied oil spill in terms of nonmarket valuation. Two seminal articles remain as pr oof of the methodological advances that resulted from these efforts. Hausman et al. (1995) used a random utility or choice modeling approach to estimate recreational losses from the Exxon Valdez, while Carson et al. (2003) use a contingent valuation survey to measure losses of nonuse value to the residents of the Continental US. To set reference points for the damages from oil spills, Carson et al. (2004) and Van Biervliet et al. (2005) use contingent valuation to obtain estimates of damages from potential (hypothetical) spills in the central coast of California and the coast of Belgium, respectively. Most recently, Loureiro et al. (2009) and Loureiro and Loomis (2012) also used the contingent valuation method to value the effects of the Prestige oil spill in Spain on residents of different Spanish provinces, as well as residents of the United Kingdom and Austria. The Deepwater Horizon (DWH) oil spill presents an opportunity to use MRIP to analyze the effects of a manmade environmental disaster before, duri ng, and after the event. On April 20, 2010, the largest marine oil spill in US history began off the coast of Louisiana in the Gulf of Mexico. By the time the leaks in the DWH offshore drilling rig were fully contained on July 15, 2010, surface oil had reached several areas along the Gulf Coast. A large expanse of the Gulf of Mexico Exclusive Economic Zone (EEZ) had been closed to fishing, accounting for 37% of the US EEZ at the height of the spill (Figure 2 1 ). State waters throughout the Gulf were also im pacted, resulting in closures ranging from 95% of state waters in Mississippi, 55% in Louisiana, 40% in Alabama, and 2% in Florida (Upton 2011). Federal and State authorities responded to the oil spill by deploying largescale cleanup and mitigation effort s throughout the affected areas.


19 For example, the state of Florida launched a massive campaign to ensure potential beachgoers and recreational anglers that coastal areas of the state were still open for business and included freefishing days during w hich the state fishing license requirement was temporarily lifted. Aside from the effects on marine ecosystems and the commercial fisheries that depend on them, the DWH spill can also be expected to have impacted recreational use of coastal and marine resources in many areas of the Gulf. Under the Oil Pollution Act of 1990 (OPA), Federal, State, and Tribal authorities have standing to claim and recover losses on behalf of the public from responsible parties. Recoverable damages include both the costs of pri mary restoration and losses in value from the time of the incident until recovery, where losses include direct use and passive use values (Jones 1997). Recreational use of natural resources falls under the direct use value category, implying that beachgoe rs and anglers may be entitled to compensation for interim losses in value in the form of compensatory restoration (Mazzotta et al. 1994; Jones and Pease 1997; Flores and Thacher 2002; Parsons and Kang 2010). More importantly, recreational demand may offer a litmus test of whether primary restoration under Natural Resource Damage Assessment (NRDA) plans have been effective by determining whether affected locations are as desirable after restoration as they were before the oil spill. In this study, we develop an approach to value quality changes in natural resources with recreational use values that occur throughout large regions and where quality changes occur asymmetrically through time and space. Our empirical application uses MRIP intercept data from the Southeast US to examine the effects of the DWH oil


20 spill on recreational anglers and estimate the relevant compensation measures that would make anglers whole pursuant to the OPA specifications. Specifically, our empirical model is designed to test the n ull hypothesis that the DWH oil spill had no impact on marine angler behavior in the Southeast US. We contribute to the literature by using two innovative aspects in our analysis, some of which are necessary due to the large scale of impacts of the DWH oil spill. First, given the large spatial extent of the DWH oil spill, we use a coarser definition of geographical zones than is usually used in nonmarket valuation, but one that permits substitution between areas that were or could have been affected by the spilled oil and those that were not. Second, we develop a quasi real time binary measure of oil spill impacts using fishery closure maps published during the DWH oil spill, which gives us a closer glance at behavioral changes that occurred as a result of the oil spill than those used in previous studies dealing with spills. Our approach is appropriate in situations where the quality change under study affects large regions, where the quality change is experienced differently at different points in time, or as was the case with the DWH oil spill, where the quality change affects large areas asymmetrically through time. A RUM to Measure Oil Spill Impacts A random utility model (RUM) of site choice is well suited to analyze situations where recreational users have alternative locations to visit (Bockstael et al. 1991). The RUM models human behavior based on observed choices assumed to be driven by the characteristics of each alternative. The RUM was introduced by McFadden (1974, 1978), who developed the conditi onal and nested logit specifications. Bockstael et al. (1987) used the nested logit to model recreational choices by swimmers in the Boston area and value the associated water quality attributes. Morey et al. (1991) also used a


21 nested RUM to control for the decision of whether or not to participate in recreational fishing in the Oregon coast and value the elimination of fishing opportunities at particular locations. Similarly, Greene et al. (1997) used a nested RUM of fishing in the Tampa Bay region to value the loss of access to fishing grounds. Kaorus (1995) nested site choice model values changes in combinations of quality for anglers in coastal North Carolina. Similarly, Thomas et al. (2010) created a model of site choice using ramp access points as nests and onwater locations as the elemental sites, and use their model to value the benefits of maintaining and improving boat ramps in the Fort Myers, Florida area. The commonly used RUM procedure developed by McFadden, the conditional logit, is limited b y the assumption of independence of irrelevant alternatives (IIA). The IIA assumption is violated when unobserved characteristics are correlated between alternatives. While the nested logit allows the researcher to establish a priori which alternatives are expected to be correlated with each other through the specification of nests, some problems may not be amenable to such specification or the specification may not be intuitive. The mixed or random parameters logit (Train 1998, 2003) and the latent class l ogit (Boxall and Adamowicz 2002) do not rely on the restrictive IIA assumption, with the added benefit that heterogeneity in the sample can be explicitly accounted for either by the estimation of distributions of the parameters in the former, or separation of the sample into latent classes in the latter. Haab et al. (2012) show the difference in results due to different specifications of the RUM in a study that estimates the marginal benefits of an additional fish caught in marine recreational trips to the Southeast US.


22 The RUM is a model of choice among a set of available alternatives. In the case of recreational fishing, the RUM models the choice an individual angler makes between available fishing sites because of the attributes of the site, such as the c osts of travel, the historic and expected catches of fish, and the popularity or accessibility of the fishing site (Bockstael et al. 1991). In our case, we also consider the presence of oil near the coastline as an indication of site quality. Following Tr ain (2003), an angler ( n ) decides to go fishing and must choose from among a set J of available alternatives ( j = 1, 2, J). Angler n selects alternative j if that is the alternative that maximizes their utility: > ( 2 1) However, the anglers utility is unobservable. Instead, we observe some attributes of the fishing locations as faced by the angler ( qnj), including travel costs ( TCnj) or the costs of accessing the location. Based on these observables, we specify an indirect function t hat relates these attributes to the anglers choice, Vnj( TCnj, qnj). Since there are likely to remain unobserved factors in the anglers utility function, we can express utility as a deterministic component ( Vnj) and a stochastic component ( nj) that captur es the factors that affect the anglers well being but are not observed, such that: = + ( 2 2) where the joint density of the random component n = [ n1,, nJ] is denoted f ( ). The probability that angler n chooses alternat ive j is thus given by = Pr ( + > + ) = Pr ( > ) ( 2 3)


23 Using the density of the stochastic terms f ( ), this cumulative probability can be expressed as = > ( ) ( 2 4) where I () is the indicator function that equals one when the expression in parenthesis is true and zero otherwise. Assuming that the stochastic terms are independent and identically distributed extreme value yields McFaddens (1974) conditional logit, where the probability takes the form = ( 2 5) The conditional logit can be estimated using maximum likelihood methods. We specify the indirect utility function ( Vnj) to be linear in the attributes of the alternatives to ease derivation of the measures used to assess the impacts of the DWH oil spill. The utility that angler n obtains from choosing alternative j can then be expressed as = + ( 2 6) where qnj is a vector of attributes and is a vector of model parameters. Preferences for the different attributes of available alternatives are reflected in the estimated parameters ( ). Heterogeneity in preferences can be introduced by allowing a cumulative density function f ( ) for the estimated parameters. In this case, the choice probability becomes = ( ) ( 2 7) which is referred to as the mixed or random parameters logit model, which can be estimated using simulated maximum likelihood methods.


24 To examine the welfare impact of a change in a quality attribute across all sites, we decompose the indirect utility function explicitly into travel costs ( TCnj) and other attributes so that angler n s utility from choosing alternative j is given by = + + ( 2 8) where is interpreted as the marginal utility of income. Now, suppose we want to evaluate the welfare impact of a proposed change in quality from the initial level ( qnj) to an alternative level ( qnj* ), where = + ( 2 9 ) The welfare measure for such a quality change is given by the difference in the sum of the indirect utilities across sites under both states of the world, weighted by the marginal utility of income as follows = ( 2 10) which, as shown by Haab and McConnell (2002), reduces to = ( 2 11) This is the per trip welfare measure or willingness to pay to prevent a decrease in quality, or conversely, to purchase an incremen t in a quality attribute. Accordingly, the sign of W depends on whether anglers perceive the quality change as an improvement or damage to the fishing experience. In our case, q is an indicator of the impacts of the DWH spill. Data Sources, Variables and M odel Specification The Marine Recreational Information Program (MRIP), formerly known as the Marine Recreational Fishery Statistics Survey (MRFSS), conducts intercept surveys of saltwater recreational anglers throughout the year (Hicks et al. 2000). These surveys


25 focus on the level and composition of catch to provide statistics on catch and fishing effort by species. The data are reported by twomonth periods referred to as waves. For this study, we assembled the unweighted intercept datasets for years 2006 through 2010, the latter being the year in which the DWH oil spill occurred. Variables of interest in the dataset are shown in Figure 2 2 Two main issues have been identified regarding the estimation of site choice models with the MRIP data (Whitehead an d Haab 2000). The first is the computing power required to estimate models with many alternatives. The second is the limited scope of demographic information collected during intercept surveys. Researchers can aggregate or eliminate sites to reduce the num ber of choices modeled and several studies have investigated the merits of each approach. Parsons and Needleman (1992) developed a site choice model of fishing in Wisconsin lakes to analyze the effect of aggregating recreational sites on estimates of the v alue of different site characteristics, and caution against the use of aggregation schemes due to the introduction of bias on welfare estimates. Parsons and Kealy (1992), using the same study region, modeled choices of recreational users engaging in differ ent activities and use the model to analyze the effect of using randomly drawn subsets of alternatives rather than the complete set to estimate the RUM parameters. These two studies suggest that using random draws from the alternative set may be more effec tive than aggregating sites as a way to ease computational requirements. These conclusions are echoed by Feathers (1994) analysis of sampling and aggregation using a site choice model of sport fishers in Minnesota. Lupi and Feather (1998) offered a pragmatic solution in which unimportant sites are aggregated, while those that are heavily visited or will be affected


26 by policy changes are kept in their elemental form. Whitehead and Haab (2000) used a different approach by examining the use of distance and hi storical catch to eliminate sites that are either too far or too unproductive to warrant inclusion as a viable choice; they found that results are not significantly affected by the elimination of nonviable choices determined using these criteria for marine recreational anglers in the Southeastern US. We use suggestions from these most recent studies to help define the size and scope of regions in this study. MRIP intercepts in the Southeast US are conducted in several locations from Louisiana to North Carolina. Several authors who have worked with MRIP have used county level aggregation to define their sites (e.g., Morey et al. 1991; McConnell et al. 1995; Whitehead and Haab 2000; Haab et al. 2001; Hindsley et al. 2011; Haab et al. 2012). We aggregate further given the objective to measure the impact of the DWH oil spill, which occurred in a large area of the Gulf of Mexico, and define a set of 10 regions. The states of Louisiana, Mississippi, Alabama, Georgia, South Carolina, and North Carolina are defined as individual regions. Florida is divided into four regions: Northwest, Southwest, Florida Keys, and Florida Atlantic (Figure 2 3 ). We then develop a binary measure of oil spill impacts using Federal fishery closure maps (Figure 2 1 ). During the DWH spill, the National Oceanic and Atmospheric Administration (NOAA) instituted a series of fishery closures in Federal waters throughout the affected areas of the Gulf of Mexico. The closures changed through time as the area affected grew until midJuly and shrank thereafter once the well was capped. We create the binary index of spill impacts using a rule of thumb that defines a region in the study to be affected by the spill at a given period of time if federal waters


27 are closed up to the state line. For each reg ion we then have six different waves or time periods with their own indicator of oil spill impacts. Therefore, we include the effects of the DWH oil spill in a spatially and temporally explicit manner. While the availability of GIS based data might suggest a finer spatial resolution, reducing the size of the region to capture such detail would necessarily reduce the overall geographic scope of the model or increase the number of discrete alternatives above what is manageable. Given that we are primarily int erested in incorporating broad substitution patterns on a scale suitable for trustees (i.e., states and tribal nations), we opt for an approach that allows us to look at the oil spill at a regional scale throughout time on a somewhat coarse manner. The final alternative set for each angler includes the 10 alternative sites in six possible seasons for a total of 60 available alternatives in a given year. The second issue with respect to using MRIP data for site choice models is the need to have estimates o f travel costs for all alternative trips in addition to the observed trip. Since the zip code of their permanent home residence is the only information we have on each angler, we created a matrix of distances between all alternative sites and all counties of residence of intercept respondents. Travel related expenses incurred by angler n traveling to site j were calculated as twice the product of the driving distance and the standard IRS mileage rate in 2009. In order to engage in a recreational activity such as fishing, participants must not only spend some of their disposable income to pay for travel related expenses, but in many cases must give up work or other wageearning opportunities. This opportunity cost of time must also be factored into travel cos t recreational demand models for a more accurate measure of the value of the recreational experience (Palmquist et al.


28 2010). In our case, the anglers opportunity cost of time is assumed to equal the median income in his or her county of residence from the most recent U.S. Census. With these assumptions, the total travel costs ( TC ) are the sum of travel and time related expenses: = 2 ( 0 55 ) + ( 2 12) where Dnj is the oneway distance between the anglers county of residence and the fishing site, = 0.33 indicates the share of the value of travel time used to account for the cost of leisure time (this is a commonly used share and was first proposed under Execut ive Order 11747), and med_incomen is the median annual income in the respondents county of residence during the year the fishing trip took place. Median income is divided by 2,080, the number of full time hours potentially worked in a year. The distance t erm is divided by 40, reflecting the assumption that road travel takes place at an average speed of 40 miles per hour (Haab et al. 2001). Total travel costs are expected to be inversely related to site choice for all modes. Our analysis rests strongly on t he two attributes developed above, the oil spill measure ( spill) and travel costs ( TC ), as well as four other types of attributes found in the MRIP dataset (Table 1). The first type of attributes, denoted Pn, are those related to the angler and include the number of days the angler has fished in the last year ( ffdays12 ), the median income in the anglers county of residence ( med_income), and whether the angler is a shore based fisher ( mode1), a for hire based fisher ( mode2) or a private/rental boat based fi sher ( mode3). The second type of attributes, which we denote Lj, have to do with the area or location in which fishing takes place and include the historical catch per unit effort for all species combined ( hckr ), the number of anglers


29 in the sample who vis ited the site ( popularity ), the number of interview locations or access points in a particular area ( sites ), and an indicator for whether the site is located in the Gulf of Mexico ( gulf ). The third type of attributes, denoted Ej, is related to the season i n which the fishing trip takes place and includes indicators for the three warmest seasons ( spring, summer and fall). The fourth type of attributes, which we denote Tnj, are related to the specific trip and include the length of the fishing trip in hours ( hrsf ), an indicator of whether the angler was targeting a particular fish species at the time ( targ ), and whether fishing took place in the ocean, close to shore ( area1), in the ocean, offshore ( area2), or in an inland or inshore location ( area3). One com monality among site choice models of recreational fishing is the use of catch as a quality attribute, primarily because catch rates are policy relevant. While triplevel data includes catch and fishing location, it does not include expected catch at alternative sites the individual considered but did not visit. McConnell et al. (1995) proposed a two stage process in which a catch equation is estimated in the first stage and used to create a quality attribute for the alternatives used in the site choice model, the second stage. This process gives a proxy of the anglers ex ante expectation of catch and thus expected quality of the fishing experience at the chosen site. Such an approach allows variation in site attributes among alternatives at the trip level. Haab et al. (2012) follow this approach to value additional catches of different marine species of management interest in the Southeastern United States using a series of site choice models that allow for heterogeneity in anglers preferences. One of their important findings is that heterogeneous angler preferences are a staple of the MRIP data, in particular fishers response to travel costs.


