Abstract
The degree to which selectivity in fisheries is malleable to changes in incentive structures is critical for policy design. We examine data for a multispecies trawl fishery before and after a transition from management under common-pool quotas to a fishery cooperative and note a substantial shift in postcooperative catch from bycatch and toward valuable target species. We examine the margins used to affect catch composition, finding that large- and fine-scale spatial decision making and avoidance of night-fishing were critical. We argue that the poor incentives for selectivity in many systems may obscure significant flexibility in multispecies production technologies. (JEL Q22, Q28)
I. INTRODUCTION
Despite decades of unprecedented technological development that has increased the scale and reduced the cost of fishing, the production process of capture fisheries still retains some of the uncertainty of its huntergatherer past. While increasingly sophisticated fish-finding equipment and modified vessels have given skippers the option of fishing in a more targeted and selective manner, the quantity and composition of catch remain partially controlled random variables–with fishermen regularly harvesting species, life stages, and sexes of fish that they, or managers, would prefer to avoid. The economic and ecological ramifications of imperfect selectivity can be significant. Lack of selectivity can limit fishermen’s ability to alter their catch composition to respond to price differentials across species or within species induced by factors such as sex, age, or size. It can also hamper fishermen’s ability to adapt to regulations limiting impacts across species, sex, size, or age classes, thereby undermining or raising the cost of achieving ecosystem-based management objectives.
The costs of nonselectivity have been particularly well documented in the literature on bycatch and discards (Crowder and Murawski 1998; Hall and Mainprize 2005; Harrington, Myers, and Rosenberg 2005; Kelleher 2005; Patrick and Benaka 2013). Bycatch often refers to the incidental catch and discard of species not specifically targeted for retention and sale, but may also refer to secondary species that may or may not be targeted yet are retained. Because of the variety of regulations that apply to bycatch in different fisheries, bycatch can have different definitions and exhibit a wide array of impacts. In some cases, bycatch may involve the capture and mortality of macrofauna such as cetaceans, pinnipeds, and sea turtles. While such encounters are often comparatively rare events, they nevertheless can have significant impacts on the stock status of endangered or threatened species and result in large welfare losses to members of society, which may place a high value on the preservation of these species (Finkbeiner et al. 2011; Lewison and Crowder 2007; Lewison et al. 2004; Moore et al. 2009). In many other cases bycatch consists of the catch of species in a multispecies fishery–or sizes or sex within a single species–for which landing of the catch is economically undesirable or legally prohibited. When this catch is discarded, a portion of the discarded fish die, causing a loss of potential fishery yield while also producing complex externalities through ecosystem interactions. In many multispecies fisheries–or fisheries with significant interactions with protected species–quotas are often set and monitored for a range of species, with the season potentially being limited by the availability of quota for bycatch species. Under the reauthorized Magnuson-Stevens Fishery Conservation and Management Act of 2006, U.S. fisheries managers have a mandate to avoid overfishing of effectively all stocks under their purview, with the result that quotas are typically set to protect the most vulnerable stocks. There is also a somewhat vague mandate that vessels reduce bycatch “to the extent practicable.” In practice, fishermen have often been unable to accommodate their catch composition to the ratios of quota allocations, often leading to premature closures of valuable target fisheries and substantial economic impacts (Patrick and Benaka 2013)– losses that should not occur if fishermen could fish with perfect selectivity at negligible cost.
The biological and technical fundamentals of nonselectivity are straightforward. When deployed in a particular spatial and temporal context, all gear types sample from the fish stocks in proximity of the gear at that point in space and time, where the expected catch composition depends on the morphological and behavioral traits of species and the selectivity of the gear toward these traits.1 Parameters of the economic and regulatory context–such as input and output prices, individual quotas for different species, or timearea closures–influence realized species composition by proximate factors, such as the choice of which spatiotemporal “urns” are sampled and the particular sampling process (i.e., the choice of gear) (e.g., Asche 2009; Asche, Gordon, and Jensen 2007; Squires 1987). In both fisheries science and a sizable portion of the fisheries economics literature, it has become commonplace, perhaps due to the legacy of highly aggregated data, to implicitly think of catch composition as the limit that results from this repeated sampling process. In this case, the heterogeneities of gear configurations and the spatial and temporal dimensions of gear usage are implicitly averaged out. The resulting estimates of catchabilities and selectivity curves are therefore treated as parameters in stock assessments,2 and effort becomes an index that is viewed as interchangeable (perhaps after some form of standardization) across all gear deployments in a season. While often a useful simplifying assumption, this practice conveys an impression of inflexibility of catch composition that may be at odds with the underlying fishing process.
If variations in the spatial and temporal deployment of gear have important implications for catch composition at the individual geardeployment level due to spatial or temporal instabilities in the mix of species or their susceptibility to a particular gear, then this implies that there may be a significant aspect of the catch composition that is behavioral in nature rather than purely technical. This creates the potential for management policy to exert some degree of control over this aspect of fishing through its influence on the incentives faced by fishermen (Branch et al. 2006; Grafton et al. 2006; Hilborn, Orensanz, and Parma 2005; Wilen 2006). In short, the realized outcomes in a fishery can be viewed as the intersection of the production possibilities afforded by the combination of the marine environment and harvesting technology and the constraints embodied in economic conditions (e.g., fish and fuel prices) and management policies. Viewed in this light, management institutions impact how potential catch compositions are translated to actual catch compositions.
Determining the potential behavioral malleability of catch composition is critical to current debates in multispecies management. Fundamental to rights-based management in a multispecies fishery is composing a range of total allowable catches (TACs) for all managed target and nontarget species; portfolios of rights to the catch of these species are then allocated to individuals or to groups of fishermen. However, due to imperfectly selective fishing gear, it is difficult to know beforehand a fisherman’s catch composition, and thus, fishermen may find it difficult or impossible to conform their catch to match their ex ante portfolio of quota allocations (Pascoe, Koundouri, and Bjørndal 2007; Squires et al. 1998; Squires and Kirkley 1996). Possible consequences include extremely high quota prices for “choke” species, a collapse in the markets for “slack” species or poorly functioning markets for choke species (Holland 2013), rampant illegal discarding, data fouling, and subverted quota markets (Copes 1986). In practice, these complications have induced some rights-based multispecies management programs to introduce flexibility through the use of “catch-quota balancing” mechanisms such as retrospective quota balancing, deemed value payments, rollover provisions, and cross-species exchanges (Sanchirico et al. 2006). These mechanisms enable fishermen to more fully utilize “unbalanced” TACs; however, they also increase the risk of overexploitation, adding to the critique that rightsbased approaches may not be appropriate for multispecies fisheries.
A handful of papers using data from nonrights-based fisheries predict that rights-based systems may face serious challenges due to weak substitution potential between species (Pascoe, Koundouri, and Bjørndal 2007; Pascoe, Punt, and Dichmont 2010; Squires et al. 1998; Squires and Kirkley 1991, 1996). In contrast, evidence from multispecies individual transferable quota (ITQ) fisheries shows that far greater flexibility in outputs is possible than often previously thought (Branch and Hilborn 2008; Sanchirico et al. 2006). These disparate findings suggest that the production technology revealed through empirical work may say just as much (or more) about the incentives facing fishermen in a particular fishery as about the policy-neutral production possibilities in the fishery. Nevertheless, changes in incentives alone may not completely alter catch composition if there are fundamental rigidities (e.g., strong complementarities) due to constraints in the natural environment or preexisting regulations (e.g., large-scale closures). Interannual variability may also occur in encountered mixes among species, which may present significant challenges for vessels balancing quotas that were based on the historical catch ratios among species.