30 We consider these findings by first specifying a negative binomial model of the catch and keep rate experienced by each angler in all observed trips ( Cnj). The predicted catch serves as an individual specific indicator of site quality, which allows greater heterogeneity in estimation of the site choice models (McConnell et al. 1995). We use three types of regressors that can be expected to influence the number of fish that an angler might be able to catch in a given trip, namely those that are related to the angler ( Pn), location ( Lj), and fishing trip ( Tnj): = + + + + ( 2 13) where 0 is the intercept term, 1, 2, and 3 are vectors of estimated parameters and nj is an error term. We refer to this model as the catch model. We then consider three types of site choice models, one for each fishing mode. This segregation allows us to estimate distinct compensation measures for shore based, for hire, and private or rental boat fishers, as well as to examine whether preferences differ across modes. Separating out the modes also has the added benefit of reducing the dimensionality of the models and allows for a better representation of site choice, which is a priori expected to vary by mode. We construct our indirect utility function by excluding characteristics of the individual anglers but maintain independent variables that can be expected to influence the probability of choosing a particular combination of fishing site and season. For instance, we include travel costs to the site ( TC ), expected catch ( Ln ( nj)), the indicator of whether the DWH oil spill was affecting a given area during a given wave ( spill ), as well as locationspecific attributes ( Lj) and seasonal indicators ( Ej) = + + + + + ( 2 14)


31 where is the marginal utility of income, 1 and 2 are parameter estimat es associated with the respective variables, 3 and 4 are vectors of parameter estimates and nj is an error term. The models are estimated for 2010, the year in which the oil spill occurred. Similarly, we use two different estimation techniques for comparison of results across specifications of the RUM, the conditional and the mixed logit. Thus, we have a total of four distinct models for each fishing mode considered. Results Site Quality Measure The independent variables in the catch model perform well as predictors of the catch and keep rate an angler experiences in a particular trip (Table 22). All the estimated coefficients are statistically significant at the 0.01 level, except for the coefficient on popularity, which is not significantly different f rom zero. The sitespecific attribute historic catch per unit effort ( hckr ) has the expected effects on the numbers of fish caught and kept. High historic catches are likely related to high stock levels and suitable fish habitat, and it is no surprise that anglers fishing in locations with high historic catches enjoy high catch rates Other than fishing experience, it is difficult to assess ex ante what the effects of angler related attributes are on catch. It is clear that fishing experience ( ffdays12) can be expected to have a positive impact on catch, as more experienced anglers possess some knowledge regarding bait, tackle, technique and fish behavior that less experienced anglers lack, and we observe this effect in our catch model. In addition, our model indicates that median income has a small negative effect on catch. We also find that anglers who fish from shore experience lower catches than those who fish from


32 private and rental boats, while those who use the for hire sector generally catch more fish than either shore or private boat anglers. Besides the kind of area in which fishing takes place, the expected effects that trip specific attributes have on catch are clear. All things held equal, longer fishing trips ( hrsf ) can be expected to yield higher catches. Similarly, targeting a particular species, which involves a series of decisions regarding bait, location, tackle, and fishing technique, among others, can be expected to result in higher catches. Both of these effects are observed in our model. In addition, we find that anglers in the Southeast generally enjoy higher levels of catch when fishing in the open ocean than they do when fishing inshore, and that anglers fishing offshore experience higher catches than those fishing in the open ocean close to shore. Choice Models Estimation of discrete choice models allows us to obtain information regarding peoples preferences toward the good or service that is being chosen. In the case of recreational fishing, the coefficients in a site choice model indicate whether the presence or absence of a given attribute increases or decreases the likelihood that an angler will choose an alternative or seasonsite combination. In our case, these parameters can also be interpreted as the marginal utility of the attribute they are associated with. For instance, an attribute with a positive coefficient can be interpreted as a desirable attribute that anglers will seek in their choice behavior. Conversely, attributes with negative coefficients are undesirable, and anglers try to avoid alternatives with these attributes. In addition, with the mixed logit we can learn not only whether anglers seek or avoid certain attributes, but also whether there is heterogeneity in preferences. By estimating the standard deviations of parameters, we can learn


33 whether most anglers prefer or avoid a certain attribute to the same degreewhen the standard deviation is not statistically significant or whether that preference or avoidance behavior is stronger in some individuals and weaker in others. In other words, a statistically significant standard deviation of a coefficient for a particular attribute indicates that anglers preferences toward that attribute are heterogeneous across the population of recreational fishers. S hore fishing mode The site attributes in the shore fishing models perform well and the model gives a good representation of angler behavior (Table 23). All the means of the parameters are statistically significant and of the expected sign. Increases in travel costs d ecrease the likelihood that anglers choose a site, which results in a downward sloping demand curve for recreational fishing. Anglers are also more likely to choose the more popular fishing locations than less popular ones. Similarly, anglers prefer locati ons in which their expected catch is high. In terms of seasons, shorebased anglers exhibit a strong preference for fishing during the fall and summer but still prefer the spring over the winter. Areas in the Gulf of Mexico are preferred over areas in the Atlantic Coast for shorebased fishing. The mixed logit results for the shorebased fishing mode show that shorebased anglers exhibit heterogeneous preferences on travel costs and the number of access points ( sites ) in the fishing location. That is, while anglers perceive travel costs as a negative attribute and prefer locations that are closer over those that are far away, some anglers are more averse than others to travel far to engage in recreational fishing. Angler preferences on the number of access points in a fishing location are a bit more complex. The conditional logit results would suggest that anglers prefer locations with


34 fewer access points. The mixed logit, however, tells a more complicated, and perhaps realistic, story. The mean of the parameter is positive and statistically significant, albeit small. However, the large and statistically significant estimates of standard deviation of the sites parameter imply that while some anglers do indeed avoid locations with high numbers of access points other anglers actually prefer locations with many access points. In this case, the attribute is seen as a positive by a group of anglers and a negative by another group. Anglers fishing from shore also avoided alternatives affected by the DWH oil spill, th ereby expressing a preference for fishing trips in locations that were not affected by the oil spill, or at times in which the spill was not affecting the chosen location. However, the standard deviation of the spill parameter shows that not only are preferences for trips affected by the DWH oil spill heterogeneous, but that some shorebased anglers choices were not at all affected by the oil spill. F or hire fishing mode The for hire fishing model also provides an adequate depiction of angler behavior with the expected signs on most parameter coefficients (Table 24). As expected, anglers are averse to high travel costs and prefer locations close by than those far away. Individuals who take guided fishing trips in the Southeast US also prefer locations that are popular fishing destinations and see the Gulf Coast as a superior alternative when compared to the Atlantic Coast. Rather than fishing during the fall, as shorebased anglers prefer to do, individuals using the for hire sector prefer to take their fis hing trips during the summertime, and springtime is preferred over the fall. Wintertime is the least preferred season of the year for guided fishing trips. The for hire models show something that might be considered an anomaly. Rather than preferring locat ions


35 that can be expected to yield high catches, anglers taking guided trips choose locations in which their expected catches are low. This abnormality may be due to lack of information by anglers regarding other locations where guided trips are available or limited availability of for hire operators in areas with high expected catches. It is also likely that anglers learn about high and low catch locations in an iterative manner after repeated trips. However, it is less likely that anglers fishing in guide d trips make enough repeated visits to discover the higher catch locations, especially since this would involve taking trips with different fishing guides. Anglers in the for hire sector also exhibit heterogeneous preferences for travel costs and the number of access points available. As in the shorebased models, some anglers taking guided fishing trips are more averse to pay high travel costs than others. Similarly, some anglers prefer locations with many access points, while other anglers prefer loc ations with few access points. Individuals taking guided fishing trips in the Southeast also perceived the oil spill as a negative attribute and avoided areas affected by the DWH oil spill during the time periods in which oil was present. For hire angler p references toward the oil spill, however, are homogeneous, implying that all anglers using this fishing mode are just as averse to fishing in areas affected by oil spills. P rivate and rental boat fishing mode The discrete choice model of private and rental boat fishing also depicts angler behavior appropriately, and the attributes used perform well explaining recreational fishers choices (Table 25). Anglers fishing from private or rental boats are more averse to travel long distances than shorebased and for hire anglers, most likely due to higher travel costs as a result of these individuals pulling their boats in trailers. Private boat anglers also prefer popular locations as well as locations in which they can expect to


36 catch higher numbers of fish. Sum mer and fall are the preferred fishing seasons, but spring is also seen as a better season for fishing than winter. As with the other fishing modes, areas in the Gulf of Mexico are preferred over those in the Atlantic Coast. Anglers fishing from private and rental boats have heterogeneous preferences for more site attributes than any other fishing mode. Similarly to anglers fishing from shore and using the for hire sector, private boat anglers preferences for travel costs and number of access points are heterogeneous. The degree of aversion to travel long distances varies across anglers, and some anglers using this fishing mode prefer locations with many access points while others prefer locations with fewer access points. In addition, preferences for expec ted catch are also heterogeneous, implying that while for some anglers the expected level of catch plays a pivotal role in the decision of where and when to fish, it plays a lesser role for other anglers. Similarly to anglers fishing from shore, individual s in the private and rental boat fishing mode exhibit a heterogeneous aversion for areas affected by the DWH oil spill. This heterogeneity appears to be greater for private anglers than shorebased ones, which is likely a reflection of greater heterogeneit y in preferences overall in the private boat fishing mode compared to other modes. WTP to avoid spilled oil The estimates of WTP for prevention of the DWH oil spill differ between estimation procedures (Table 26). In the conditional logit models, for inst ance, we find that the mean monetary compensation due to anglers is $125 for anglers in guided trips, $65 for anglers fishing from shore, and $10 for anglers using private and rental boats (Figure 2 4). On the other hand, the estimates of WTP we obtain fro m the mixed logit model are more conservative at $34 in the for hire sector, $28 in the shorebased


37 sector, and $1.60 in the private and rental boats sector (Figure 2 5). The relative magnitude of WTP across modes, however, is quite similar across estimati on procedures, with the compensation for anglers in guided trips being the largest of the three modes and that for anglers using private boats being the smallest. We obtain confidence intervals for WTP using the Krinsky Robb (1986) procedure with 10,000 dr aws. To compute estimates of total damages to marine anglers in the Southeastern United States as a result of the DWH spill (Table 8), we multiply our per trip measures of WTP for prevention of the spill by the total number of marine fishing trips estimate d by MRIP in the States considered in our study (Table 7). The total damages as estimated using the conditional logit model have a 95% confidence interval spanning between $1.3 and $1.6 billion, with a mean of $1.4 billion. Our most conservative estimate o f total damages obtained using the mixed logit model has a 95% confidence interval that spans between $459 million and $625 million and is centered at $542 million. Given that the mixed logit addresses the IIA problem associated with the conditional logit, this conservative estimate of total damages of $542 million is the most credible estimate. Discussion The most important finding of this study is the rejection of the null hypothesis that the DWH oil spill had no impact on marine recreational fishing in t he Southeastern United States. In fact, the 2010 DWH oil spill in the Gulf of Mexico had a negative and statistically significant impact on marine anglers across all fishing modes, and using choice modeling we have found that the monetary compensation due to them is in the order of half a billion dollars. However, while this estimate represents losses in non-


38 market values, it does not reflect the totality of economic losses as a result of the DWH oil spill. Our estimates reflect losses in direct use value o f the Gulfs fishing grounds to marine recreational anglers only. Losses to commercial fishing, the tourism industry and other interests who depend on recreational expenditures are not accounted for in our study; neither are losses to other recreational us ers of the Gulf of Mexico, such as beachgoers, scuba divers, and recreational boaters. Finally, and perhaps more importantly, our study does not measure losses in ecosystem service flows or passive use values as a result of the oil spill. Another aspect that merits clarification is the relative magnitude of our estimates of WTP for prevention of the DWH oil spill. Common sense could suggest that anglers whose fishing expenses are highest can be expected to have the highest WTP for prevention of the DWH oi l spill. However, our compensation measures are not related to fishing related expenses but to behavioral responses to two attributes of fishing trips. In particular, the compensation measure developed here (Equation 211) is inversely related to anglers response to travel costs (consisting of the standard mileage and opportunity cost measures using Census data) and directly related to their response to oil spill impacts. For instance, the response of private boat anglers to travel costs is the largest of the three modes, showing that anglers who haul their own boats are less willing to travel long distances than other anglers who are just traveling with their fishing gear (shorebased) or those who have no need to carry fishing gear or boats (for hire). Si nce the compensation measure is inversely related to the response to travel costs, this makes the compensation measure for private boat anglers smaller than that for other modes, other things held equal. Similarly, private boat anglers response to the oil


39 spill is the weakest of the three modes. A possible explanation is that anglers driving their own vessel have more flexibility to find fishing spots that are oil free, even when the region they are fishing in is experiencing the impacts of the oil spill. This weaker response translates into smaller compensation measures for private boat anglers and explains why the monetary compensation for shorebased anglers is consistently larger than that for private boat anglers. The use of the mixed logit allows us t o find the attributes of fishing trips for which anglers have heterogeneous preferences. Our findings in this respect are similar to those from Haab et al. (2012), who also find preferences for travel costs to be heterogeneous among anglers in the Southeas tern US. In addition, we also find heterogeneity in preferences for site popularity, expected catch, and the number of elemental sites among these same anglers, depending on the fishing mode. Our models may differ from those in Haab et al. in terms of the number of heterogeneous attributes due to our higher level of aggregation and the consideration of different groups of anglers. The aggregation in Haab et al. is county level, a finer scale than our regional aggregation. This may be the reason for the disc repancy in heterogeneity of preferences for the number of elemental sites, as our regions have many more elemental sites than their counties, as well as site popularity, since our sites are essentially different. Similarly, their focal groups are anglers w ho target particular groups of fish species, while ours are anglers who use particular fishing modes. The sample sizes used in each model are much larger in our study, but their study uses a much larger number of models. The catch attribute also plays more important role in the Haab et al. study than in ours, as it is the attribute that is valued. In that sense, the


40 catch attribute serves the same purpose in Haab et al. as the oil spill attribute serves in this study. Another critical attribute for which anglers appear to have heterogeneous preferences is the DWH oil spill itself. That different people would respond differently to an oil spill is not an entirely surprising result or one without a precedent in the literature. Researchers investigating the ef fects of the Prestige oil spill found that Spanish tourists actually increased their visits to the affected coastline during and after the spill (Garza Gil et al. 2006). Loureiro et al. (2006) call this the solidarity effect as individuals choose to visi t the affected coastline as a result of public interest surrounding the oil spill and to experience first hand the magnitude of the disaster. In the case of the DWH oil spill, it is possible that many marine anglers actually took trips to affected areas as a result of free fishing days in which license restrictions were lifted, or just to see the effects of the oil spill personally.


41 Table 21. Model Variables Type Name Description spill Dummy for oil spill effects on intercept zone TC Round trip travel cost to site ($) P n ffdays12 Number of days fished in last year P n med_income Median annual income in respondents county of residence ($) P n mode1 Dummy for 1 st mode (shore fishing) P n mode2 Dummy for 2 nd mode (for hire fishing) P n mode3 Dummy for 3 rd mode (private or rental boat) L j hckr Historic mean number of fish caught and kept by wave and mode a L j popularity Number of intercepts in region L j sites Number of access points in region L j gulf Dummy for Gulf of Mexico E j summer Dummy for May August E j spring Dummy for March April E j fall Dummy for September October T nj C nj Observed catch (all species) per hour fished T nj nj Predicted catch (all species) per hour fished T nj hrsf Hours spent fishing during intercepted trip T nj targ Dummy for targeting any species T nj area1 Dummy for fishing in the ocean, close to shore T nj area2 Dummy for fishing in the deep ocean, offshore T nj area3 Dummy for fishing inland or inshore a Fish kept are those that have been landed and the angler is taking for consumption, not fish caught during the trip and used as bait. Table 22. Negative binomial predicted catch model of fish caught and kept Variable Coefficient Standard Error Z ffdays12 0.0023 0.0003 8.28 med_income 0.0000081 0.0000 6.70 mode1 0.1575 0.0344 4.58 mode2 0.1590 0.0546 2.91 hckr 3.2739 0.1192 27.47 popularity 0.0000007 0.000000 0.35 hrsf 0.2069 0.0074 28.14 targ 0.4374 0.0270 16.19 area1 0.1372 0.0320 4.29 area2 0.5068 0.0468 10.82 constant 2.1457 0.0798 26.89 Model Statistics: Log likelihood 59604.009 LR chi2(10) 4440.2 Pseudo R 2 0.0359 Significant at the 0.01 level.


42 Table 23. Models of shore fishing site choice Conditional Logit Mixed Logit Baseline (2009) Event (2010) Baseline (2009) Event (2010) Variable Coef. Z Coef Z Coef. Z Coef Z Travel Cost 0.0096 102.69 0.0096 105.63 0.0243 47.53 0.0260 48.09 Popularity 0.0269 62.64 0.0229 54.86 0.0423 43.40 0.0357 39.40 Exp. Catch 0.6339 3.00 1.2719 5.80 0.7996 3.45 1.3466 5.73 Spill 0.6306 17.74 0.7227 13.26 Summer 0.6088 31.23 1.0185 50.88 0.6081 30.67 1.0168 50.11 Spring 0.3474 14.24 0.4664 18.93 0.3483 14.27 0.4649 15.44 Fall 0.6830 30.73 1.0391 48.58 0.6760 27.14 1.0494 48.18 Gulf 1.1743 38.20 1.2025 40.68 1.9179 31.00 1.8459 30.77 Sites 0.0704 19.05 0.0116 3.10 0.0469 7.79 0.0391 6.46 SD (Travel Cost) 0.0132 38.89 0.0140 40.12 SD (Popularity) SD (Exp. Catch) SD (Spill) 1.1820 8.97 SD (Sites) 0.1886 18.82 0.1832 17.88 Model Statistics: Log likelihood 52583.05 68906.44 4154.9 4749.74 Pseudo R2 0.3478 0.3891 Alternatives 60 60 60 60 Cases 18,463 21,628 18,463 21,628


43 Table 24. Models of for hire fishing site choice Conditional Logit Mixed Logit Baseline (2009) Event (2010) Baseline (2009) Event (2010) Variable Coef. Z Coef Z Coef. Z Coef Z Travel Cost 0.0036 78.15 0.0034 78.89 0.0077 45.85 0.0073 47.16 Popularity 0.0064 21.14 0.0042 13.99 0.0105 23.48 0.0094 21.21 Exp. Catch 0.3125 19.41 0.4901 27.18 0.2159 12.80 0.3701 19.85 Spill 0.4289 12.43 0.2528 6.90 Summer 1.0014 39.66 1.1944 45.70 0.9955 39.12 1.1257 42.87 Spring 0.6943 22.56 0.8882 30.68 0.6682 17.40 0.8409 26.48 Fall 0.6169 19.40 0.7456 24.71 0.5851 15.19 0.6867 16.57 Gulf 0.7816 30.38 0.8255 33.02 0.9536 27.14 0.9162 27.48 Sites 0.0030 0.99 0.0398 12.63 0.0157 4.10 0.0107 2.61 SD (Travel Cost) 0.0082 40.48 0.0079 39.60 SD (Popularity) SD (Exp. Catch) SD (Spill) SD (Sites) 0.0926 9.65 0.1323 17.32 Model Statistics: Log likelihood 12182.08 15545.85 2506.58 2521.87 Pseudo R2 0.1257 0.1423 Alternatives 60 60 60 60 Cases 11,835 13,338 11,835 13,338


44 Table 25. Models of private and rental boat fishing choice Conditional Logit Mixed Logit Baseline (2009) Event (2010) Baseline (2009) Event (2010) Variable Coef. Z Coef Z Coef. Z Coef Z Travel Cost 0.0137 137.86 0.0140 139.24 0.0382 55.46 0.0384 61.53 Popularity 0.0282 74.67 0.0271 70.65 0.0479 52.76 0.0416 55.72 Exp. Catch 0.6210 9.59 0.7935 12.75 0.2145 2.30 0.3281 3.70 Spill 0.1392 5.67 0.0636 2.01 Summer 0.5501 41.32 0.8295 54.56 0.5518 41.14 0.8411 54.46 Spring 0.2401 14.09 0.4209 23.36 0.2354 13.47 0.4458 24.07 Fall 0.3358 20.40 0.7687 46.74 0.3351 17.38 0.8044 46.73 Gulf 1.6001 63.57 1.5871 62.54 2.7791 44.64 2.4053 48.12 Sites 0.0382 13.05 0.0288 9.13 0.0171 3.29 0.0081 1.49 SD (Travel Cost) 0.0191 47.74 0.0193 50.63 SD (Popularity) 0.0156 17.33 SD (Exp. Catch) 1.2498 5.29 1.5079 7.22 SD (Spill) 1.3786 16.45 SD (Sites) 0.0433 2.65 0.0538 3.13 Model Statistics: Log likelihood 116517 .35 126430 .99 10149 11309.3 Pseudo R2 0.3917 0.4134 Alternatives 60 60 60 60 Cases 36,327 37,346 36,32 7 37,346


45 Table 26 Per trip willingness to pay for prevention of the Deepwater Horizon oil spill Conditional Logit Mixed Logit 95% CI 95% CI Mean Lower Upper Mean Lower Upper Shore $65.57 $57.99 $73.05 $27.76 $23.53 $32.14 For Hire $125.63 $104.54 $146.15 $34.53 $24.22 $44.96 Private $9.95 $6.40 $13.47 $1.64 $0.01 $3.31 Table 27 MRIP participation estimates by state and mode (2010) State Shore For Hire Private Alabama 812,275 33,505 840,377 Florida 10,038,324 579,314 13,866,273 Georgia 335,219 7,359 530,224 Louisiana 728,670 78,886 3,054,931 Mississippi 596,544 6,842 629,207 North Carolina 3,313,215 165,304 2,199,055 South Carolina 1,142,892 76,834 1,078,462 Total 16,967,139 948,044 22,198,529 Table 28 Total loss estimates (in Millions USD) Conditional Logit Mixed Logit Mean 95% CI Mean 95% CI Lower Upper Lower Upper Shore $1,112.50 $983.90 $1,239.50 $471.00 $399.20 $545.30 For Hire $119.10 $99.10 $138.60 $32.70 $22.90 $42.60 Private $220.90 $142.10 $299.00 $36.40 $0.20 $73.40 Total $1,451.36 $1,296.99 $1,601.65 $542.25 $459.93 $625.39


46 Figure 21. Federal fishery closure in the Gulf of Mexico in response to the Deepwater Horizon oil spill, July 13, 2010 Figure 2 2 .Density histograms of important variables


47 Figure 23 Ten coastal regions in the Southeastern U.S. Figure 24 Conditional logit estimates of per trip willingness to pay for prevention of the DWH spill


48 Figure 25 Mixed logit estimates of per trip willingness to pay for prevention of the DWH spill


49 CHAPTER 3 USING PORTFOLIOS TO MANAGE NATURAL RESOURCES Introduction Use of natural resources often requires the balancing of alternatives and groups of alternatives and their associated payoffs. Private users face choices with different costs and benefits, as well as differing levels of risk depending on each alternatives propensity to natural and market variability. For instance, farmers must choose among crops to be planted and fishers must choose which species t o target while considering variation in weather, prices, input costs and other factors. But in a society where public trustees manage use of natural resources, it is these trustees who have the responsibility for such decisions. Therefore, natural resource trustees must make the same choices that individual farmers, fishers, and foresters make on a daily basis, except at a much larger scale that may sometimes include entire landscapes or seascapes. Portfolio selection is a tool that can help natural resourc e managers by systematically weighing returns and risks of different alternatives. The portfolio approach has its roots in the economics of finance. Financial investors must choose from among a pool of assets with varying rates of return. Harry Markowitz (1952) recognized that investors could ameliorate their investment risk by holding a variety of assets whose returns fluctuate in opposite directions. In addition, he proposed the portfolio selection model to systematically choose the combinations of assets that would yield the maximum returns at the lowest possible risk. In recent years, the portfolio selection method has been proposed and illustrated as a tool for managing natural resources so as to yield the highest possible returns at minimum risk.


50 The c oncepts and language from portfolio selection have been permeating the study of natural resources for some time. In the context of fisheries management, for instance, Lauck et al. (1998) liken the use of marine reserves to the practice of portfolio diversi fication as a measure to reduce risk. Similarly, Hanna (1998) argues that the creation of property rights institutions in fisheries that are more amenable to management using ecosystem or species portfolios is necessary to deal with the multiple objectives inherent in fisheries management. Edwards et al. (2004) propose the use of portfolio selection to manage capture fisheries with multiple objectives at the ecosystem level and discuss the institutional changes required to make portfolio management of fisheries feasible. Their proposal rests on the similarities between individual species in an ecosystem and financial assets. In the context of biodiversity conservation, Figge (2004) draws an analogy between market assets and conservation units genes, species or ecosystems and argues that the diversification inherent in portfolio management can be an effective tool for conservation. Ecologists have also recognized a portfolio effect at work in natural systems (Tilman et al. 2006; Schindler et al. 2010), by whic h ecological communities with high diversity tend to produce more stable streams of ecosystem services. Similar to investors, ecosystems appear to benefit from holding a diverse set of species, as the effects of natural variability are partially dampened by diversification. Empirical applications of the portfolio approach for natural resource management in the literature have mostly dealt with issues in fisheries, although some innovative applications in the fields of spatial planning, range management and climate change adaptation have also been proposed. The fisheries related applications include


51 decisionsupport tools for individual fishers, seafood processors and brokers, aquaculture firms, and fisheries managers dealing with single and multiplespecies problems. Baldursson and Magnusson (1997) were the first to implement the approach and the problem they studied was that of fisheries managers regulating the catch of Icelandic cod. Their decisionsupport tool was designed to aid in the optimal targeting o f different ageclasses of cod and their model is explicitly tied to an agestructured population model, where fish of different ages and sizes have different harvest costs and prices. Larkin et al. (2003) take a similar approach and use portfolio selection to examine the problem of diversification of whiting and pollock products in the Pacific U.S. from the perspective of seafood brokers, seafood processors, and fisheries managers, with a fisheries manager portfolio that is tied to a population model for P acific whiting. Similarly, Perruso et al. (2005) examine the optimal targeting portfolios of U.S. based fishers operating in the pelagic long line fleet in the South Atlantic, Caribbean, and Gulf of Mexico, and use their model to study the impact of area c losures on fishers optimal targeting strategies. The problem of multi species or ecosystem based fisheries management was studied by Sanchirico et al. (2008), who developed optimal portfolios of catch for several commercially caught seafood species in the Chesapeake Bay using sustainability constraints. In the same vein, Radulescu et al. (2010) use portfolio selection to develop optimal catch strategies for an aquaculture facility in Romania cultivating several species of fish. Two commonalities in these s tudies are the definition of assets as commercially harvested marine organisms and the measure of returns in monetary units, usually profits or revenues.


52 Outside of fisheries, the first empirical application of portfolio theory to natural resources was Koellner and Schmitzs (2006) study on grassland biodiversity and risk adjusted returns from experimental plots reported in the ecological literature. While their approach did not go as far as selecting optimal portfolios, an innovative aspect of their study is the measurement of returns in nonmonetary units, namely biomass yields. In another inventive study, Crowe and Parker (2008) find optimal seed portfolios for regeneration of white spruce forests in Canada under a variety of climate change scenarios. In the realm of spatial planning, Hills et al. (2009) measure the risks and returns of biomes in different seascapes using qualitative indices to plot the corresponding portfolio frontiers and use their approach to find inter related threats to biomes and measures of biome resiliency within the seascape, although portfolio selection is not used. Most recently, Halpern et al. (2011) develop optimal portfolios of spatially defined commercial and recreational activities in the Channel Islands, California, and use the spatial variability of the portfolio as a measure of social equity in the distribution of returns. The main innovations of the nonfisheries empirical portfolios are the inclusion on nonmonetary measures of return as well as spatial and scenariobase d assets. To help guide the development and implementation of portfolios for natural resource management we briefly review the portfolio selection method and develop a series of essential questions for using this approach. Available data and intent of the decisionmaker in large part dictate how these essential questions are to be answered. This in turn determines the type of portfolio model that results and the kind of management questions that can be analyzed. We also discuss the time series


53 properties o f returns and sustainability constraints, two aspects of natural resource portfolios that are of special importance in fisheries related applications. Finally, two illustrations of the approach dealing with the choice of Total Allowable Catches (TACs) of d ifferent species in marine ecosystems are provided. Specifically, we develop portfolios of catch for the Colombian Pacific Coast and the Gulf Coast of Florida using different measures of return. We use our results to assess the efficiency of the observed p ortfolio of ecosystem wide landings in both locations and discuss the advantages and shortcomings of the portfolio selection approach to natural resource management. Portfolios for Natural Resource Management Portfolio selection can be applied in situations where there are several options available, each with their own observable stream of potential payoffs. For our purposes, we refer to each of these options as an asset, and the payoff from each asset is referred to as the assets return. In such situations, the decisionmaker must choose which assets to hold or invest in. The deciding agent, whom we refer to as the portfolio manager, is presumed to be seeking high returns at low levels of risk. Portfolio managers trading in financial markets can choose fro m among a myriad of assets such as bonds, stocks, derivatives, futures, options and swaps (Cvitanic and Zapatero 2004). For each of these assets, the investor has expectations regarding returns and the variation in those returns. Observations of past retur ns, coupled with other information at the investors disposal, are often useful in the formation of these expectations. The starting point for portfolio selection is then a vector of expected returns from the n available assets at time t denoted ( t ), and a matrix of covariance in returns at time t denoted ( t ). In practice, portfolio developers will rely on time series data of returns in past time periods to obtain the vector of expected returns and the covariance


54 matrix. If the assets series is perceiv ed to be stationary, a measure of central tendency can be used to obtain the expected returns. If the series is nonstationary or if recent time periods are expected to have a stronger influence on the value and covariance of expected returns, portfolio developers can use methods such as the valueat risk, vector autoregression, or conditional heteroskedasticity (Sanchirico et al. 2008) to improve the performance of their selection models. The portfolio manager then chooses a vector of weights for all asset s, denoted c( t ), which dictates how much of the asset is purchased or held. In financial portfolios, the weights dictate what portion of the total investment is allocated to purchasing each asset. The expected returns of the portfolio are given by = ( ) ( ) ( 3 1) which is essentially a weighted average of the returns of all assets included. Similarly, the variance of the portfolio, which is seen as a measure of the inherent risk, is given by = ( ) ( ) ( ) ( 3 2) The efficient, or minimum risk set of portfolios can be found by choosing the c(t) vectors that solve the programming problem Minimize ( ) ( ) ( ) Subject to ( ) ( ) ( ) ( 3 3) where M ( t ) is a minimum returns target that is iteratively changed so as to recover the set of c( t ) vectors that solve the problem for a wide range of expected returns. Additional constraints can be placed on the programming problem to obtain solutions that reflect the objectives and limitations of the portfolio manager. For instance, financial portfolio managers may include constraints on the amount of funds to be invested, as