The resolution of this debate will ultimately rely on empirical evidence, and yet, there are remarkably few papers utilizing data before and after major policy changes in multispecies fisheries to examine the nature of trade-offs between species (Branch and Hilborn 2008; Paul, Torres, and Felthoven 2009). We contribute to closing this gap in the literature in two ways. First, we utilize detailed data at the individual haul level from a large trawl fishery before and after a major policy change. In the initial management structure, the catch of all species, including bycatch species, was regulated by the common-pool assignment of multiple TACs for each species. Under this common-pool incentive system, reaching a bycatch species catch limit frequently closed the target fishery prematurely. Under the new management structure, individual vessels operated under a multispecies catch share system, with individual accountability for their catch of target and bycatch species. We find dramatic evidence of a shift in overall catch composition away from bycatch species and toward valuable target species, as well as far less variability in the target/bycatch ratio. Second, we conduct a detailed analysis of the many margins of behavior (Smith 2012) that fishermen employed to alter their catch composition. We show that fishermen were able to alter their catch composition substantially through their choices of when and where to fish on fine and coarse scales. We find evidence that large-scale shifts in fishing grounds, larger and more immediate reactions to undesired catch compositions, and reduced fishing at night have all contributed significantly to the observed changes. Importantly, these margins of change were all available to fishermen before the institutional change and yet were not adopted. This suggests that management systems that provide few incentives for selective fishing may obscure fishermen’s ability to alter their catch composition. More generally, new incentives can induce a wide variety of significant changes in fishing behavior, but, depending on spatial and temporal variation in catch composition, the extent and costs of these behavioral changes may be more or less substantial.
II. FISHERY BACKGROUND
The Bering Sea-Aleutian Islands (BSAI) nonpollock groundfish trawl fishery is a relatively small fleet of vessels that fish the waters of the Eastern Bering Sea and the Aleutian Islands in the North Pacific using bottom trawl gear to catch a variety of groundfish species. Vessels embark on trips of one to two weeks in length and minimally process harvested fish onboard. Twenty-three vessels have actively participated in the fishery since the early 2000s, ranging in size from 91 to 295 feet (median = 154), with horsepower ranging from 850 to 7,000 (median = 2,250).
The fishery can be divided into two broad subfisheries: one in the Aleutian Islands, dominated by trawling for Atka mackerel (with some targeting of Pacific cod, Pacific ocean perch, and a range of rockfish species), and a second in the Bering Sea fishery, a highly mixed fishery on the relatively shallow shelf area of the Eastern Bering Sea, where the dominant targets are a range of flatfish species (e.g., yellowfin sole, rock sole, flathead sole, and arrowtooth flounder) and Pacific cod. Targeting behavior, and the resulting spatial-temporal allocation of effort over the course of a season, is largely determined by a combination of natural, economic, and regulatory factors.
Pre-2008 Regulation
Management of the BSAI groundfishes is conducted by the National Marine Fisheries Service (NMFS) and the North Pacific Fishery Management Council (NPFMC). Management policies have largely consisted of a complicated system of regulations including spatial and temporal closures and limits on participation, target and prohibited species catch, and discards. Prior to 2008, the BSAI nonpollock groundfish trawl fishery operated as a limited license program with fleetwide TACs allocated to each target species. Regulators at the NMFS Alaska Regional Office monitored the progress of fleetwide catch against the TACs throughout the season, using data collected by onboard observers and weekly production reports. If regulators anticipated that a particular species’ TAC would be exceeded, the fishery was closed to “directed fishing,” resulting in a substantial reduction in the proportion of that particular species that could be retained.
Management also restricts the catch and retention of prohibited species. Prohibited species are valuable targets to fishermen outside of the BSAI groundfish fishery and, based on long-standing policy, cannot be retained by the BSAI groundfish fleet. Prohibited species include Pacific halibut, king, Tanner and snow crab, Chinook and chum salmon, and herring. The most problematic prohibited species prior to 2008 was Pacific halibut, a species that coexists in relatively small numbers with the groundfish target species. Given the limitations on the selectivity of trawl gear, some quantity of halibut bycatch by the BSAI groundfish fleet is unavoidable. The BSAI Groundfish Fishery Management Plan (NPFMC 2013b) allocates quotas for prohibitive species catch (PSC), which are strictly limited regardless of the species’ biomass. Prior to 2008, PSC was allocated to the BSAI groundfish fleet as a common property TAC.
Due to the complementary role one target species plays in the catch of another, the exhaustion of one species’ TAC often resulted in the premature closure of another. In response, seasonal TAC allocations were divided into finer seasonal suballocations, resulting in a redistribution of fishing effort over a wider portion of the year with a series of subfisheries with alternating closures and reopenings. The fleetwide allocation of PSC quota was divided between these various target fisheries based on their anticipated usage of the quota, in an effort to ensure the potential targeting of all fisheries in a year. Given the common property nature of the Pacific halibut allocations, there was a general lack of individually costly effort to avoid halibut bycatch (Abbott and Wilen 2010, 2011), resulting in many subfisheries–particularly rock sole and yellowfin sole–closing prematurely when a binding halibut TAC was reached, leaving millions of dollars of unharvested target-species quota.
The act of dividing TAC allocations into finer seasonal suballocations gave rise to a complex pattern of seasonal target fisheries over the course of a given year (Figure 1). The BSAI groundfish fishery began January 20 each year, with the early weeks of the season devoted to a rock sole fishery in the Eastern Bering Sea and an Atka mackerel fishery in the Central Aleutian Islands. In most years before 2008, the rock sole fishery was closed prematurely due to a binding halibut TAC. Subsequently, vessels primarily targeted yellowfin sole for the majority of the spring in the Bering Sea shallows before it too typically suffered from a premature closure due to excessive halibut bycatch. Afterward, species targeting became rather diversified with additional openings and closures of the rock sole and yellowfin sole fisheries. The year typically ended with directed fishing by some vessels for Atka mackerel in the Aleutian Islands. The nature of this management system forced a focus on different species at certain time periods, even if those species were available at that period only in low quantities or with large amounts of bycatch.
Weekly Production and Selected Fishery Openings and Closures for the Bering Sea-Aleutian Islands Groundfish Fisheries in 2006
Post-2008 Regulation
In 2008, Amendment 80 (A80) to the BSAI Groundfish Fishery Management Plan (NPFMC 2007a) was implemented. The provisions of A80 were designed to facilitate increased target catch and profits, reduce bycatch and discards, and increase flexibility while complying with target and prohibited species TACs.