55 well as nonnegativity constraints on the weights to ensure that no short sales are allowed. There are some important elements in which the application of portfolio selection to natural resources diverges significantly from its use in finance. Natural resource managers may be acting on behalf of society rather than as private individuals, and their motives may not always be profit related. Natural resources are also very different from financial assets, not only in terms of their diversity, but also in terms of how their returns are measured and distributed over time. We can also expect the difference in the managers objectives as well as physical limitations on managers and resources to dictate programming constraints of a very different nature. We can therefore identify four essential questions that deal with 1) the nature and objectives of the portfolio manager, 2) the definition of an asset to be included in the p ortfolio, 3) the way in which returns are measured and expected to be distributed, and 4) the definition of constraints in the programming problem whose solution yields the optimal portfolios. These four questions, whose answers turn out to be inter related, are discussed in turn. Four Questions for Portfolio Development Who is the Portfolio Manager? Financial portfolio managers are generally investors acting on their own behalf or that of some other individual with the objective of gaining as much monetary returns as possible under a tolerable amount of risk. Firms or individual natural resource users such as farmers, fishing vessels, seafood brokers and processors, for instance, will act under similar motivations. In natural resource related decisionmaking, however, portfolio managers can also be acting on behalf of a community or society as a whole. Fisheries and range managers, land and seascape planners, and agencies in charge of


56 planning and preparation for long term environmental changes, are some examples of potential natural resource portfolio managers. The main difference between these social planners and the individuals and firms who also manage natural resources lies in the objectives they pursue and the geographic and temporal scales at which their decisions must be taken and hence the range of effects they may have. Notably, for the social planner or resource trustee the maximization of revenues is only one among a competing set of interests. Other relevant objectives may include biomass yields, social equity, employment levels, adaptation capacity, and ecosystem resilience, among many others. Likewise, the temporal and spatial scale in which natural resource trustees and social planners operate is much wider than that of the individual resource user and the financial portfolio manager. Decisions regarding natural resource use and managem ent can have impacts that span throughout entire ecosystems and regions. In some cases, such as largescale deforestation and use of fossil fuels, the impacts of such decisions can be felt throughout the globe. The consequences of these decisions can also affect multiple generations, and in those cases these effects must be taken into account by the decisionmaker. Natural resource managers, therefore, face optimization of objective functions that can be much more complex than that of the financial portfoli o manager not only in terms of objectives but also in considerations of space and time. What Constitutes an Asset? Once the temporal and spatial scale of the problem under consideration is defined and the management objectives are set out, the next step is to define what the assets of concern will be. Financial assets are generally not physical entities but rather contractual agreements that bind parties to some payoff be it positive or negative


57 under uncertain states of the world. These contractual agreements are tied to seemingly unpredictable events such as commodity prices, corporate profitability, performance of other assets, or in some cases the payoff is guaranteed simply by the future solvency of one of the parties involved. Returnyielding natural r esources are similar perhaps with the exception that natural variability, rather than human events, is the major source of uncertainty in payoffs. Natural resources can be considered assets in the broad sense if their management can in any way yield returns to individuals or society. This includes consumptive uses such as planting crops and harvesting wild animals as well as nonconsumptive uses such as recreation and services related to ecosystem functions. These assets and their respective returns can be defined in a temporal sense if returns are realized at different times throughout the year. For example, if the same asset can yield different returns at different times of the year due to seasonal fluctuations in prices, then two or more assets can be def ined for low and high value periods. Similarly, if what could be considered the same asset yields different returns according to spatial location, then several assets can be defined, one for each relevant location. Also, the assets returns may not be relevant under the present state of the world but rather under a variety of future scenarios. For instance, an asset can be seen as yielding a return not today but at a future time for which a set of likely scenarios has been developed. This is the case of por tfolios that seek optimal adaptation alternatives for environmental change scenarios, where returns are not based on past performance but rather on the assets expected performance under a variety of scenarios where inferences on the relationships between asset performance and environmental conditions can be made.


58 How can we Measure Returns? Once the management objectives and the assets under consideration are defined, measures of return for each asset must be identified. Financial assets such as stocks gen erally yield returns in the form of dividends and valorization. To measure the return of such assets, a financial portfolio manager needs only to consider the costs of acquiring the asset and the potential benefits or earnings that the asset might bring. F inancial assets therefore yield returns that are predominantly monetary in nature, which is fitting given that financial portfolio managers are mostly concerned with maximizing earnings. In the case of corporate stocks future performance of the asset is ge nerally unknown. However, portfolio managers can look at previous performance to aid them in forming expectations of future dividends and valorization of the asset. For instance, financial investors can develop an annual rate of return measure using the co sts of acquiring the stock at the beginning of the year, the value of the stock at the end of the year, and any dividends paid throughout the year, and use the distribution of such a rate of return through time to form an expectation of next years rate of return. If the average rate of return is considered stable or stationary, the financial investor needs only to consider the mean return of assets and their covariance to develop optimal portfolios of stocks. In cases where returns from natural resource management are monetary or can be readily monetized, such as yields from marketable crops, harvests of wild finfish and shellfish or timber, the natural resource portfolio manager can rely on these monetized yields to created measures of returns from each as set. If management costs for each asset are also available, then measures of net benefits similar to those used for financial stocks can be developed. In some cases, however, benefits from natural


59 resource management cannot be easily quantified in terms of money. This typically occurs when markets for the good or service in question do not exist, which may happen for a variety of reasons generally stemming from the public good characteristics of some natural resources. In these cases, it may be better for t he portfolio manager to measure returns from each asset directly in terms of the service provided rather than attempt to monetize the service flows in question. For instance, portfolio managers that seek to maximize yields can measure returns in terms of biomass, while those whose objective is to maximize recreational use can measure returns in terms of visitor days. If observable measures of returns are not available, the portfolio manager can use a series of subjective rules or judgments to place a nonmo netary value on the returns of each asset. For instance, Hills et al. (2009) use a participatory approach to develop categorical measures of returns and risks of different biomes in a seascape. The key aspect of any decision rule to assign values to returns of different assets is that such rules follow the strong monotonicity assumption of consumer preference theory by giving larger values to assets that are perceived to be more valuable and larger risk scores to assets that are deemed more risky. While por tfolio selection may not be as useful in these cases as in those where observable measures of return are used, the development of these measures and their comparison across different assets may in itself be of use to managers and decisionmakers. In cases where observable measures of return exist, the portfolio manager must pay particular attention to the distribution of returns over time. Specifically, the existence of increasing or decreasing trends or seasonality in returns must be examined so as to form more informed expectations of future returns. Natural resource management may


60 be plagued with situations where ecological or humaninduced change may be affecting the distribution of returns through time. For example, many fish stocks have been experienci ng decreasing trends in landings due to resource over exploitation or dynamic changes in the structure of ecosystems (Mullon et al. 2005; Pinsky, et al. 2011). In these cases, using simple averages to create expectations of landings in future years may be misleading. In cases such as this where the distribution of returns is nonstationary, use of statistical methods that allow for correction of measures of expected value and covariance is warranted. What Constraints do DecisionMakers Face? Once management objectives have been set and assets and their respective returns have been defined, the portfolio manager must consider any constraints that should be used in the optimization. Constraints are placed to ensure that solutions remain in the feasible realm and they may include limitations on the amount of resources under the managers disposal, such as labor, time and capital inputs, as well as factors that may limit the sustainability of the system or those that are related to the policy environment under w hich the portfolio manager operates. Financial portfolios are generally constrained by the amount of capital to be invested, and in some instances nonnegativity constraints are used to ensure that no short sales are included in the optimal solutions. In situations where the amount of existing resources or the growth and recruitment of living assets limit the possible levels of harvest, it is appropriate to place constraints to ensure that solutions remain within these limits. Harvest or management costs can also be explicitly included as a portfolio constraint rather than implicitly in the development of measures of return. Alternative policy scenarios, such as area closures


61 or different levels of maximum sustainable yields, can also be set as constraints Finally, the optimization algorithm finds the efficient frontier by including a minimum return (maximum allowable risk) constraint when the procedure is set to find the minimum risk (maximum returns) portfolio. Answers to these four questions are largely responsible for the objectives and scope of the portfolio selection model that can be constructed. Similarly, the answer to one question plays a large role in how the following questions can be answered. For instance, the choice of portfolio manager in gr eat part determines the scope of the analysis and hence the scale at which assets must be defined. Similarly, both the choice of portfolio manager and definition of assets dictate the way in which returns are to be measured. On the same vein, all these fac tors affect any special constraints that should be included in the selection model. To illustrate the possible answers to the four essential questions, Table 31 shows the different applications of portfolio selection to natural resources and the way these questions were answered, and Figure 31 shows a downward directed flowchart of possible answers to these questions Time Series Properties of Returns in Natural Resource Portfolios One aspect of natural resource portfolio development that has received ver y little attention is the time series properties of returns. As seen in Table 31, of all the NRM portfolio studies, only Sanchirico et al. (2008) address this issue. In essence, the problem arises when returns in the past differ significantly from current returns and returns that can be expected in the future. Alternatively, the relationship between assetswhich is captured using the covariancemay have changed through time. In natural resource management the main reason for such structural changes in returns or relationships between assets is environmental or ecological change. In fisheries, for


62 instance, the decline in worldwide capture fisheries is a factor that must be accounted for by portfolio managers. In this section we focus on returns that change through time. To illustrate this problem, we show the historical trends in revenues from Gag grouper ( Mycteroperca microlepis ) in Floridas Gulf Coast from 1986 to 2009 (Figure 32). Even though the time series is too short to make a judgment on the long t erm returns from gag, it is clear that revenues increased after 1995, peaked in 2001, and have experienced a decline since then. The decline of gag is usually attributed to overly high fishing pressure from commercial and recreational fishers coupled with a red tide event in 2005 that resulted in a largescale fish kill. Currently gag are considered overfished in the Gulf of Mexico and a recovery plan has begun to be implemented (Salamone 2012). This combination of anthropogenic and natural pressures has ensured that returns or revenues from gag will remain at low levels at least in the short term. If fisheries managers wanted to include the stock of gag grouper as an asset in a natural resource portfolio, the nave advisor would probably suggest that the mean of the observed returns be used as the expected returns for the stock of gag. This value is represented by the dotted line in Figure 32. The mean is well below the historic peak in returns, a level that is unlikely to be observed in the near future. However, if levels of fishing revenue near the mean were to be observed in the following year it would amount to a fiveyear high. An informed advisor with information regarding the current status of gag would probably point out the difficulties that the s tock is currently undergoing and advocate for a lower expectation of fishing revenue for gag grouper in the following years.


63 The fisheries manager in data constrained ecosystems, however, may have very little or no information regarding the current state o f fish stocks. In such systems, the fisheries manager must make decisions regarding allowable catch and other restrictions without the benefit of ample knowledge on the biology and behavior of the fish in question or the status and health of the overall st ock. If the fisheries manager only had data on historic revenues at her disposal, she could use it to create a more informed expectation about the performance of the stock in the near future using an integrated autoregressive moving average or ARIMA model. Such models take into account seasonality, reaction lags and trends in the time series to formulate predictions of returns in the following periods (Cryer and Chan 2008). ARIMA models differ from each other by the number of parameters used, and the choice of parameters itself depends on the characteristics of the time series. In our analysis, we use the forecast package developed by Hyndman and Khandakar (2008) to formulate expectations regarding future fishing revenues, landings and prices. Given that not all fish stocks are expected to have the same reaction lags or be experiencing similar trends, a unique ARIMA model is required for each time series. Annual time series, such as the ones used in this analysis, can be modeled using a nonseasonal ARIMA ( p d q ) process, which can be represented by ( ) ( 1 ) = + ( ) ( 3 4) where { Yt} represents the observed time series, {t} is a white noise process with mean zero and variance 2, B is the backshift operator, d is the order of differencing, and () and () are polynomials of order p and q respectively. The analyst must choose the parameters p d and q that best explain the variation in the time series. This process


64 generally requires careful examination of the time series, including differenci ng the series d times to obtain a stationary series, and examining the autocorrelogram and partial autocorrelogram of the stationary series to determine the number of significant lags, which determine the choice of parameters q and p respectively (Cryer and Chan 2008). To find the ARIMA model that best describes each time series we use the auto.arima function available in Rs forecast package (Hyndman and Khandakar 2008). This package uses a stepwise algorithm to find the best values of the ARIMA parameter s. The value of d is selected based on successive Kwiatkowski Phillips Schmidt Shin (KPSS) root tests while p and q are found by iteratively using Akaikes Information Criteria (AIC) to find the most parsimonious model. The auto.arima forecast for gag rev enues, shown as the dotted line in Figure 32, is lower than the average level of revenues and is likely to be a better indicator of expected revenues Sustainability Constraints Using Historical Catch Natural resource portfolios, like their financial counterparts, are not inherently bounded by the laws of thermodynamics and may yield solutions that are not feasible in the real world. For instance, the optimal solution may involve harvest rates that exceed the current stock of the natural resource. To prevent such solutions, the portfolio selection problem can include a set of constraints that ensure that thresholds of feasibility are not exceeded. In fisheries applications, for example, there is only a given amount of fish that can be harvested even if fi shers possess perfect technology that allows them to harvest up to the last fish. However, harvesting up to the last unit of fish, or even up to an extinction threshold, will ensure that the stock of fish will cease to exist and all future returns disappear. Fisheries managers and fish harvesters therefore


65 benefit from the existence of some level of maximum harvests that limits the amount of fish taken in any given period and hence ensure the survival of the stock for future time periods. Modern fisheries s cience has developed a variety of stock assessment methods that allow estimation of the size of the fish stock as well as maximum sustainable yield levels of harvest. In many cases, fisheries scientists can determine minimum size limits for harvest that ensure the health of the population and support higher yields. Size limits are especially important in the management of fish and shellfish species with complex life histories, as limiting the size of individuals harvested may ensure that there are enough br eeding individuals in the population to maintain healthy stock levels. The fisheries portfolio manager can take into account the knowledge of fisheries scientists by setting constraints to the portfolio selection problem. For instance, in a portfolio probl em that seeks to find the optimal combinations of harvest from different fish stocks, maximum sustainable yield targets developed by fisheries assessment scientists can be incorporated to the selection problem by constraining harvest rates in the solution set to be below these targets. If this kind of information on the size of the stock and sustainable harvest targets has not been developed, the fisheries manager might instead take a precautionary approach and limit harvests to be below some previously obs erved level that is deemed appropriate. In the analyses that follow we use sustainability constraints similar to those developed by Sanchirico et al. (2008), which ensure that prescribed landings of a given stock do not exceed their historical maximum. The fisheries manager can easily set the threshold level relative to this


66 historical maximum by changing the value of one of the parameters of the optimization problem (). Namely, in each time series we observe a historical maximum level of harvest, measured in tons, which we denote hmax. A vector of maximum harvest levels for all assets, Hmax, is then created. If the selection model is optimizing biomass yields directly, as in the Colombian Pacific Coast example that follows, then the constraint we impose is simply ( ) ( ) (3 5) where expected returns ( t ) are given in tons and dictates the fraction of the historical maximum landings that will be allowed for harvest ( = 1 in the examples that follow). Alternatively, if the selection model is set to optimize monetary returns as in the Florida Gulf Coast example, the creation of the Hmax vector is followed by the creation of a vector of expected prices, Pe, using ARIMA forecasts of observed prices. We can then create a vector of maximum revenues, max = HmaxPe, which is used in the constraint ( ) ( ) (3 6) where expected returns, ( t ), are given in dollars. It is important to note, however, that rather than using observed historical maxim um levels of harvest, the portfolio manager could easily replace Hmax with a vector of threshold levels of landings mandated by policy or recommended by fisheries scientists. Alternatively, the fisheries managers could set < 1, to ensure that only levels below the historical maximum landings are allowed.


67 Portfolios of Total Allowable Catches Using Biomass The Colombian Pacific Coast We first turn our attention to the Colombian Pacific Coast (CPC), a tropical region where multiple species are harvested predominantly by industrial and artisanal fishers (Figure 3 3). The Colombian Pacific coastline spans nearly 1,300 km from the border with Panama in the north to the border with Ecuador in the south and includes two remote islands (Gorgona and Malpelo) and four states or departments: Choco, Valle del Cauca, Cauca, and Nario (Corporacion Colombia Internacional 2006). There is one major port city, Buenaventura, but otherwise the region is remote and road access is virtually nonexistent. This is also one of the poorest regions in the country and government services are quite limited. The remoteness of the region and the lack of government services translate into major challenges for the fishing sector that include high prices of fuel and fishing gear, lack of access to credit, and lack of access to markets for the sale of seafood products (Conde Castro 1991). As can be expected for a remote area with little government services, fisheries data is quite limited. Involvement of the national government is limited to providing fuel subsidies to the industrial fishing fleet and collecting data on landings. There is also a random sampling of landings to measure and identify the different marine species that are caught throughout the area, but some fishing towns are never sampled. The purpose of these sampling efforts is to determine whether juvenile fish are being harvested and if so recommendations are made regarding regulations that could prevent such taking of juvenile fish (Corporacion Colombia Internacional 2006). The CPC data used in this study is a time series of landings by 28 species groups spanning from 1995 to 2008 obtained from the Colombian Institute of Agriculture and Fisheries


68 (Instituto Colombiano Agropecuario ICA), housed within the Colombian Ministry of Agriculture and Rural Development. Data on prices or revenues could not be found in public records, and it is likely that such data is not collected. Given that the data available for the fisheries in this region include only landings in metric tons by species group, we use that as our measure of returns. The problem is solved from the viewpoint of the regional fisheries manager and the objective is to obtain the largest amount of fish biomass, regardless of species. Each species group is defined as an individual asset. In this case, this presents some problems, as some species groups include a variety of marine species (Table 32). For instance, tunas ( atunes ), snappers ( pargos ) and groupers ( meros ) are all grouped as a single entity, preventing managem ent at a finer scale. A sustainability or maximum catch constraint is set to ensure that our solutions do not dictate landings greater than the historic observed maximum CPC Results The optimization in equation (33) yields the efficient frontier that represents the set of portfolios of catch that result in the highest level of biomass harvested with the lowest possible expected variability. In other words, the efficient frontier shows the minimum level of risk or variability in harvests attainable at diff erent levels of total harvest. All points to the right of the frontier are in the feasible region but are inefficient, as returns can be increased without increasing risks. All points to the left of the frontier are unattainable. The frontier for the CPC i s shown in Figure 34, where circles represent individual species groups and the cross represents the current portfolio of catch.


69 There are a few features of this plot that deserve special attention. First, note that at the low risk, low returns end of the plot (left hand side) we see that fisheries managers can eliminate risk by closing the fishery, a solution in which risk and return are virtually zero. Increasing harvests from that point requires managers to cope with higher levels of risk. Second, the expected returns and risk from individual assets are included in the plot as circles, and all lie to the right of the frontier in the low return areas of the feasible region. Third, the current portfolio of catches, which is included in the plot as a small cross, also lies to the right of the frontier. This implies that the current combination of harvest rates for all species groups in the CPC is inefficient, and that reorganization could lead to higher returns (more landings) with no increases in allowable risk being necessary. We calculate the extent of harvest inefficiency by comparing the actual portfolio to efficient portfolios that lie directly to the left (E 1) or above (E 2) the actual portfolio (Table 33). Results indicate that the ecosystem wide fi shery could benefit by reducing the fisherys risk or variability of landings by 46% without significant effect on expected landings if some reallocation of species targeted effort occurred. Similarly, expected fishery returns or total biomass extractedco uld be increased by 26% without significant increase in the level of risk allowed. Along with the efficient frontier, which simply summarizes the returns and variance of the minimum variance set of portfolio returns, we also obtain the matrix of efficient harvest policies corresponding to each point in the frontier. That is, for each point in the efficient frontier, there is a unique set of prescribed policies that dictate the harvest level that must be sought from each species group. Since in this case ret urns


70 are given directly by biomass or landings, this matrix identifies the efficient set of prescribed catch policies. For visualization we use a transition plot that shows the prescribed levels of catch for each species group at different levels of allowable risk (Figure 3 5) Portfolios of Total Allowable Catches Using Revenues The Florida Gulf Coast We now consider the case of the Florida Gulf Coast (FGC), which spans north and west to the border with Alabama near Pensacola, and south and east to the sou thwest tip of Everglades National Park near Lake Ingraham. There are several major fishing ports in this region including Apalachicola, Port St. Joe, Steinhatchee, and Fort Myers, among many others. We exclude the Florida Keys from this analysis. Besides some remote areas of Everglades National Park, the entire coastline is well connected to the rest of the country by a modern system of roads and highways and several major airports are located very close to the coastline. Major cities in this region include Pensacola, Panama City, Tampa, Saint Petersburg, Sarasota, Fort Myers and Naples. Fisheries in state waters are managed by the Florida Fish and Wildlife Conservation Commission, while those in federal waters are managed by the Gulf of Mexico Fishery Manag ement Council. In addition, several offices of the National Marine Fisheries Service (NMFS) are located in Florida, including the Southeast Regional Office in Saint Petersburg and the Southeast Fisheries Center in Miami, both of which are heavily engaged i n research and management of these fisheries and regularly conduct stock assessments. In many ways, the FGC with its modern infrastructure, scientific support for management, and stateof the art fisheries data collection programs, is the antithesis to


71 the remote Colombian Pacific Coast. Information on the health of recreationally and commercially important fish stocks is routinely collected and fishing regulations and programs such as catch and size limits, transferrable quotas, and fishing seasons are dev eloped and implemented in a transparent fashion through the Gulf of Mexico Fishery Management Council. In terms of portfolio development, information on returns in terms of biomass and fishing revenues is readily available on NMFS website. In particular, time series beginning in 1986 are available for 69 species groups landed by commercial vessels in the FGC (Table 34 ). Previous to 1986 several species were identified only to the family taxonomic level, and data on returns for many important species cannot be recovered. We therefore use time series data on fishing revenues and landings between 1986 and 2009 and exclude all previous years. We extract information on average annual prices by dividing fishing revenues by the corresponding level of landings, and obtain vectors of average annual prices in dollars per metric ton. Both prices and revenues in the time series are normalized to 2009 dollars using the Producer Price Index. Once again we set up the portfolio selection problem from the fisheries manager s perspective and use individual species or species groups as available in the dataset as our assets. In this case, however, we measure returns in terms of fishing revenues rather than landings, as we had done for the CPC. Thus, the objective is to maximize ecosystem wide fishing revenues while minimizing ecosystem wide risk by choosing the optimal set of prescribed catch levels for all species groups. Finally, we set a sustainability constraint to prevent landings of all species to exceed their observed h istoric maximum


72 FGC Results Solving the selection problem yields the efficient frontier in Figure 36, where circles represent the expected fishing revenues and risk from individual species and the cross represents the current portfolio of landings. In this case, the efficient frontier represents the set of portfolios of catch that result in the highest level of fishing revenues with the lowest variability in revenues. Conversely, the efficient frontier shows the minimum level of risk or variability in fis hing revenues attainable at different levels of total harvest. Similar to the CPC, the current portfolio of catches in the FGC is inefficient as it lies to the right of the efficiency frontier. Through a reorganization of targeting effort (Table 35), fish eries managers could achieve a 16% increase in fishing revenues without the need to increase the allowable level of risk. Fisheries managers could also attain the same level of revenues as under the current portfolio of catches with a 44% reduction in the level of allowable risk through reallocation of targeting effort. Along with the efficient frontier, solving the portfolio selection problem yields the matrix of expected revenues from the minimum risk set of catch policies. This information can be visualized in the transition plot shown in Figure 3 7, which shows the efficient level of expected fishing revenues from each species group in the selection model at each level of allowable risk. The minimum variance set of expected revenues can be used to obtain the set of prescribed catch policies, which determine the harvest level that should be sought from each species at different levels of allowable risk. To do so, we divide the prescribed fishing revenues in the minimum variance set by the expected price of each species group. As previously mentioned, expected prices are


73 estimated using an ARIMA model. This information, visualized in the transition plot in Figure 38, shows the prescribed level of landings for different levels of allowable risk Discussion Portfolio selection, in its most basic form, is a flexible tool that can be used for natural resource decisionmaking. The approach can be used in a variety of natural resource decisionmaking contexts, and the essential questions presented here should se rve as a guide to future natural resource portfolio developers. While the selection approach rests heavily on historic measures of returns from different natural resources, information related to carrying capacity, sustainability, and biophysical limits obtained by scientists studying the resources can be systematically integrated into selection models through the use of optimization constraints. Similarly, policy or regulatory constraints can also be taken into account. Some of the approaches to Ecosystem Based Management currently under development rest on the creation of highly complex models of ecosystem dynamics that include nutrient cycles, ocean currents and food web interactions in a temporally and spatially explicit manner (Plagyani 2007). Compared to many of those models, portfolio selection may seem an overly simple approach and hence one with limited practical value. However, some of the results from portfolio selection are quite unique and obtaining similar information from more complex models of ecosystem dynamics may be quite difficult. For instance, the portfolio selection approach as illustrated here allows the natural resource manager to quantify in terms of the ecosystem returns being used the costs of inefficient decisionmaking. Specifical ly, the distance between the current portfolio of decisions, which is shown as the cross in Figures 34 and 3 6 for the CPC


74 and FGC respectively, and the efficient frontier can be easily calculated. In the case of the CPC and the FGC, the inefficiencies found range between 16% and 26% in terms of revenue and around 45% in terms of risk. If there were significant costs to reorganization of catch, for instance, such measures could be used on the benefits side of a benefit cost analysis. Selection models like the ones developed here could be used to aid in fisheries policy making as a tool complementary to stock assessments and reference point rules. For instance, managers could substitute the sustainability constraints we use with science based rules such as m aximum allowable harvests developed with stock assessment methods or reference points. These policies would replace the sustainability constraint thereby making all prescribed policies more precautionary than mandated by maximum allowable harvest policies. The implication is that resource users may actually gain from obtaining harvests that are lower than those mandated by science based policies, at least when judged from a risk return framework. There is also important information conveyed in the matrix of prescribed landings corresponding to the minimum risk set of portfolios, which we visualize as transition plots. A case in point is the prescription for tuna landings in the Colombian Pacific Coast. For some time it has been argued that this region is in a uniquely favorable position to become a significant source of different species of tuna. While the tuna fishery has been quite important in the region for some time, the increased incidence of El Nio Southern Oscillation (ENSO) events, which heavily aff ect salinity and water temperatures in this region, have changed the migratory patterns of several species of tunas in a way that makes it more likely for them to be present in this region (Conde


75 Castro 1991). While these oceanographic dynamics are not explicitly included in the portfolio selection model, the approach does prescribe an increase in tuna landings, and tuna is a major player in all the minimum risk portfolios. Another interesting facet of the selection approach to natural resources is the rel ationship between diversity and returns. Figge (2004), Koellner and Schmitz (2006), and Schindler et al. (2010) among others have argued that there is a relationship between diverse portfolios and stability of returns. We also see this effect in our portfolios for fisheries management, as the high return portfolios are highly diversified. In fact, it seems that high returns from fisheries are only possible through a diverse composition of landings. However, the low return, low risk portfolios are usually dominated by few species, and other species only enter the minimum risk set of portfolios as sustainability constraints for the dominant species become binding. This is indicated in the transition plots by areas that stop increasing in size and become flat. This feature of our models is quite different from what is commonly seen in financial portfolios. The low risk, low return efficient financial portfolios, for instance, are generally characterized by a varied combination of assets. In particular, a large n umber of assets with negative covariances are present in this section of the frontier, as the negative covariances attenuate risk and allow for very low risk portfolios. The highreturn, high risk portfolios, on the other hand, are dominated by a few highrisk, high return assets, and portfolios in this section of the frontier are not very diverse at all. The large discrepancy between the diversity of low return and highreturn portfolios in financial and natural resource applications may be due to the norm alization of financial returns, which are generally calculated on a per dollar invested basis. Such


76 normalization is not required in natural resource portfolios, and in fact may be difficult to achieve, as the meaning of investment in a natural resource portfolio is quite different from that in a financial portfolio. This is an important issue and merits further research. There remain some important limitations to the portfolio selection approach to natural resource management. In situations where there are competing alternatives with inherently different measures of return, use of the portfolio approach can be complicated by an equivalency problem. For instance, inclusion of the recreational fishing sector in the Florida Gulf Coast into a portfolio select ion model like the one presented here is complicated by the nonmarket nature of recreational benefits. While a selection model could easily be constructed to analyze the risk return tradeoffs between different recreational activities, one to analyze the t radeoffs between commercial and recreational activities would be much more difficult to build. In essence, it is difficult to compare alternatives whose returns are measured in different units. Construction of equivalency measures between activities that are inherently different is likely to open up new areas of application for the portfolio approach to natural resource management


77 Table 31. Empirical portfolios in the natural resource management literature Study Manager Assets Space/Time considerations Measure of returns Returns stationary Constraints Baldursson and Magnusson (1997) Fisheries manager Cod age cohorts N/A Profits Yes Fish growth and recruitment; fishing costs Larkin et al. (2003) Seafood broker Fish Processor Fisheries Manager Products made from Pacific Whiting and Pollock Monthly variability Profits Yes Fish population dynamics Perruso et al. (2005) Representative vessel Commercial fish species N/A Profits Yes Trip limits per species Koellner and Schmitz (2006) Range Manager Plant species in experimental grasslands N/A Biomass; Yield in monetary units N/A None Sanchirico et al. (2008) Fisheries Manager Commercial fish species N/A Revenues No; Value at Risk Method for covariances Sustainable fishing levels Crowe and Parker (2008) Conservation Planner Tree seed sources Adaptation Regions Climate Change Scenarios Adaptation capacity Yes None Hills et al. (2009) Coastal Planner Biomes in the seascape Spatially defined biomes Subjective value and risk scale N/A None Radulescu et al. (2010) Fish Farm Owner Species in a fish farm N/A Profits Yes Max. Sust. Yield Halpern et al. (2011) Coastal Planner Commercial fisheries Spatial planning units Revenues Trips Yes Total effort Spatial closures Recreational activities


78 Table 32. Colombian Pacific Coast species groups, maximum landings in sustainability constraints, and year in which maximum landings were observed Group Max. Landings (tons) Year Atunes 78025 1997 Berrugate 590 2000 Sardina 28747 1997 Bagres 488 2000 Cojinua 568 2000 Lenguados 171 2001 Lisa 43 2004 Meros 295 2000 Mojarras 101 2004 Pargos 802 2000 Pelada 401 1996 Plumuda 2708.45 2001 Robalo 341 1996 Corvina 347 2006 Sierra 912 1998 Tiburn 2625 2002 O trospeces 4686 1997 L angosta 9 2001 C amarontigre 2829.65 1997 Cangrejos 865.69 2005 Coliflor 688 2003 camaronblanco 2686 1999 O trocamaron 910 2004 Jaiba 13 2008 Almejas 10 2000 Calamar 295 1996 Caracol 29 1997 O tromolusco 864 1997 Table 3 3. Measure of inefficiency in the Colombian Pacific Coast Returns a Risk b Actual 62.92 275.75 E 1 62.21 148.08 E 2 79.42 273.99 Gross Gain 16.50 127.67 Percent Gain 26.2% 46.3% a Returns are measured in thousand tons b Risk is measured in millions of tons


79 Table 34 Florida Gulf Coast species groups, maximum landings in sustainability constraints, and year in which maximum landings were observed Group Max. Landings (tons) Year Group Max. Landings (tons) Year Ballyhoo 468.7 1995 Octopus 47 1996 Bluefish 321.2 1988 Eastoyster 1669.9 1987 Catfishes 57.1 1987 Permit 79.7 1991 Cobia 81.2 1996 Flpompano 253 1990 Bluecrab 5792.9 1998 Bluerunner 817.9 1988 Softbluecrab 72.5 2000 Scads 1411.2 1993 Stonecrab 3153.1 1998 Scamp 138.2 1993 Atlcroaker 60 1986 Porgies 299.1 1993 Dolphinfish 480.9 1991 Blkseabass 234.7 1994 Blkdrum 262.7 1986 Sandtrout 83.7 1986 Finfishmeal 3350.3 1992 Spottrout 612.2 1988 Finfishfood 399.7 1987 Sharks 2553.4 1990 Flatfish 106.1 1991 Sheepshead 291.2 1994 Gag 1488.1 2001 Shellfish 756.6 1999 Blkgrouper 616.8 1989 Brownshrimp 1070.8 2006 Redgrouper 4001.4 1989 Pinkshrimp 10774.7 1996 Snowygrouper 119.7 2003 Rockshrimp 2079.1 1992 Warsawgrouper 43.4 1986 Whtshrimp 971.3 2005 Yedgegrouper 450.3 2000 Graysnapper 312.6 1987 Yfingrouper 185.5 1986 Lanesnapper 61.4 1991 Grunts 386.6 1994 Muttonsnapper 164.5 1987 Atlherring 3382.6 1989 Redsnapper 469.2 1986 Hogfish 53 1993 Silksnapper 90.4 1992 Crevjack 1748.3 1990 Vermsnapper 1123.8 2009 Kingwhiting 33.1 1990 Ytailsnapper 994.6 1993 Ladyfish 2629.7 1990 Snappers 171.5 1987 Leatherjackets 169.9 1993 Spot 149.8 1990 Spinylobster 3264.1 1996 Squids 66.2 1998 Slipperlobster 24.9 1994 Swordfish 371.1 1996 Kingmackerel 947.8 1993 Goldtilefish 226 1988 Spmackerel 1616.7 1992 Bigeyetuna 18.1 1989 Menhaden 8613 1987 Blkfintuna 29.1 1990 Mojarras 229.3 1993 Yfintuna 1755.1 1986 Strmullet 10668.3 1989 Wahoo 29 1997 Whtmullet 326 1989


80 Table 35. Measure of inefficiency in the Florida Gulf Coast Returns a Risk b Actual 104.56 198.89 E 1 102.69 110.63 E 2 121.34 198.40 Gross Gain 16.78 88.26 Percent Gain 16% 44.3% a Returns are measured in millions US dollars b Risk is measured in trillions US dollars Figure 31. Possible answers to the four essential questions for portfolio development


81 Figure 32. Historical returns from Gag grouper (Mycteroperca microlepis) in the Florida Gulf Coast. The dashed line represents the mean returns, while the dotted line represents the ARIMA forecast of returns for the following year.