Implementation of A80 resulted in a number of changes to fishery regulations. First, A80 granted a defined share of the total A80 TAC for the six primary target species (yellowfin sole, rock sole, flathead sole, Pacific cod, Atka mackerel, and Pacific Ocean perch) to each vessel in the previous limited-entry program according to its catch history. Second, vessels could vest their shares in either a cooperative formed by participating members or in the limited-access common-pool fishery. Cooperatives are given considerable flexibility as to how catch entitlements are internally allocated. Leasing arrangements and/or nonarm’s-length methods of internal reallocation are all feasible, and some trading between cooperatives is allowed as well. Vessels that join the limited-access fishery vest their shares in a common pool that is available to all vessels in the limited-access fishery, similar to preA80 management. In addition, cooperatives are allocated shares of PSC TACs according to their holdings of target quota shares. In practice, vessels have primarily fished their own target and PSC allocations, although some stacking of quota on vessels has occurred and there are a significant number of exchanges of quota of one species for another as company owners determine which species they expect they will need or hold in excess. Third, at inception, vessels were required to retain and process at least a certain percentage of their catch into final products. This percentage increased annually and was intended to greatly reduce the high rate of discards prior to A80. Because of legal issues, this provision was subsequently eliminated in 2010 but was in place for the first two years immediately following A80 implementation.
The NPFMC also implemented Amendment 85 (NPFMC 2007b) concurrently with A80, with the purpose of allocating the TAC of BSAI cod across competing gears and sectors. Under this rule, the A80 sector received 13.4% of the TAC, with that allocation being divided between cooperatives and the limited access sectors. This allocation presented challenges to the A80 fishery for at least three reasons. First, this allocation is considerably lower than the historical harvest share of the A80 sector (Anderson and Concepcion 2011). Second, even in the absence of active targeting of cod, there is often significant incidental catch of cod involved in fishing for other major species such as yellowfin sole and rock sole (Figure 1). Third, the allocation of cod creates a “hard cap”: all fishing in the BSAI must cease for cooperative members if the cooperative exceeds its allocation. These factors have caused many within the A80 sector to substantially reduce their targeting of cod. In many cases this concern has risen to the point that many vessels now claim to avoid targeting this once-important target species (Abbott and Haynie 2012), even actively avoiding it as bycatch (Anderson and Concepcion 2011, 2012).
All of the 22 vessels that actively participated immediately prior to A80 implementation remained in the fishery. Initially, 16 vessels (from seven different companies) formed a single cooperative known as the Alaska Seafood Cooperative, while the remaining 6 vessels (five from a single company) elected to stay in the limited-access sector.3 In 2008, one company replaced a sunken vessel in the limited access sector, giving a total of 23 active vessels in the fishery since the early 2000s. There has also been some consolidation within Alaska Seafood Cooperative as two smaller vessels in the cooperative no longer actively participate in the fishery.
III. DATA
The primary data for our analysis are confidential observer data on the location and catch of each vessel from the North Pacific Groundfish Observer Program4 (NPGOP). Onboard observers record the deployment and retrieval location and times for every trawl, as well as additional information such as the total catch and tow depth. Observers also randomly select (without preannouncement) particular hauls for species composition sampling. Once a haul is selected, the contents are themselves randomly sampled to provide assessments of the weight or numbers of target and PSC catch. Before A80, vessels longer than 124 feet were required to carry an observer on all fishing days, while smaller vessels were only required to carry an observer on 30% of fishing days and had discretion over when to satisfy these coverage requirements. Since 2008, all vessels, regardless of length, are required to maintain “200% coverage,” meaning that two observers are onboard at all times, yielding species composition information for virtually every post-A80 haul. We therefore have an unusually thorough record of fishing activities and catch composition of all vessels larger than 124 feet both before and after A80 and all vessels after its implementation.5 The NPGOP is highly regarded and the data undergo an extensive auditing process before being used for fisheries management.
In addition to observer data, we also utilize complete vessel-level data on the production weight of final products for each target species, as well as estimates of the initial catch weight embodied in the final products. These data are self-reported to NMFS and were provided weekly prior to 2009 and daily thereafter. For information on product prices we rely on commercial operator annual reports from each catcher-processor. Finally, we supplement all of our data with data from the 2008 Amendment 80 Economic Data Report, a mandatory survey implemented after rationalization. We use it here for data on vessel attributes and to confirm commercial operator annual report revenue data.
IV. HALIBUT AND COD AVOIDANCE
As described above, the fishery was overwhelmingly constrained by its bycatch of Pacific halibut prior to A80 implementation, resulting in target species quota being left unharvested with the premature closure of many target fisheries. A significantly lower cod allocation under Amendment 85 also evoked concern that noncod target quota would be underutilized due to cod’s potential new role as a “choke” species. In this section, we set out to answer the following question: Have fishermen been effective at avoiding Pacific halibut and cod since the implementation of A80?6
Perhaps the clearest indicator of successful halibut and cod avoidance by fishermen postA80 is the fact that cooperative members have avoided reaching their collective allocations of halibut and cod in every year since A80 implementation. Indeed, halibut bycatch by the Alaska Seafood Cooperative was between 70% and 83% of the annual allocation, while cod catch ranged between 78% and 99% of its annual allocation. This has enabled the fleet to harvest the vast majority of its target quotas of yellowfin and rock sole in all years since the inception of A80. As a result, there has been a consistent increase in the number of vessel-days spent fishing per year, and the newfound availability of halibut PSC in the fall months as a result of bycatch savings earlier in the season has allowed for a much more robust late-season fishery for yellowfin sole than previously. Seasonal allocations of quota under the common-pool system–which focused fishing in pulses of high activity–were replaced under A80 by annual quotas for A80 target and prohibited species. Given this newfound flexibility, vessel owners were able to smooth their participation over the season. This elongation and smoothing of participation occurred as the intensity of fishing per day, as measured by actual trawling time per day, declined by approximately 30% for cooperative vessels, with total annual fishing hours falling below their pre-A80 values.7
We employ multiple statistical measures to investigate the quantitative impact of A80 on bycatch8 avoidance. We examine the annual mean catch rates of bycatch species (catch per hour of fishing) before and after A80 and compare this measure to the mean share of catch of bycatch species to total catch. Catch rates of halibut or cod (catch per unit effort [CPUE]) control for the intensity of fishing effort and are therefore conducive to comparing across years given the propensity for year-to-year variation in fishing effort. The catch of halibut or cod as a share of total catch (STC) provides a measure of bycatch that controls for annual variation in overall harvest levels. This is particularly important given the increased utilization of target species allocations after A80 and the year-to-year variation in target species TACs. STC also provides an inverse measure of “bycatch productivity,” from the perspective that bycatch is an input to the production process of harvesting target species. The relationship between CPUE and STC can be seen through the decomposition bycatch/hour = [total bycatch/total catch] × [total catch/ hour] → CPUE = STC × [total catch/hour]. Thus, each measure provides distinct information regarding bycatch avoidance. The complementary nature of these measures is reflected by the fact that it is possible for halibut CPUE to increase post-A80 even if halibut STC declines if overall productivity–as measured by total catch per hour–increases significantly.
These annual measures fail to capture any dynamic components of bycatch avoidance that take place within a season. Catch rates for halibut and cod vary substantially within any given year, with some parts of the year being more prone to high bycatch rates than others. Examining the evolution of catch rates within a season provides information regarding the distribution of bycatch avoidance within a year and how this distribution has changed since the implementation of A80. We investigate the intra-annual distribution of bycatch avoidance using weekly halibut and cod STC.