82 Figure 33. Map of Colombia. The Pacific Coast spans south to the border with Ecuador and North to the border with Panama.


83 Figure 34. Efficient frontier of biomass harvest from 28 groups of marine species i n the Colombian Pacific Coast. Circles in the plot represent the return and risk of individual species groups. The cross in the plot represents the current portfolio or combination of harvest rates.


84 Figure 35. Transiti on map of prescribed harvest policies in the CPC. The curve corresponds with the efficient frontier and the harvest policies for the individual species groups can be traced for each level of allowable risk.


85 Figure 36. Efficient frontier of biomass har vest from 69 groups of marine spec ies in the Florida Gulf Coast. Circles in the plot represent the return and risk of individual species groups. The cross in the plot represents the current portfolio or combination of harvest rates.


86 Figure 37. Transition map of fishing revenues from prescribed harvest policies in the Florida Gulf Coast. The curve corresponds with the efficient frontier and the expected revenues for the individual species groups when following the prescribed catch policies can be traced for each level of allowable risk.


87 Figure 38. Transition map of prescribed harvest polic ies in the Florida Gulf Coast. The curve is obtained by dividing the expected revenues from each species (Figure 3 7) by the expected price of each species. The individual species groups can be traced for each level of allowable risk.


88 CHAPTER 4 MODELING IMPACTS OF HUMAN ACTIVITIES ON MARINE ECOSYSTEMS: A STYLIZED APPROACH USING ECOPATH WITH ECOSIM Introduction The Ecopath software, as initially conceived by Polovina (1984), was intended as a modeling routine that allowed the estimation of mean annual biomass, production, and consumption for components of an ecosystem. This was accomplished by creating a biomass budget box model in which steady state equilibrium solves simultaneously for the unknown parameters in the system. The main feature was a top down approach to modeling the ecosystem. Rather than starting from primary production and working its way up to the apex predators, the modeling framework began with the apex predators and other species groups at high trophic levels that were known to exist in the system, and then estimated the amount of primary production required to sustain those populations. Polovinas application to the tropical reef system of the French Frigate Shoals is also notable in that no fishing mortality is included. Therefore, no economic analysis was possible in this early version of Ecopath. Several advancements in theoretical ecology were included in Christensen and Pauly s (1992) improved version of Ecopath. Notably, measures of ecosystem maturity proposed by Odum (1969) were included to provide some insights into the stability of the ecosystem. Routines to calculate trophic levels, ascendency, trophic aggregation, and a v ariety of other ecological indices were also included. This version of Ecopath also included the first hints of economic analysis. A routine to calculate direct and indirect impacts of predation, similar to the routines in IMPLAN to calculate direct and in direct economic impacts, was included. Hannons (1973) interpretation of Leontiefs Input Output framework to model direct and indirect impacts of predation in an


89 ecosystem provided the basis for this development. Furthermore, the application to the Schlei Fjord ecosystem included mortality due to fishing, but no reference was made as to the specifics of the fishing fleet. The version of the software freely available today ( ), Ecopath with Ecosim (EwE), includes several improvements and devel opments that have been added in the nearly 30 years since Polovinas pioneering work. The static, steady state equilibrium framework of the early Ecopath versions has now been replaced by a temporary mass balance assumption. A dynamic modeling capability k nown as Ecosim allows exploration of the impacts of fishing and environmental disturbances, as well as optimal fishing policies. Another capability known as Ecospace allows the dynamic models created with Ecosim to be replicated over a spatial grid (Christ ensen and Walters 2004). The objective of this Chapter is to explore the capabilities of EwE to perform economic analysis. To do so, a simple hypothetical model of an ecosystem with multiple species groups and multiple harvesting fleets (fisheries) is developed, and the steps required in building such a model are outlined. Special emphasis is given on the economic or financial data requirements for building the model. This model is then used to explore the different routines included in EwE that explici tly deal with the relationship between the fishing fleets and the ecosystem Ecopath: A Simple Model of a Tropical Ecosystem The model used in this exploration is a modified version of that pr esented in Christensen (2009). As such, the model is meant to re flect historical conditions in the North West Shelf (NWS) of Australia as described by Sainsbury et al. (1993). A flow


90 diagram of the modified model of the ecosystem, which due to modifications should be considered a hypotheti cal ecosystem, is presented in F igure 4 1 Mass Balance The hypothetical ecosystem is composed of nine species groups: seabirds, piscivores, squid, small fish, benthos, zooplankton, phytoplankton, macroalgae, and detritus. One of the important features of the ecosystem is the habitat interaction that exists between macroalgae and small fish. Small fish find refuge in macroalgae, and so their predation mortality is lower when macroalgae density is high, and viceversa (Christensen, 2009). The basic input parameters and diet compositi on matrix are shown in Tables 41 and 42. There are three commercial fishing fleets operating in the hypothetical ecosystem. Trawls are the largest fleet as determined by the volume of landings. They target piscivores, which can be sold at a premium price. However, their fishing gear incidentally removes a considerable volume of macroalgae from the ecosystem. This incidental removal is discarded inside the system. On the other hand, fishers using traps also target piscivores, but do so with a minimal impact on the ecosystem, as there are no incidental catches. Finally, jiggers target squid. However, their gear is notoriously detrimental to seabirds, which get caught in the fishing gear and drown while attempting to eat the squid. This incidental bycatch is also discarded inside the system. The landings and discards figures are shown in Tables 43 and 44. Applying a forcing function to account for the relationship between small fish and macro algae as shown by Christensen (2009) concludes the first stage of modeling the hypothetical ecosystem in EwE. Since seabirds also predate on small fish, the forcing function must be applied to the predation mortality that they exert on small fish as well


91 as that from piscivores and squid. We now have a mass balanced model of the hypothetical ecosystem Economic Input to Mass Balance Stage As discussed in the previous section, only landings and discards information for each fishing fleet are required to obtain a mass balanced model of the ecosystem. Other information abou t the fishing fleets is not required to study the ecosystem, but must be added if the user intends to use EwEs capa bilities for economic analysis. Inclusion of this kind of information will enrich the outputs obtained from the mass balancing routine, as w ell as open the possibility of exploring fisheries policy scenarios in Ecosims dynamic simulations. The types of data that can be used as an input, and the outputs obtained when using this data are described below Financial information on cost structure Different fishing fleets can be expected to have different cost structures depending primarily on the species they target and the gear they use. Furthermore, some fleets may be more profitable than ot hers for a variety of reasons. In theory, the cost stru cture of a fishing fleet will dictate how fishing effort will respond to changes in relative prices of inputs and outputs, as well as how the fleet may respond to management regulations. EwE allows the user to input financial information about the cost str ucture and profitability of each fleet. In particular, the percentage of total revenue (total value of the catch) that is required to cover fixed costs, effort related variable costs, sailing related variable costs, and profits in a given year (or the unit of time defined in Ecopath) can be added as an input ( Appendix A). The figures are meant to portray the average or representative vessel in the fleet, and each of the cost and profit categories is entered separately (Christensen et al. 2005). The fina ncial


92 information on cost structure for the three fleets in the hypothetical ecosystem is shown in T able 4 5 Market and nonmarket prices Prices are a critical set of information in most economic activities. The expected market price at which a vessel can sell its catch is a major determinant of fishing decisions. Market prices can fluctuate seasonally, and some fleets may be able to sell their catch at premium prices due to this seasonality. Also, some fleets may operate with highly selective gear that al lows them to catch only the largest fish, and may therefore receive a price premium for this selectivity. EwE allows the user to enter a price per unit weight for each species caught by each fleet, allowing different fleets to sell their catch at different prices. This may happen if one gear is better at targeting larger fish (or fish of particular sizes) or if the product can be transported to market fresh rather than frozen, for instance. The price matrix allows the user to enter prices for all species, b ut only those species that are also represented in the landings matrix of each fleet are taken into account in EwEs calculations. The market price matrix for the hypothetical ecosystem is presented in Table 46. While the most obvious value of an ecosys tem is the value of the catch extracted from it (i.e., the total revenue of fishing fleets), an ecosystem and its components may also be valuable in situ for nonextractive uses or nonuses. Some authors have pointed out that one of the key concepts of Ecosystem Based Management is the complete valuation of all components of the ecosystem, including benefits arising from the existence of nontarget species (McLeod et al. 2005). The nonmarket price matrix in EwE is meant to represent the value of a resource in the ecosystem for nonexploitative uses (Christensen et al. 2005). In the case of the


93 hypothetical ecosystem, we have given seabirds a nonmarket value, as we assume that people enjoy watching birds, and some people feel distress by knowing that seabir ds die during the operations of jigger vessels. Therefore it is assumed that society receives a benefit when seabirds are conserved, and conversely is negatively affected when seabirds perish due to fishing activities (Appendix B). The nonmarket value mat rix for the hypothetical ecosystem is presented in Table 47. Economic Output of Mass Balancing Once the economic figures are included in Ecopath, the mass balance procedure calculates the total revenue, cost and profit of each fleet, as well as the value of each species. Let Lij be the landings of species i ( i = 1, 2,, n ) by fleet j ( j = 1, 2,, m), and Pij be the price of species i when caught by fleet j The catch value of species i by fleet j is: = ( 4 1) Similarly, the total revenue of fleet j is given by: = ( 4 2) Now let the percentages from the financial information on the cost structure of each fleet be expressed as proportions, where PFCj is the proportion of fixed cost, PECj the p roportion of effort related costs, and PSCj the proportion of sailing related costs, all for fleet j The total cost of the fleet can then be expressed as: = + + ( 4 3) The total profits of the fleet can similarly be expr essed by: = 1 + + ( 4 4)


94 Once more, let the non market price of species i be represented by NMPi, and the biomass of species i be represented by Bi. The nonmarket value of species i in the ecosystem can be expressed as = ( 4 5) The total value of species i is given by = + ( 4 6) And finally, the total economic value of the ecosystem can be expressed as = + = ( 4 7) The mass balance economic output table for the hypothetical ecosystem is shown as T able 48. Ecosim: Time Dynamic Simulation and Policy Optimization As a software that deals primarily with ecological dynamics, EwEs capacity to conduct economic analysis may not pass the neoclassical economics litmus test (see Holland et al.s (2010) discussion on what does and does not constitute economic analysis). However, the features available in EwE may prove interesting and useful if one defines fisheries economics very broadly as t he study of the interactions between coastal and marine ecosystems and the human beings who use and enjoy them. The Ecosim features that deal with this relationship are outlined in this section. Fishing Effort Forcing Function One of the questions concerni ng this relationship between humans and ecosystems that may be of interest to fisheries managers is the effects in the ecosystem of different fishing rates or levels of fishing eff ort from the different fleets. The first feature available in Ecosim is a sc reen that allows the user to create a rel ative effort


95 forcing function. As a base (effort = 1), Ecosim uses the fishing mortalities established by the user in the mass balance stage. In our hypothetical ecosystem, for example, a relative effort value of 1 for the trawl fleet removes 0.7 tons per square kilometer of piscivores in one year as landings, as well as 1 and 3 tons per square kilometer of benthos and macro algae a year as incidental catch, respectively. A doubling in relative effort would double these f ishing attributed mortalities. In essence, the user is allowed to draw a time series of fishing effort for each fleet, taking the mass balance fishing mortali ties from Ecopath as the base. Alternatively, the user can specify a time series of relativ e fishing effort targeted at particular species groups. One can then Run Ecosim, and a time series of abundance trends for each species in the ecosystem is calculated a nd plotted in the same screen. Furthermore, the user can ask Ecosim to produce a serie s of plots with predicted time series of a variety of ecological indicators by each species group. These plots can be exported as a csv file. Figure 4 2 shows the expected biomass trends of the economically important species groups in the hypothetical ecos ystem, if fishing effort of all fleets were to remain at its base value for 20 years. Bionomic Dynamics: Fleets as Dynamic Predators An alternative to sketching the relative fishing effort trends is to allow Ecosim to simulate these trends in an open access scenario. To do this, the user must check the fleet/effort dynamics box in the main Ecosim screen. This option simulates fishing effort dynamics using two time scales for fisher response. In the short term, fishing effort can respond to potential income from fishing, but only to a point constrained by the current size of the fleet. In the long term, the investment and depreciation d ecisions for increasing or decreasing the fleet size (capital capacity to fish) are modeled in a way


96 analogous to the growth of a biological population. That is, the fleet responds to profits (surplus) by growing in size, and to losses (deficits) by reduci ng the size of the population. These responses are represented by two state variables for each fleet j Ejt is the current amount of active, searching gear (scaled to 1 as the Ecopath base), and Kjt is the fleet effort capacity, such that Ejt Kjt. At eac h time period t mean income per unit effort is calculated as = ( 4 8) Where qij represents the catchability coefficient for fleet j going after species group i A mean profit rate for each fleet is also calculated as = ( ) ( 4 9) For each time step, the short term effort response for the following time period is predicted as a sigmoid function of income per unit effort and current fleet capacity: = ( 4 10) where IHj represents the income level required for half of fleet j to be deployed or enticed to become active (i.e. the income level where Ejt = 0.5 Kjt), and represents a heterogeneity parameter for fishers: high values of imply that fishers see income opportunity similarly, while low levels imply that fishers deploy their effort at a wide range of income levels (Christensen and Walters, 2004; Christensen et al. 2005). The parameters IHj and can be set by the user and are unique to each fleet, but iterative simulations with the model of the hypothetical ecosystem suggest that these changes are small, and that the system tends to settle at a unique open access or bionomic equilibrium, regardless of changes in the price of fish, profit margins, and


97 values of the bionomic parameters described above. For comparison, the biomass trends of the economically valuable species in the hypot hetical ecosystem are shown in F igure 43. Notable in the bionomic simulations for the hypothetical ecosystem is the growth of the jigger fishery, which ends up driving seabirds to extinction. Recall that the jiggers are assumed to remove seabirds as incidental catch, and the jigger fleet is predicted to grow dramatically. In an effort to manipulat e the outcome of the simulation, the price of squid (the target species of the jigger fishery) was reduced dramatically, and the profit margin of the fishery was eliminated. This manipulation did not change the outcome of the simulation. Manipulation of the bionomic parameters IHj and changes the time series of effort and biomass, but the end point appears to be the same. These results defy economic theory and basic intuition. The bionomic dynamics tool appears to be an interesting feature of EwE. However the conclusions obtained from simple manipulation of the input parameters suggest that this tool can be improved dramatically with some basic economic intuition. Fishing Policy Search Invoking the fishing policy search option in Ecosim opens a new window that allows entering the number of jobs per catch value created by each fishing fleet. While it is possible that some government agencies or fisher groups maintain detailed information on the number of jobs created by each fishing fleet, it is more common to encounter employment statistics for the commercial fishing sector as a whole. The number of jobs created by each fleet can therefore be calculated using the catch value ( CVij) statistic created by the Ecopath mass balance procedure discussed previously.