Changes in Annual Mean Catch Rates
Halibut bycatch rates–as measured by the annual mean of both CPUE and STC for the Bering Sea portion of the fishery (halibut bycatch is less significant in the Aleutian Islands)–are considerably lower for cooperative members post-A80 (Figure 2).9 A distinct break exists between 2007 and 2008, with mean halibut CPUE falling from 100 to 50 kg/ h and mean halibut STC falling from 1.5% to 0.8%. Both measures of halibut bycatch remain well below 2002–2007 levels after the policy change in spite of some rebound in halibut CPUE in 2009 and 2010. By contrast, both measures of halibut bycatch in the noncooperative portion of the fleet were essentially unchanged in 2008 and skyrocketed to historically high levels in 2009 and 2010. These patterns are highly robust to econometric methods that control for potential variations in the timing of fishing and variations in the intensity of participation by different vessels (see Figure A1 of the online appendix10). There is thus strong evidence of a structural shift in the halibut bycatch composition of cooperative vessels that cannot be explained by simple shifting of effort across the season or between vessels, and is not shared by noncooperative vessels.
Means and 95% Cluster-Robust Confidence Intervals of Daily, Vessel-Level Halibut and Cod Catch (Catch per Unit Effort [CPUE] and Share of Total Catch [STC]); Clusters Defined over Combinations of Vessel, Week, and Year
In the case of cod, cooperative members traditionally had mean CPUE and STC over twice those of noncooperative members in the Aleutian Islands (Figure 2). Prior to 2008, the cooperative fleet had traditionally divided its effort between actively targeting Atka mackerel (which typically has low cod bycatch) and cod in the Aleutian Islands, whereas the noncooperative portion of the fleet rarely participated in the latter fishery. In the Bering Sea, the noncooperative fleet had a much more consistent presence in the winter rock sole roe fishery, which often has high cod bycatch rates and where cod often serves as an important secondary target (Abbott and Haynie 2012). Nevertheless, the cooperative fleet traditionally had substantially higher mean CPUE and STC for cod in the Bering Sea (Figure 2). After A80 there is a strong and persistent reduction in the annual mean for both cod CPUE and STC in the Aleutians for the cooperative fleet, even though there is essentially no variation for noncooperative vessels.11 In the Bering Sea fishery, a dramatic reduction in cod CPUE and STC occurred after 2008, but with a partial rebound in CPUE in subsequent years. Given that cod can be retained, this rebound is economically desirable as cod catch did not rise to levels that restricted the fisheries for other species. Controlling for vessel and week of fishing suggests that the post-A80 average STC for cod in the Bering Sea has remained well below levels seen from 2002 to 2007 (Figure A212). By contrast, vessels outside the cooperative have seen post-A80 increases in cod by both metrics.
The sharp reduction in halibut and cod bycatch by the cooperative fleet cannot be explained by reductions in assessed halibut or cod biomass (Hare 2011; Witherell and Peterson 2011) in the Bering Sea over this interval (Figure 3).13 The total estimated halibut biomass, derived from fisheries-independent bottom-trawl survey data of the Eastern Bering Sea, decreased only slightly between 2007 and 2008 and then increased in subsequent years–a pattern consistent with the pattern of noncooperative bycatch in the final three years of the sample and the rebound in CPUE for cooperative vessels in 2009–2010. Similarly, while the estimated biomass of cod has declined from its recent peak in the early 2000s, this decline has been gradual and not indicative of the sharp changes observed in catch composition after 2007. Overall, the structural shift in halibut and cod STC for cooperative vessels runs counter to the contemporaneous shifts in biomass, suggesting that the mechanisms underlying the changes in STC are behavioral.
Stock Assessment Estimates of Biomass
Explanations of the sudden increase in cod avoidance on the basis of price also fall short. Cod prices rose steadily from 2002 to 2008, with nominal prices for the median vessel in 2002 of $1.59/lb for finished product rising to a 2008 price of $3.42/lb. These prices collapsed in 2009 to $1.91/lb. While potentially reinforcing incentives for cod avoidance, this price collapse does not adequately explain the significant changes since 2008. First, the cod price decrease reflected a decrease in fish demand with the global recession; some other Bering Sea species (most notably yellowfin sole) experienced similar price decreases, so the price decrease of cod alone does not explain the shift away from cod relative to other target species. Second, the price decrease for cod occurred in 2009, whereas the Alaska Seafood Cooperative adjustment away from cod occurred most dramatically in 2008, when the price of cod was at its peak. This strongly suggests that the avoidance of cod has its roots in a dramatic increase in its perceived shadow cost after A80 rather than in exogenous factors.
Changes in the Intra-annual Distribution of Catch Rates
The distinct break between 2007 and 2008 in the annual mean CPUE and STC suggests cooperative members have been successful at avoiding cod and halibut since the inception of A80, but reveals little about the seasonal nature of this avoidance. Figure 4 shows the intra-annual distribution of bycatch avoidance using weekly halibut and cod STC in the Bering Sea for cooperative members only. Apart from the fact that weekly STC for halibut fluctuates around a lower annual mean for postA80 years, the most notable change is the reduced variation through the season. In particular, there are fewer extreme fluctuations toward high bycatch rates, indicating a tightening of the distribution around the mean. This pattern is especially strong in the first months of the season. In the final months of the fishing year, the mean and variability of 2008–2010 STC levels are not notably distinct from pre-A80 levels.14 Reductions in mean halibut bycatch rates were driven by shrinkage of the upper tail of the bycatch rate distribution, with the most dramatic reductions occurring in the first part of the season and with only minor reductions noted in the final third of the season. Interviews with vessel owners and skippers indicate a conscious attempt to maintain a STC that will allow vessels to catch their targets throughout the year, knowing that there is uncertainty about how much halibut will be present on the grounds at different portions of the year and in different target fisheries. These patterns suggest that much of the purposeful halibut avoidance occurred in the first eight months of the year, especially in the winter fishery for rock sole. Given the uncertainty faced by fishermen in the initial years of this program, it is possible that this pattern of behavior reflects a high degree of caution in the initial months to avoid the prospect of having insufficient halibut to participate in the fall yellowfin sole fishery. By the time the fall fishery arrives, it becomes clear that cooperative members have sufficient halibut quota remaining to avoid premature closure of the fishery. Fishermen relaxing their avoidance measures in light of this knowledge may explain the similarities between preand post-A80 fall halibut STC.
Weekly Bering Sea Halibut and Cod Share of Total Catch (STC) for Cooperative Member Vessels
The most dramatic reductions in cod STC have occurred in the first two months of the season during the rock sole roe fishery (Figure 4). Significant reductions persist into the summer months, and, as with halibut, there is less apparent interweek variability in the proportion of cod caught. However, in the final third of the season a notable change in pattern occurs. In the first post-A80 year (2008), cod catch rates remained low, perhaps a sign of the encountered species mix or of caution in the use of cod in a new management regime with substantial a priori uncertainty. However, in 2009 and 2010 the catch rates of cod increased dramatically in the fall fishery, substantially exceeding pre-A80 rates. The overall pattern, as with halibut, is one of “cod saving” early in the season through avoidance of a targeted fishery in the Aleutians and less secondary catch of cod in winter/spring flatfish fisheries, followed by a relaxation and even increased targeting of cod later in the season.15 Because Pacific cod can be retained and marketed, it is not surprising that we see specific targeting in some fall months when the cod are no longer needed to enable the harvest of other species.
While weekly variability in both cod and halibut STC clearly declined post-A80, there were weeks during post-A80 years where particular vessels struggled to find cod catch rates that would allow them to catch their quota through the year. Between years, there have been significant differences in the average cod catch rates per month, making it reasonable for vessels to assume that coming months could potentially resemble the highest rates encountered in previous years. The pattern of fishing we find is broadly consistent with a fishing strategy in which both cod and halibut are seen as potentially binding to one’s fishing prospects. While possible under secure individual multispecies catch shares, such a pattern was not incentive-compatible under the previous common-pool system.
V. THE MULTIPLE MARGINS OF SELECTIVITY
The previous section established that there were substantial and unprecedented changes in the composition of catch after multispecies catch rights were vested in the hands of individual vessel owners. A natural question in light of these changes is “how?” How were fishermen able to so dramatically alter the composition of undesirable species in their catch without major structural changes in gear technology? The answer to this question is important well beyond the context of this case study as it can help to provide insight on the interface between gear technology, the spatiotemporal distribution of species, and the incentives for or against selectivity inherent in the management system. Such an understanding can help to shed light on which features of the technical, biological, and economic systems may permit flexible substitution across species–potential that may be obscured by institutional factors.
Our investigation of how fishermen were able to alter their catch compositions focuses on three margins of behavioral change: (1) large-scale adjustments of fishing grounds, (2) short-term movements away from ephemeral bycatch hot spots, and (3) reductions in nightfishing.16 A theme that emerges from these margins is the critical nature of the “when” and “where” decisions of fishing. Our findings suggest that the presence of spatiotemporal heterogeneity of species distributions and relative catchabilities may allow for substantial flexibility on the scale of days, weeks, or a season–even as the fishing technology, conditional on choice of spatiotemporal context, samples from “whatever is there” according to fixed proportions.
Large-Scale Spatial Adjustments
Broad spatial shifts in fishing effort across fishing grounds may exert a significant effect on the composition of catch in this fishery. For instance, the behavioral effects of a largescale spatial closure in the mid-1990s shifted the secondary catch composition of the winter rock sole fishery away from red king crab and toward increased halibut and cod (Abbott and Haynie 2012). The first two panels of Figure 5 present spatial kernel densities of fishing trawls in the January-April fishery for cooperative members during the 2005–2007 and 2008–2010 intervals, while the third panel differences these densities to reveal areas that saw relative increases or declines in the postA80 era.17 A notable feature of the fishery at this time of the year is the intense concentration of effort to the south and west of the Red King Crab Savings area (RKCSA), with roughly two-thirds of effort lying in this relatively small area. Much of this effort occurs in the first few weeks of the season during the rock sole roe fishery, with well over 90% of trawls in the first seven weeks of the season lying within this region both before and after 2008.18 Subsequent fishing activity in March and April tends to be more spatially diffuse, reflecting changes in target species (often toward yellowfin sole).
Distribution of January-April Fishing Effort for Cooperative Vessels in 2005-2007 and 2008–2010 (top row) in Comparison to Areas of High and Low Catch Rates of Cod and Halibut (second and third rows)
The most striking change in the distribution of effort after A80 is the almost complete forsaking of the extreme southwestern grounds and accompanying consolidation of effort in areas just south of the RKCSA and in the Special Savings Area of the RKCSA. The fourth and fifth panels of Figure 5 show kernel densities of cod and halibut STC for 2005-2007, and the final two panels repeat this information for 2008–2010.19 These figures show that the northerly and slightly eastward shift of the cooperative fleet during this period avoided significant and temporally persistent “hot spots” for both halibut and cod–a partial reversal of the southwesterly shift of effort that increased both halibut and cod STC in the wake of the RKCSA closure in 1995 (Abbott and Haynie 2012).20
Large-scale avoidance behavior of cod and halibut is also evident in the midseason from May to August (Figure 6). Effort shifted dramatically out of historically codand halibutrich waters in the southeastern grounds, and especially cod-rich (and to a lesser degree halibut-rich) waters in the far northwestern grounds, and toward more consolidated and central grounds to the north and west of the RKCSA. These grounds had low to moderate levels of cod and halibut bycatch in pre-A80 years that largely persisted after 2008. Visual comparison of the patterns of fishing effort shifts and halibut/cod hot spots during the September-December period (Figure A521) shows no consistent evidence of either halibut or cod avoidance in the broad spatial choices of vessel captains.
Distribution of May-August Fishing Effort for Cooperative Vessels in 2005-2007 and 2008–2010 (top row) in Comparison to Areas of High and Low Catch Rates of Cod and Halibut (second and third rows)
Avoidance Behavior
Even after adjusting their large-scale fishing grounds, vessels may come upon bycatch hot spots that are ephemeral in nature, reflecting fine-scale fluctuations in phenomena such as currents, temperature, sea ice, or predator/ prey mobility. Significant gains in bycatch avoidance may be achieved by vessels moving relatively short distances after encountering such a hot spot. However, this may be costly in a common-pool fishery (Abbott and Wilen 2009), and prior attempts to institute voluntary spatial avoidance of bycatch hot spots in this fishery have had little success (Abbott and Wilen 2010).
We examine whether A80 increased the propensity of vessels to make significant movements–measured by the distance between the retrieval and deployment locations of subsequent hauls–when the previous haul contained a large share of halibut.22 In general, interhaul movements are small (Figure A623), with a median distance of just 1.5 nautical miles (NM) and 25% of hauls less than 1 NM apart. Interspersed among these shortdistance hauls are movements of 10 NM or less, with distances of 20 NM or more occurring only about 5% of the time. Measuring changes in the direction of hauls (Figure A724) reveals a strongly bimodal pattern, with trawls that roughly retrace the bearing of the first trawl (~ 180°) about three times more common than hauls that continue on the previous trajectory (~ 0¤).25 This portrays a dominant pattern of fishing in which vessels repeatedly trawl a ground, retracing the path of previous trawls after moving a short distance. These “mining” events are interspersed with periodic shifts to new grounds or searching behavior, where these shifts often occur over relatively short distances.
Overall, we find convincing evidence that A80 had a positive impact on the relationship between interhaul movements and the occurrence of high halibut STC in previous hauls for members of the cooperative. A cross-tabulation of movement events–defined as an interhaul movement greater than 3 NM–versus halibut STC in the previous haul (Table 1) displays only a negligible relationship between halibut concentration and movement probability prior to A80 implementation. In contrast, post-A80 years display a progressively higher share of movements as the halibut share increases, so that the probability of movement increases by a factor of 2 for the highest bycatch and 1.5 for intermediate rates.