98 Recall that the total revenue obtained by each fleet can be obtained by summing over the catch values of the portions caught by each fleet. The total market value of the ecosystem can be calculated as = ( 4 11) and the proport ion of total value that can be attributed to each fleet is therefore = ( 4 12) If the total number of jobs created by the commercial fishing sector in the economy is TJ then the jobs per catch value by each fleet can be calculated as = ( 4 13) The jobs by fleet for the hypothetical ecosystem, assuming that commercial fishing creates a total of 300 jobs in this hypothetical economy, is shown in T able 49. To find an optimal fishing policy, the user is allowed to define an objective function to be optimized. This objective can be a combination of net economic value (profits), employment, mandated rebuilding of target species, ecological stability and other index measures. The fishing policy search option in Ecosim uses t he DavidsonFletcher Powell (DFP) nonlinear optimization procedure to iteratively improve an objective function by changing relative fishing rates. The parameter variation scheme used by DFP is known as a conjugate gradient method, where alternative par ameter values are tested so as to locally approximate the objective function as a quadratic approximation. The DFP is one of the most efficient algorithms for complex, nonlinear optimization available (Christensen and Walters, 2004).


99 Maximizing a simple o bjective function The most simple policy optimization option involves defining the weights wi for an objective function of the form = + + + ( 4 14) w here TPj represents total profits from fleet j ER is a mandated rebuilding index that increases as rebuilding targets are met, and ES is an ecological stability index similar to those described by Odum (1969). Once the weights are defined, the output from the optimization procedure can be visualized as a table, a trend plot of iterations, and a triangular figure that describes the way the objectives are being met. This procedure allows the sole owner of the ecosystem to define her objectives and use the ecosystem model to define what the best course of action is. Iteration trends for the objectives and relative effort when maximizing the objective function when w1 = 1 and all other weights are assumed to be zero (profit maximization) in the hypot hetical ecosystem are shown in F igures 44 and 45. One o f the most useful features of EwEs policy optimization routine is that the levels of fishing effort that maximize the objective function defined by the user are automatically sketched into the dynamic model of the ecosystem as an effort for cing function This allows the user to monitor the biomass and other trends in the ecosystem if the prescribed policy where to be put in place. Biomass trends for the economically valuable species if the profit maximizing policy is followed are shown in F igure 4 6. Alter native: prevent cost > earnings A very interesting phenomenon arises in multispecies or ecosystem models of fishing regarding the impacts that one fishery may have on another through the trophic interactions of the target species. Such ecosystem externali ties may dictate the sole

PAGE 100

100 owner of the ecosystem to maintain a fleet that operates at a loss if doing so results in an increase of the population of the species targeted by a profitable fleet. The example of the hypothetical ecosystem may serve to illustr ate this phenomenon. In our case, the trawling fleet targets piscivores, which eat squid. Furthermore, trawls remove macroalgae as incidental catch, therefore allowing higher predation on small fish by both piscivores and squid. Therefore, the trawl fishery provides a positive ecosystem externality to the jigger fishery by removing some of the squids predators and by facilitating a higher availability of food for the squid stock. If the trawl fishery was losing money and the squid fishery was still profit able, a sole owner of the ecosystem would want to keep the trawls operating at a loss just to reap the benefits trawlers provide to the jigger fleet through complex ecosystem interactions. To separate the scenarios in which a sole owner would have control of all the fleets operating in the ecosystem and that in which fleets make independent profit maximization decisions, Ecosim offers the option prevent cost > earnings, which limits fleets to operate only at profitable levels. Ecosim also allows the user to visualize the extent of ecosystem externalities with the Maximize by Fleet Values option, which automatically pops up a window with a matrix of ecosystem relationships between fishing fleets. The idea is that such a matrix could be used to estimate th e compensation required by different fleets for benefiting or adversely affecting other fleets that operate in the ecosystem. Maximizing social utility of a risk averse portfolio Ecosim also provides the user with an option that could be interpreted as that of maximizing social welfare. The social manager is seen as having an objective function of the form

PAGE 101

101 = ( ) + ( ) ( 4 15) where the first term in the left hand side of the equation refers to the profits of all fleets o perating in the ecosystem, the second term refers to the existence value of all biomass groups, and the third term is a variance index that increases when the ecosystem is changed from the Ecopath mass balance base (Christensen and Walters 2004). This function therefore attempts to balance profitability of the fleets operating in the ecosystem with negative ecosystem effects that such profit maximization may bring about. In the hypothetical ecosystem, for instance, maximizing for this risk averse social u tility function prescribes a much more conservative policy than that prescribed in the sole owner profit maximization scenario. To maintain an emphasis in profitability of the fisheries, the relative weights chosen were w1 = 1, w2 = 0.1, and w3 = 0.1. The iteration and predict ed biomass trends are shown in F igures 47, 4 8 and 4 9. Discussion While not developed explicitly as a tool for economic analysis, and somewhat lacking in economic theoretical background, EwE is an interesting tool for fisheries model ing and management. If we allow the definition of fisheries economics to be broadly the study of the interaction between marine and coastal ecosystems and the people that use them, then EwE is definitely a tool with a fruitful prospect. The capacity to mod el trophic interactions that is easily available in EwE is most likely not within the toolkit of most economists, even though the problems studied by fisheries economists are heavily affected by these trophic interactions. A large part of the economic anal ysis capabilities of EwE rely in simple and straightforward common sense.

PAGE 102

102 The bionomic dynamics routine holds particular promise in integrating economic and ecological models within the EwE framework. As the simple manipulations described here suggest, how ever, the routine needs improvement. Careful inclusion of bioeconomic theory in this or a similar routine may be the key to successful integration of these models. The fishing policy search routines allow the user to play around with different objectives of fisheries management, and in doing so it is easy to examine the inherent tradeoffs between these objectives. When considering EwE and similar platforms for ecosystem modeling, however, fisheries decision makers must realize that the capacities to conduct analysis of economic and social impacts are very limited. For instance, the cost structure in EwE is always assumed to be linear in parameters, which in turn implies that the marginal cost and productivity of inputs remains constant across all possible levels of fishing effort. This is a problematic assumption that oversimplifies constraints and decisionmaking at the vessel level. Similarly, fishing effort in EwE does not respond to changes in output prices or input costs. This assumption of exogeneity of fishing behavior is perhaps the biggest shortcoming of the EwE platform when judged from the socioeconomic viewpoint. The exogeneity of fishing effort, which extends to the entire human component in the EwE platform, implies that this software is inappropriate for conducting the wide majority of analyses that fisheries economists perform. For instance, the issues of overcapitalization and capital stuffing, the race for fish, transferrable fishing quotas, among others, cannot be addressed using EwE. Simi larly, all issues related to property rights and fishing regulations must be ignored. The first step in addressing this problem

PAGE 103

103 is changing the way in which fishing effort is modeled to make it endogenous to parameters set by the user. In essence, the cali bration of an ecosystem model that realistically includes the human component must require at least as much effort in the ecological aspects as in the socioeconomic aspects. It is important to note ease of learning, availability of documentation, and user friendliness of the software as major advantages of using EwE, especially as it relates to building ecosystem models with a diverse suite of ecological features. However, the documentation on conducting economic type analyses on EwE is not as rich or as c lear, and the shortcomings of the platform caused by the exogeneity of fishing effort make it inappropriate for most economic analysis. EwE and its users would definitely benefit from inclusion of economists in the development of successive versions of the software.

PAGE 104

104 Table 41. Basic input parameters for the hypothetical model Group Name Habitat Area (fraction) Biomass in habitat area (t/km) Prod./ biom. (/year) Cons./ biom. (/year) Ecotrophic efficiency Seabirds 1 1 0.3 1 Piscivores 1 5 0.3 1 Small Fish 1 15 1 4 Squid 1 2 2 10 Benthos 1 40 5 20 Zooplankton 1 50 20 80 Macro algae 1 5 0.5 1 Phytoplankton 1 100 150 Detritus 1 100 Table 42. Diet composition matrix Prey / Predator 1 2 3 4 5 6 1 Seabirds 2 Piscivores 0.25 3 Small Fish 0.5 0.45 0.3 4 Squid 0.25 0.05 0.05 5 Benthos 0.5 0.05 6 Zooplankton 0.95 0.7 7 Macro algae 8 Phytoplankton 1 9 Detritus 0.95 10 Import 11 Sum 1 1 1 1 1 1 Table 43. Landings in the hypothetical ecosystem (t/km2) Group Name Trawls Jiggers Traps Total Seabirds 0 Piscivores 0.7 0.4 1.1 Small Fish 0 Squid 0.3 0.3 Benthos 0 Zooplankton 0 Macro algae 0 Phytoplankton 0 Detritus 0 Sum 0.7 0.3 0.4 1.4

PAGE 105

105 Table 44. Discards in the hypothetical ecosystem (t/km2) Group Name Trawls Jiggers Traps Total Seabirds 0.1 0.1 Piscivores 0 Small Fish 0 Squid 0 Benthos 1 1 Zooplankton 0 Macro algae 3 3 Phytoplankton 0 Detritus 0 Sum 4 0.1 0 4.1 Table 45. Cost structure of the fishing fleets in the hypothetical ecosystem Name of fleet Fixed cost (%) Effort related cost (%) Sailing related cost (%) Profit (%) Total value (%) Trawls 45 30 15 10 100 Jiggers 45 25 20 10 100 Traps 40 20 35 5 100 Table 46. Market prices in the hypothetical ecosystem (hypothetical currency) Group Name Trawls Jiggers Traps Seabirds 0 1 1 Piscivores 250 1 250 Small Fish 1 1 1 Squid 1 110 1 Benthos 1 1 1 Zooplankton 1 1 1 Macro algae 1 1 1 Phytoplankton 1 1 1 Detritus 1 1 1 Table 47. Nonmarket prices in the hypothetical ecosystem (hypothetical currency) Group Name Value/unit biomass Seabirds 100 Piscivores 0 Small Fish 0 Squid 0 Benthos 0 Zooplankton 0 Macro algae 0 Phytoplankton 0 Detritus 0

PAGE 106

106 Table 48. Economic output from the mass balance procedure in Ecopath Group \ value Trawls Jiggers Traps Catch value Non market value Total value Seabirds 0 100 100 Piscivores 175 100 275 0 275 Squid 33 33 0 33 Benthos 0 0 0 Macro algae 0 0 0 Total value 175 33 100 308 100 408 Total cost 157.5 29.7 95 282.2 Total profit 17.5 3.3 5 25.8 Table 49. Employment by fleet in the hypothetical ecosystem Fleet Catch Value Proportion J obs Trawls 175 0.568 170.5 Jiggers 33 0.107 32.14 Traps 100 0.325 97.4 Total 308 1 300 Figure 41. Flow chart of the hypothetical ecosystem. Straight lines represent trophic flows, while dashed lines represent habitat interactions.

PAGE 107

107 Figure 42 Expected biomass trends for economically important species in the hypothetical ecosystem if fishing effort is assumed constant for 20 years Figure 43 Biomass trends in the hypothetical ecosystem when openaccess or bionomic dynamics are simulated Figure 44 Iteration trends of the different objectiv es when maximizing profits in the hypothetical ecosystem 0 1 2 3 4 5 6 0 10 20 Seabirds Piscivores Squid 0 1 2 3 4 5 6 0 5 10 15 20 Seabirds Piscivores Squid 0 0.5 1 1.5 2 2.5 3 3.5 0 5 10 Econ. Social Ecol.

PAGE 108

108 Figure 45 Recommended effort trends to allow profit maximization in the hypothetical ecosystem Figure 46 Biomass trends in the hypothetical ecosystem if the profit maximizing policy is follow ed Figure 47 Iteration trends of the different objectives when maximizing for risk averse social utility in the hypothetical ecosystem 0 1 2 3 4 5 0 5 10 Trawls Jiggers Traps 0 2 4 6 8 10 0 5 10 15 20 Seabirds Piscivores Squid 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0 5 10 15 20 25 Econ. Social Ecol.

PAGE 109

109 Figure 48 Recommended effort trends to allow maximization of the risk averse utility function in the hypothetical ecosystem Figure 49 Biomass trends in the hypothetical ecosystem if the risk averse social utility maximization policy is followed 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 0 5 10 15 20 Trawls Jiggers Traps 0 1 2 3 4 5 6 0 5 10 15 20 Seabirds Piscivores Squid

PAGE 110

110 CHAPTER 5 CONCLUSIONS The three essays in this dissertation explore a few of the many emerging issues in mankinds relationship with the natural world, with a particular emphasis on marine resources. In the first essay we explored the issue of water pollution from a major oil spill, the DWH, and its effect on recreational use of marine resources. As the worlds population continues to grow and incomes increase, we can expect that recreational demand for marine resources will continue to expand. The same is true of the global dem and for energy, and much of it will come from offshore oil drilling or will be transported by sea. The problem of oil spills interfering with recreational use of marine resources, therefore, is likely to be of prime importance in the near future. The methods and analysis developed in that essay, therefore, are likely to maintain relevance and their use is likely to grow. Similarly, the systematic approach to decision making in natural resource management developed in the second essay may have its most productive days still ahead. As pressure on our natural resources increases, managers will be forced to make difficult decisions that may require choosing between competing alternatives. Portfolio selection, which has been important in financial decisionmaking for the past 50 years, may yet prove quite useful in t he field of natural resources. The essential questions for portfolio development outlined in the second essay will aid practitioners in the creation of their own portfolio selection models for natural resource management. Finally, the modeling of entire ecosystems, discussed in the third essay, will continue to grow as a field due to the increased recognition of the importance of ecosystem interactions. While the marine sciences have pioneered this appr oach, other

PAGE 111

111 disciplines that deal with natural resource management are likely to follow. Integration of economics into these models will likely prove to be a fruitful area of research. The analysis in this essay may help pave the way for a more complete in tegration of economics and human behavior into EwE and other existing ecosystem models.

PAGE 112

112 APPENDIX A COMPONENTS OF A VESSELS TOTAL OPERATION COSTS EwE allows the user to enter the percentage of the total value of the catch, or total revenue of a representative vessel in a fishing fleet, that is allocated to the different kind s of fixed and variable costs. Here are some common operating expenses of a fishing vessel, below the EwE cost category to which they should be ac counted in: Fixed costs a) Regular maintenance b) Major repairs c) Depreciation and capital equipment d) Overhead e) Insurance Effort related variable costs a) Bait, tackle, ice, and other supplies b) Wages or salary for crew and captain (or share of catch) c) Owners time (if ve ssel is owner operated) Sailing related variable costs a) Boat fuel and oil Profits a) Whatever money is left, if any, after covering all other expenses. Meant to reflect the return of owning a vessel in a given fishery.