Tabulation of the Decision to Move (Assuming a Threshold of > 3 NM) or Not for Cooperative Member Vessels in the Bering Sea Fishery versus Halibut Share of Total Catch of the Previous Haul
The pattern of increased movement probabilities after large halibut bycatch events in post-A80 years is also robust to a more rigorous econometric analysis. We estimate the probability of moving after the realization of a halibut bycatch “event,” while controlling for an extensive set of variables that might influence the movement decisions of fishermen.26 We estimate multiple linear probability models (using heteroskedasticity robust standard errors) that differ only by the threshold used to define a movement (1, 3, 5, or 7 NM). Bycatch “events” are categorical variables indicating whether halibut STC lies within particular bounds. We define categories of 2%5% (moderate), 5%–10% (high), and > 10% (very high) halibut (where hauls with 0%–2% halibut are the omitted base category) and interact these indicator variables with annual dummy variables (with 2007 as the base year).27 Estimation results for cooperative members (Figure 7) demonstrate a significant increase in the probability of movement in response to extreme halibut bycatch after A80–an increase of probability of between 0.22 and 0.24 (p < 0.0003) at the 3 NM threshold for halibut bycatch rates > 10%. These shifts are more muted for lower levels of halibut bycatch, with no significant shifts in behavior in the 5%–10% range and in-creases in the probability of 3 NM movements for the 2%–5% halibut range of 5.5%–7.8% (< 0.025). These increases in probability are in addition to estimated increases in movement for hauls with zero or very low levels of halibut. In contrast to these large changes, we find no discernible increase or deviation from historical bycatch avoidance patterns for noncooperative vessels. Estimates of movement probabilities analogous to those in Figure 7 for noncooperative vessels display an increase for the baseline level of halibut only (Figure A828). This lack of increase in avoidance behavior is likely explained by the fact that halibut remained a common-pool resource after 2008 for this portion of the fleet, reducing the private benefit of costly evasive maneuvers to these vessels relative to cooperative members that have individual allocations of halibut quota.
Estimated Changes in Movement Probabilities (Base Year = 2007) for Cooperative Vessels across Different Distance Thresholds, Conditional on Halibut Percentage in the Previous Haul
We also find evidence of seasonality in the bycatch-induced movement of cooperative members, which closely aligns with the analysis of large-scale movements earlier in this section. Estimating the linear probability models above using four-month subseasons reveals a notable increase in the avoidance of halibut in the 5%–10% and > 10% ranges between January and April (Figure A929). In contrast, increased avoidance of halibut is limited to only the > 10% range between May and August, with no apparent increases in reactions to 5%–10% bycatch events and mild reactions to halibut bycatch in the 2%–5% range (Figure A1030). Finally, there is little statistical evidence for any A80-driven changes in reactionary avoidance in the final four months of the year (Figure A1131).32
To what extent does a bycatch-induced movement result in a reduced bycatch rate in the subsequent haul? To answer this question, we estimate a regression of the change in the halibut STC per haul conditional on indicator variables for three increasing ranges of movement. By estimating this regression on samples restricted to cases where the initial catch rate of halibut is in the 2%–5%, 5%–10%, or > 10% range, we estimate the mean effect of movement conditional on the severity of previous bycatch. Estimates for these regressions reveal strong statistical evidence that movements of 3 NM or more following a high bycatch event (5%–10% and > 10%) can have a significant quantitative effect on subsequent bycatch (Table 2). A movement of at least 3 NM after a bycatch event in the 5%–10% range reduces halibut STC by an extra 0.01 (a 10% to 20% reduction in STC), with no evidence that longer movements lead to lower bycatch rates. Movements between 3 and 7 NM after a bycatch event in the > 10% range reduce halibut STC by 0.025, while even longer movements yield a mean STC reduction of approximately 0.05.33 These estimated reductions are robust to the inclusion of a full set of vessel, spatial, and temporal fixed effects.
Estimates of the Change in Halibut Share of Total Catch (STC) for Movements of Different Distances and Conditional on Halibut STC in the Previous Haul for Cooperative Members
Reductions in Night-fishing
There is well-established scientific evidence that many demersal species exhibit diel patterns of behavior (e.g., foraging) or differential reactions to approaching gear that make them more or less vulnerable to trawl gear at night (e.g., Casey and Myers 1998; Ryer 2008), providing some rationale for shifting fishing effort away from nighttime hours to avoid harvesting species that are more susceptible at night. To investigate the salience of this phenomenon in our context, we examine the extent to which A80 impacted night-fishing activity and estimate its implied effect on halibut bycatch reduction.
Overall, we find evidence of a significant shift away from night-fishing for cooperative members since the inception of A80. For our investigation, we calculate the duration of dayversus night-fishing in each haul utilizing trawl deployment/retrieval times and locations, in conjunction with estimates of the implied length of day and sunrise and sunset times for each haul (Forsythe et al. 1995).34 Since any reductions in night-fishing occurred in a context where the overall number of fishing hours was declining, regardless of the time of day, we focus on changes in the share of night-fishing over time. Figure 8 graphs the share of night-fishing hours by week in the Bering Sea for the cooperative fleet, where the bold, black broken line reflects the natural proportion of nighttime hours in that week.35
Proportion of Weekly Fishing in Nighttime Hours by Year for Cooperative Vessels in the Bering Sea
Aside from the pronounced seasonality of daylight hours in this northern fishery, another fairly robust feature is for effort in all years to skew toward daylight hours. This tendency becomes noticeably more pronounced in the years following A80, with particularly large reductions in night-fishing occurring in the first half of the season.
This pattern of reduced night-fishing in post-A80 years is also robust to a more rigorous econometric analysis. We estimate the change in the proportion of nighttime fishing hours for each vessel-day using the fractional logit estimator (Papke and Wooldridge 1996) while controlling for seasonality, vessel effects, and the natural cyclicality of the proportion of nighttime hours to daylight hours. The fractional logit estimator models the expected proportion of nighttime hauls using a logit link function, effectively estimating a binomial logit model with a continuous lefthand side variable.36 The odds ratios for the annual dummy variables in this regression– which are equal to the proportional change in the ratio of nighttime to daytime fishing hours relative to 2007–are remarkably constant across all pre-A80 years (Table 3). From 2008 on, however, the relative odds of night-fishing fell to between 76% and 78% of 2007 levels. The average marginal effects of the 2008-2010 dummy variables–which equal the change in the predicted proportion of nightfishing for all observations in a given year from switching its annual dummy variable “on” and “off”–reveal highly significant reductions of about 0.05 in the proportion of night-fishing hours in a day, equivalent to a decrease of between 15% and 18% relative to the proportion of night-fishing hours in 2007.
Odds Ratios and Average Marginal Effects of Annual Dummy Variables (Base = 2007) from a Fractional Logit Regression of the Proportion of Daily Fishing Hours Prosecuted at Night in the Bering Sea by Cooperative Vessels
There is also a pronounced seasonality to the reduction in night-fishing. Estimates conducted on four-month subsets of the data suggest a consistent move away from night-fishing in the first eight months of the year, followed by a fall fishery in which the diurnal pattern of fishing is statistically indistinguishable from previous years. This pattern is behaviorally consistent with the lack of spatial avoidance of halibut observed at the end of the season.
We investigate the possible rationale for this shift in the temporal distribution of effort by comparing catch rates for hauls based upon the proportion of time spent fishing during the day versus night. In doing so, we must exercise caution since other factors that influence catch rates are not randomized over the time of day. The desired mix of particular species may influence the timing of fishing. Fishing location varies across the season and is therefore correlated with day length. A single vessel fishing in a given location in two different parts of the year may face very different distributions of species abundance due to temporal instabilities or, if compared over similar time frames, may nevertheless experience substantially different catch patterns due to different targeting strategies or fishing skill. To minimize these biases, it is ideal to compare catch rates across hauls for a given vessel across narrow intervals of time and space but with different mixes of day- and night-fishing. To do this, we posit the following relationship for the catch for a given species for a vessel i, a spatial unit s, a particular fishing day t, and on any particular haul h:
[1]
cist is a vessel/area/day effect (unobserved heterogeneity in catchability) that may be correlated with the haul-varying variables xisth or PctNightisth the proportion of Durationisth fished at night.37 Estimation can be conditioned on cist by using the fixed-effects Poisson estimator (Cameron and Trivedi 2005; Hausman, Hall, and Griliches 1984), which models the conditional likelihood of the distribution of total catch across multiple hauls within a vessel/area/day combination, exploiting “within” variation in the percentage of night-fishing between hauls for identification. To implement the estimator, we set the spatial unit to the ADFG STAT6 zones and include average trawl depth and its square as the elements of xisth .
The estimates reported in Table 4 are consistent with observations from previous fish behavioral studies and experimental flatfish trawls. The exponentiated γ coefficients (incidence rates) measure the proportional expansion/contraction of the catch rate–conditional on duration–from shifting a haul entirely from dayto night-fishing. The incidence rates reported in Table 4 therefore indicate significant reductions in CPUE at night for all target species.38 Halibut, on the other hand, experiences a CPUE increase of 16% at night, a pattern consistent with the literature and fishermen’s observations (Ryer, Rose, and Iseri 2010). The estimates derived from seasonal subsamples reveal some limited amount of seasonal variability for target species. However, the dominant finding is that halibut bycatch rates are 64% higher at night during the first four months of the fishery, but with no significant diel pattern later in the season.39
Incidence Rates of Proportion of Night-fishing Hours on Catch per Hour for Cooperative Vessels in the Bering Sea Fishery for 2008-2010
Conversations with vessel captains suggest that the rationale for reducing night-fishing is driven by a combination of halibut avoidance and as a matter of efficient time allocation. As an integrated catching and production unit, vessels have some ability to allocate necessary fishing downtime (e.g., time spent in search or transit or time for the factory to catch up in its production) to less productive (and less halibut intensive) hours. Ceteris paribus, it is certainly more cost-effective to shift effort to the daytime, where catch rates per hour are 25% higher overall. Reductions in halibut bycatch that come with daytime fishing also provide a strong incentive to reduce night-fishing in the winter and early spring. By taking the ratio of incidence rates we find that the nighttime ratios of rock sole and yellowfin sole to halibut bycatch are only 43% and 55% of their daytime values, so that a unit of halibut bycatch yields roughly twice the revenue during the day compared to night. Vessel captains have specifically stated that they have an objective of filling up the holding tanks onboard their vessels before nightfall so as to maintain factory productivity through the night while minimizing fishing in less productive and high-bycatch nighttime conditions. The fact that reduced night-fishing is not present in the fall fishery when the diel gradient for halibut disappears–even as nighttime target catch rates remain persistently lower at night–and when other measurable halibut avoidance behaviors are also eroding strongly suggests that halibut avoidance plays a dominant role in the allocation of effort across the day.
By providing fishermen strong individual incentives for bycatch avoidance, A80 drove fishermen to exploit previously underutilized substitution possibilities in their daily allocation of fishing time to reduce their bycatch. Given the magnitude of the diel effects on catch and bycatch rates, one may question why the temporal substitution is not even greater. The answer to this question may lie in the inherent limits to substitutability provided by temporal and technical constraints. Given the integrated nature of production in a factory trawler, it is important to coordinate fishing and processing activities to maintain an optimal throughput of product, and there is limited capacity (and negative implications for product quality) to store daytime catch for processing at night. Furthermore, the fixed daily allocation of daylight hours (which are particularly scarce in the winter/spring fishery) implies that only so much effort can be shifted within a day before further avoidance of night-fishing implies a reduction in daily fishing hours. While we do observe substantial reductions in fishing hours per day, this reduction in intensity does come at a cost, necessitating longer trips and seasons. Some smaller vessels cannot store enough fish onboard to continue to process through the night, so avoiding night-fishing for these vessels would be especially costly. One final counterincentive to reducing night-fishing may lie in the fact that avoidance of cod is actually facilitated by night-fishing, with ratios of rock sole and yellowfin sole to cod being 70% and 176% higher at night. Thus, the fact that fishermen are concerned about the avoidance of more than one species in a multispecies setting may limit the efficacy of a particular margin of avoidance when that margin operates in qualitatively different directions for distinct bycatch species.
VI. DISCUSSION
We have demonstrated that fishermen operating in a complex multispecies fishery were able to alter the composition of their catch away from species that were anticipated to constrain their fishing opportunities over the season. We also provided a detailed analysis of the operation-level decisions that led to these reductions. It is notable that these margins involved changes in the behavior of fishing decision makers rather than the adoption of new gear. By proactively varying the spatial and temporal aspects of gear deployment–the “when” and “where” of fishing– and by increasing the spatial reaction to negative signals from the environment, captains were able to substitute away from halibut and cod and increase the share of valued species in their catch.
The overwhelmingly behavioral as opposed to technical nature of short-run adaptations to the new incentives came with one exception: the adoption of halibut excluder devices under certain fishing conditions by the cooperative fleet in 2008 and beyond. These devices utilize differences in size and behavior to deflect halibut through an escape hatch on top of the net, while the majority of target species swim through holes in the deflection device and are caught. Unfortunately, we have no reliable data on whether an excluder was used in any particular instance. However, interviews with multiple captains suggest that the adoption of excluder gear is not yet widespread, with only a couple of vessels using the gear consistently. Captains have provided two arguments for this slow adoption. First, some captains maintain that while the halibut CPUE is lowered by excluders, the attendant reductions in target CPUE result in more muted improvements in halibut per unit of target catch and revenues, reducing incentives for adoption. Second, switching between conventional and excluder gear involves a nontrivial cost, making occasional use of excluders unattractive. As a result, while most vessels own excluder gear, they have mostly limited its use to particularly halibut-rich fisheries–such as those for flathead sole and arrowtooth flounder–instead choosing to alter their behavior along the more cost-effective margins of space and time. Nevertheless, it is telling and completely consistent with the incentives faced in the fishery that this technology was developed in a cooperative venture between industry and National Oceanic and Atmospheric Administration personnel in the late 1990s and yet was not adopted in any significant way until after A80 in 2008 (Gauvin and Rose 2000; Rose and Gauvin 2000). It was not until vessels faced individual bycatch constraints that vessel operators found it in their interest to utilize the new technology, despite its widespread availability. To do otherwise would have meant bearing the full cost of lost target catch and the direct cost and inconvenience of the gear modification itself, while receiving only a small fraction of the benefits. Viewed in this way, even this “technological” aspect of bycatch avoidance was ultimately driven by operators’ behavior in adopting available technologies.40
Our findings suggest that vessel operators have been able to skillfully adjust their usage of what is commonly viewed as a nonselective technology. Yet, this substitution potential was latent until changes in management institutions altered the incentives of fishermen, causing them to internalize the costs of fishing less selectively. Importantly, many if not all of the margins of selectivity exercised by the fleet were common knowledge before 2008 but were rarely used. As early as the mid 1990s the fleet instituted a voluntary bycatch avoidance program that pooled spatiotemporal information across vessels to facilitate avoidance of bycatch hot spots, but an evaluation of this program found no effect on either spatial behavior or halibut bycatch rates (Abbott and Wilen 2010). Examinations of the implicit shadow cost of halibut revealed by the spatial choices of vessels in the decades prior to A80 show avoidance that falls well short of the fleetwide optimum (Abbott and Wilen 2011; Haynie, Hicks, and Schnier 2009). This highlights that while informationbased strategies for enhancing selective fishing (e.g., Gilman, Dalzell, and Martin 2006) may have merit, the mere possession of information may not be sufficient to induce significant changes in catch composition if fishermen’s incentives are not also properly aligned.