PAGE 113

113 APPENDIX B T OTAL ECONOMIC VALUE OF AN ECOSYSTEM The total economic value of an ecosystem is the sum of the following value components (Pearce and T urner 1990): Use Values Direct Use Value: Value for actual use or enjoyment of a n ecosystem or its components. Values for fishing and other extractive uses of species are examples of dire ct use. Most commodities traded in markets fall under this category. Indirect Use Value: Non consumptive services provided by ecosys tems fall under this category. Scenic vistas provided by coastlines and storm mitigation services provided by coastal wetl ands are good examples. While these values are sometimes captured in the price of related commodities (i.e., housing), these goods are almost never traded in markets. Option Value: Represent a preference to preserve an ecosystem for some unspecified use i n the future. Conservation of coral reefs and the species associated with them to keep open the option of finding medically valuable species in the future is an example. Extinction of a species, for instance, repr esents a complete loss of option value. Non Use Values Bequest Value: Preference to conserve an ecosystem or its components for ones heirs or for future generations in general. Conservation of an ecosystem or a species for the sake of ones children or gr andchildren is an example of bequest value. Existence Value: Relates to the satisfaction received by individuals for knowing that an ecosystem or one of its components exists, even though no direct or indirect use is involved.

PAGE 114

114 Treatment of Nonmarket values in EwE The current version of EwE assumes that there is a direct linear relationship between biomass of the species group and its nonmarket price or value, so that a doubling in biomass of a species group results in a doubling of its nonmarket va lue (C hristensen et al. 2005). However, the relationship between nonmarket value a nd extinction is more complex. If extinction is unacceptable from societys point of view, then the nonmarket value of the last unit harvested before extinction occurs is very hi gh (Clark et al. 2010). EwEs current treatment of nonmarket values does not account for this social aversion to extinction of species groups. Ecosim can go around this problem by use of the Log Utility optimization option, which makes extinction of a species with nonmarket value unacceptable.

PAGE 115

115 LIST OF REFERENCES Arndt A, Van Neer W, Hellemans B, Robben J, Volckaert F, Waelkens M (2003) Roman trade relationships at Sagalassos (Turkey) elucidated by ancient DNA of fish remains. Journal of Archaeolog ical Science 30: 10951105 Baldursson FM, Magnuss on G (1997) Portfolio Fishing. Scandinavian Journal of Economics 99 :389 403 Bockstael NE (1995) Travel cost models. In : Bromley D (ed) Handbook of Environmental Economics. Blackwell, Cambridge MA Bockstael NE McConnell KE, Strand IE (1991) Recreation. In : Braden JB, Kolstad CD (eds) Measuring the Demand for Environmental Quality North Holland, New York B ockstael NE, Hanemann WM, Kling CL (1987) Estimating the value of water quality improvements in a recreati onal demand framework. Water Res ources Research 23:951 960 Boxall PC, Adamowicz WL (2002) Understanding heterogeneous preferences in random utility models: a latent class approach. Environmental and Resource Economics 23 :421 446 Brown L, Gardner G, Halweil B (1999) Beyond Malthus. W.W. Norton & Company, New York Brodziak J, Link J (2002) Ecosystem based fishery management: What is it and how can we do it? Bulletin of Marine Science 70: 589 611 Carson RT, Mitchell RC, Hanemann WM, Kopp RJ, Presser S, Ruud PA ( 2003) Contingent v aluation and lost passive use: d amages f rom the Exxon Valdez oil spill. Environmental and Resource Economics 25:257286 Christensen V (2009) Ecopath with Ecosim: linking fisheries and ecology. In: Jorgensen S, Chon TS, Recknagel F (e ds) Handbook of Ecological Modelling and I nformatics. WIT Press, Southampton Christensen V, Pauly D (1992) ECOPATH II a software for balancing steady state ecosystem models and calculating network charact eristics. Ecological Modelling 61: 169 185 Christen sen V Walters C (2004) Ecopath with Ecosim: methods, capabilities, and lim itations. Ecological Modelling 172:109139 Christ ensen V, Walters C, Pauly D (2005) Ecopat h with Ecosim: A User's Guide. Fisheries Centre University of British Columbia, Vancouver

PAGE 116

116 Clark C, Munro G, Sumaila U R (2010) Limits to the Privatization of Fishery Resources. Land Economics 86:209 218 Conde Castro I (1991) Apuntes economi cos de la pesca en Colombia II. Ens ayos de Economia 2:121144 Corporacion Colombia Internacional (2006) Pe sca y Acuicultur a Colombia 2006. Ministerio de Agricultura y Desarrollo Rural, Bogota Crowe KA, Parker WH (2008) Using portfolio theory to guide reforestation and restoration under climate change scenarios. Climatic Change 89 :355370 Cryer JD, Chan KS (2008) Time Series Analysis with Applications in R. Springer, New York Cvitanic J, Zapatero F (2004) Introduction to the Economics and Mathematics of Fi nancial Markets. MIT Press, Cambridge Edwards SF, Link JS Rountree BP (2004) Portfolio management o f wild fish stocks. Ecological Economics 49:317329 FAO Fishe ries and Aquaculture Department (2010) The State of World Fisheries and Aquaculture. FAO, Rome Feather PM (1994) Sampling and aggregation issues in r andom utility model estimation. American Journal of Agricultural Economics 76: 772 780 Figge F (2004) Bio folio: applying portfolio theory to biodiversity. Biodive rsity and Conservation 13:827849 Flores NE, Thacher J (2002) Money, who needs it? Natural resource damage assessment. Contemporary Economic Policy 20:171178 Garza Gil MD, PradaBlanco A, Vazquez MX (2006) Estimating the short term economic damages from the Prestige oil spill in the Galician fisheries and tourism. Ecologi cal Economics 58:842849 Gordon HS (1954) The economic theory of a commo n property resource: the fishery. Journal of Political Economy 62: 124 142. Greene G, Moss CB, Spreen TH (1997) Demand for recreational fishing in Tampa Bay, Flor ida: a random utility approach. Marine Resource Economics 12:293305 Grigalunas TA, Anderson RC, Brown GM, Congar R, Meade NF, Sorensen PE (1986) Estimating the cost of oil spills: Lessons from the Amoco Cadiz incident. Marine Re source Economics 2:239262

PAGE 117

117 Haab T, Whitehead JC, McConnell KE (2001) The economic value of marine recreational fishing. NO AA Technical Memorandum NMFS SEFSC 466 Haab T, Hicks RL, Schnier K, Whitehead JC (2012) Angler heterogeneity and the species specific demand f or marine recreational fishing. Marine Resource Economics 27:229251 Haab T, McConnell KE (2002) Valuing Environmental and Natural Resources: The Econometrics of Non market Valuation. Edward Elgar, Northampton, MA Halpern BS, White C, Lester SE, Costello C, and Gaines SD (2011) Using portfolio theory to assess tradeoffs between return from natural capital and social equity across space. Bi ological Conservation 144:14991507 Hanna S (1998) Institutions for marine ecosystems: Economic incentives and fishery management. Ecologi cal Applications 8:S170S174 Hannon B (1973) The structure of ecosystems. Jour nal of Theoretical Biology 41:535546 Hausman JA, Leonard GK, McFadden D (1995) A utility consistent, combined discrete cho ice and count data model assessi ng recreational use losses due to natural resource damage. J ournal of Public Economics 56: 1 30 Hicks RL, Gautam AB, Van Hoorhees D, Osborn M, Gentner B (2000) An introduction to the NMFS Marine Recreational Fisheries Statistics Survey with an emphasis on economic valuation. Mar ine Resource Economics 14:375385 Hills J, Carlisle M, Le Tissier M, Muir D, Robinson C (2009) Landscape scale analysis of ecosystem risk an d returns: A new tool for ICZM. Marine Policy 33:887900 Hindsley P, Landry CE, Gentner B (2011) Addressing onsite sampling in recreation site choice models. Journal of Environmental Ec onom ics and Management 62:95110 Holland D, Sanchirico J, Johnston R, Joglekar D (2010) Economic Analysis for Ecosystem Based Management: Applications to Marine and Coastal Environments. Earthscan, Washington DC Hyndman RJ Khandakar Y (2008) Automatic time series forecastin g: The forecast package for R. Journal of Stati stical Software 27:122 Johannes R (1982) Traditional Conservation Methods and Protected M arine Areas in Oceania. Ambio 11:258261 Jones CA (1997) Use of nonmarket valuation methods in the co u r troom: recent affirmative precedents in natur al resource damage assessments. Water Resources Update 109: 1018

PAGE 118

118 Jones CA, and Pease KA (1997) Restorationbased compensation measures in natural resource liability statutes. Contemporary Economic Policy 15:111 122 Kaoru Y (1995) Measuring marine recreation benefits of water quality improvements by t he nested random utility model. Resourc e and Energy Economics 17:119136 Koell ner T, Schmitz OJ (2006) Biodiversity, ecosystem function, and investment risk. BioSci ence 56:977985 Kurlansky M (1998) Cod: A biography of the fish that changed the world. Penguin Books, New York Larkin SL, Sylvia G, Tuininga C (2003) Portfolio analysis of optimal seafood product diversific ation and resource management. Journal of Agricul tural and Res ource Economics 28:252271 Lauck T, Clark CW, Mangel M, Munro GR (1998) Implementing the precautionary principle in fisheries manag ement through marine reserves. Ecologi cal Applications 8:S72S78 Loureiro ML, Ribas A, Lopez E, and Ojea E (2006) Estimated costs and admissible claims linked to the Prestige oil spill. Ecological Economics 59:4863 Loureiro ML, Loomis JB, Vazquez MX (2009) Economic valuation of environmental damages due to t he Prestige oil spill in Spain. Environmental and Resource Economics 44:537553 Lupi F, Feather PM (1998) Using partial aggregation to reduce bias in random utility travel cost models. Water Resources Research 34:35953603 Marean CW, Bar Mattews M, Bernatchez J, Fisher E, Goldberg P Herries AIR, Jacobs Z, Jerardino A, Karkanas P, Minichillo T, Nilssen PJ, Thompson E, Watts I, Williams HM (2007) Early human use of marine resources and pigment in South Africa during the Middle Pleistocene. Nature 449:905908 Markowitz H (1952) Portfolio Selection. The Jo urnal of Finance 7: 7791 Mazzotta MJ, Opaluch JJ, Grigalunas TA (1994) Natural resource damage assessment: The role of resource restoration. Nat ural Resources Journal 34:153 178 McConnell KE, Strand IE, Blake Hedges L (1995) Random utility models of recreational fishing: Catching fish using a Poisson process. Marine Resource Economics 10: 247 261 McFa dden D (1974) Conditional logit analysis of qualitative choice behavior. In: Zarembka P (ed) Frontiers in Econometrics Academic Press, New York

PAGE 119

119 McFadden, D (1977) M odelling the Choice of Residential Location. Discussion paper 477. Cowles Foundation at Yale University, New Haven McLeod K, Lubchenco J, Palumbi S, Rosenberg A (2005) Scientific Consensus Statement on Marine Ecosystem Based Management. Signed by 221 acade mic scientists and policy experts with relevant expertise and published by the Communication Part n ership for Science and the Sea at Morey ER, Shaw WD, Rowe RD (1991) A discretechoice model of recreational participation, sit e choice, and activity valuation when complete trip data are not available. Journal of Environmental Economics and Management 20:181201 Mullon C, Freon P, Cury P (2005) The dynamics of collapse in world fisheries. Fish and Fisheries 6:111120 National Res earch Council (2006) Dynamic changes in marine ecosystems: fishing food webs, and future options. National Academies Press, Washington DC Odum E (1969) The strategy of ecosys tem development. Science 164:262270 Parsons GR, Kang AK (2010) Compensatory restoration in a random util ity model of recreation demand. Contemporary Economic Policy 28:453463 Parsons GR, Kealy MJ (1992) Randomly drawn opportunity sets in a random ut ility model of lake recreation. Land Economics 68:93106 Parsons GR, Needleman MS (1992) Site aggregation in a random utility model of recreation. Land Economics 68:418433 Pauly D, Christensen V, Dalsgaard J, Froese R, Torres FJ (1998) Fishing down marine food webs. Science 279: 860 863 Pearce D, Turner R (1990) Economics of Natural Resources and the Environment. Harvester Wheatsheaf, Hetfordshire, UK Perruso L, Weldon RN, Larkin SL (2005) Predicting optimal targeting strategies in multispecies fisheries: a portfolio approach. Marin e Resource Economics 20:2545 P insky ML, Jensen OP, Ricard D, and Palumbi SR (2011) Unexpected patterns of fisheries collapse in the world's oceans. Proceedings of the National Academ y of Sciences 108:83178322 P itcher T, Hollingworth C (2002) Fishing for fun: Where's the catch? In: Pit cher T, Hollingworth C (eds) Recreational Fisheries: Ecological, Economic and Social Evaluation. Blackwell, Oxford

PAGE 120

120 Plagyani EE (2007) Models for an ecosystem approach to fisheries. Fisheries Technical Paper 477. FAO, Rome Polovina J (1984) Model of a Reef Ecosystem: The ECOPATH Model and Its Application to French F rigate Shoals. Coral Reefs 3:111 Radulescu M, Radulescu CZ, Rahoveanu MT, Zbaganu G (2010) A portfolio theory approach to fishery management. Studies in Informatics and Control 19: 285294 Sainsbury K, Campbell R, Whitelaw A (1993) Effects of trawling on the marine habitat on the North West Shelf of Australia and implications for sustainable fisheries management. In : Hancock D (ed) Sustainable Fisheries through Sus tainable Habitat. Bureau of R ural Sciences Proceedings, AGPS, Canberra Salamone D (2012) A plan to save gag grouper. room/fact sheets/a plan to save gaggrouper 8589942 081 (accessed January 19, 2013) Sanchirico JN, Smith MD, Lipton DW (2008) An empirical approach to ecos ystem based fishery management. Ecological Economics 64:586596 Schindler DE, Hilborn R, Chasco B, Boatright CP, Quinn TP, Rogers LA, Webster MS (2010) Population diversity and the portfolio e ffect in an exploited species. Nature 465: 609612. Smith T (1994) Scaling Fisheries: The Science of Measuring the Effects of Fishing, 18551955. Cambridge University Press, Cambridge Thomas MH, Lupi F, Harding DB (2010) A policy tool for evaluating investments in public boat ramps in Florida: a random utility model approach. Southern Agricultural Economics Association Annual Meetings, Orlando, FL Tilman D, Reich PB, Knops JMH (2006) Biodiversity and ecosystem stability i n a decade long grassland experiment. Nature 441:629632 Train KE (2003) Discrete Choice Methods with Simulation. Cambridge University Press, New York Train KE (1998) Recreation demand models with taste differences over people. Land Economics 74:230239 Up ton HF (2011) The Deepwater Horizon Oil Spill and the Gulf of Mexico Fishing Industry. Congressional Research Service Report for Congress, Washington DC Van Biervliet K, Le Roy D, Nunes PALD (2005) A contingent valuation study of an accidental oil spill al ong the Belgian coast. In : Maes F (ed) Marine Resource Damage Assessment: Liability and Compensation for Environmental Damage. Springer, Dordrecht, The Netherlands

PAGE 121

121 Whitehead JC, Haab T (2000) Southeast Marine Recreational Fishery Statistics Survey: Distan ce and catch based choice sets. Marine Resource Economics 14:283298

PAGE 122

122 BIOGRAPHICAL SKETCH Sergio Alvarez was born in Manizales, Colombia, in 1982, the year of the dog. He attended Colegio Granadino from kindergarten until graduating high school in 2000. Sergio then moved to South Florida, where he attended Broward Community Col lege and obtained an Associate of Arts degree in agricultural s ciences. After that, he came to the University of Florida in Gainesvil le, where he earned a Bachelor of Arts degree in environmental s ciences and completed graduate studies in the Food and Resource Economics (FRE) Department. He holds Mast er of Science and PhD degrees in natural resource economics Sergio is interes ted in the interaction between mankind and nature in general, but his graduate research focused mostly on fisheries issues. Some specific areas where he has conducted research include choice modeling, nonmarket valuation, ecosystem and bioeconomic modeling. Besides fisheries, Sergio is interested in protected area management and the conservation of biodiversity. Sergio has been married to Cassie since 2008, and they try to coexist with a German Shepherd, miniature C ollie mix named Rumi and a black cat named Malinche. He enjoys outdoor activities, and has recently developed a special interest in kayak and canoe fishing, which he does mainly for research purposes.