While our results indicate that vessel operators are decidedly flexible in adjusting their species mix using nonselective fishing gear, realized catch composition remains the outcome of a partially controlled random variable, the distribution of which is driven by the underlying mixture of stocks. Despite the added flexibility created by A80, fishermen continue to operate in a fishery constrained by environmental, regulatory, and technological factors, thereby making incidental catch of nontargeted species a necessary component of catch. The TAC-setting process under the BSAI Fishery Management Plan does not guarantee that quotas will be perfectly balanced with the underlying biomass composition, so that matching quotas with catch continues to be a challenge. This suggests that catch-quota balancing mechanisms such as those described by Sanchirico et al. (2006) may still play an important role in optimal prosecution of multispecies fisheries.41
Finally, our results highlight the difficulty in assessing the potential for cross-species substitution in fisheries on the basis of ex ante data alone. While some (Pascoe, Koundouri, and Bjørndal 2007; Pascoe, Punt, and Dichmont 2010; Squires et al. 1998; Squires and Kirkley 1991, 1996) have used such data in the context of primal or dual production economic models to argue that policy interventions such as multispecies ITQs may fail, these techniques depend on the combination of technology, biological conditions, and behaviors observed in the data to estimate the parameters on which these inferences rely. Since these data are often generated under radically different incentive structures (i.e., different species-specific shadow prices) than the prospective system, a purely empirical approach superimposed upon a flexible functional form for a production, cost, or revenue function is likely to reflect far more about the incentives for substitutability under the status quo regime than the technological possibilities under a potential new management regime. Depending on the behavior of fishermen under status quo institutions, many possible catch combinations may be unknown prior to the management change simply because fishermen have never had the incentive to seek out those catch combinations.
Generating more robust ex ante predictions of policy interventions will likely require that greater care be given to modeling fishing as a process–by using information derived from microdata and in-depth conversations with fishermen to understand the fundamental decisions that drive the quantity and composition of catch and their outcomes in terms of expected catch, revenues, and costs. In many cases these decisions, such as choices of when and where to fish, may map only imperfectly to the notion of conventional inputs (such as fishing hours or measures of crew labor) often used in production economic models. By nesting these decisions in a structural economic (i.e., optimization) model it is possible to simulate the impacts of changes in management policies on system outcomes, including changes in species mix, although the effectiveness of these predictions may vary. There are strong precedents for such modeling in fisheries economics (e.g., Haynie and Layton 2010; Reimer, Abbott, and Wilen 2014; Smith and Wilen 2003). Such an approach will demand more of researchers in terms of the need for significantly more detailed microbehavioral data focusing on the pertinent margins of fishermen’s behavior than is typically gathered, as well as a willingness to impose potentially strong assumptions on the structural decision process of fishermen. However, these difficulties merely reflect the inherent difficulties associated with a modeling approach that addresses what policy makers actually desire: a prediction of the state of the fishery under a new set of incentives.
Acknowledgments
We thank Ron Felthoven, Steve Kasperski, Jim Wilen, and seminar participants at Arizona State University, the 2011 PERC Workshop on Lessons Learned in Rights-based Fisheries Management, the 2011 AFS Annual Meeting, the 2011 NAAFE Forum, and the AERE 2013 Summer Conference for their valuable feedback. We are grateful to Jason Anderson of the Alaska Seafood Cooperative, as well as John Gauvin, Bill Orr, Dave Wood, and Robert Hezel for their invaluable information on the A80 fleet. Part of this work was completed while Matthew Reimer was a predoctoral fellow at the Center for Environmental Economics and Sustainability Policy at Arizona State University. The findings and conclusions in the paper are those of the authors and do not necessarily represent the views of the National Marine Fisheries Service. This work was funded by NOAA Award NA10NMF4370286.
Footnotes
The authors are, respectively, associate professor, School of Sustainability and Center for Environmental Economics and Sustainability Policy, Arizona State University, Tempe; economist, Resource Ecology and Fisheries Management Division, Alaska Fisheries Science Center, National Marine Fisheries Service, NOAA, Seattle, Washington; and assistant professor, Institute of Social and Economic Research, Department of Economics, University of Alaska–Anchorage.
↵1 Throughout this paper we refer to “selectivity” as between species rather than between different age or size classes within a species. However, many of the same arguments would also apply in these cases.
↵2 We do not mean that fisheries scientists assume that these parameters never vary. Stock assessment methods may allow for time-varying estimates of these parameters.
↵3 Prior to 2010, the Alaska Seafood Cooperative was known as the Best Use Coalition. In 2011, all vessels in the limited-access sector decided to form a second cooperative under the name of the Alaska Groundfish Cooperative, eliminating the limited-access sector.
↵4 See https://alaskafisheries.noaa.gov/sustainablefisheries/observers/.
↵5 Over 62% of pre-A80 hauls provide species-specific catch data, and over 99% of post-A80 hauls contain this information.
↵6 We limit our analysis of prohibited species to Pacific halibut since the PSC quota for three crab species has remained comparatively slack. Discussions with fishermen and comparisons of seasonal crab catch with quota allocations to the A80 sector confirm that while fishermen may try to avoid excessive crab bycatch, it is of secondary concern and has not posed a serious risk of restricting the target fishery.
↵7 See the online appendix (http://le.uwpress.org) for a detailed analysis of the impacts of A80 on the temporal distribution of fishing effort.
↵8 For symmetry of terminology, we often refer to cod as “bycatch” even though it can be retained and sold. Its status as a target versus a bycatch species is inherently “fuzzy” and is endogenous within the season and vessel specific.
↵9 All figures in this section use the full sample of vessels, including vessels without full observer coverage pre-A80. While this could introduce bias into our estimates, limiting the sample to only fully covered vessels made no notable difference.
↵10 The online appendix is available at http://le.uwpress.org.
↵11 This reduction is closely related to the shift of three cooperative vessels–each of which had very high historical proportions of cod and very little Atka mackerel–from the Aleutian fishery to the Bering Sea, confirming anecdotal information from cooperative members that targeting of cod has been drastically reduced, with cod quota being preserved for use as bycatch in the Bering Sea fishery.
↵12 Available at http://le.uwpress.org.
↵13 Variations in oceanographic conditions, notably bottom temperature and the size of the Bering Sea “cold pool,” have a notable effect on the distribution of halibut and other groundfish species. However, the expansion of the cold pool that occurred post-A80 is predicted to have shifted the center of gravity of halibut biomass to the southeast, increasing the susceptibility of halibut to the fleet (Kotwicki and Lauth 2013).
↵14 Estimates of a post-A80 indicator variable from a spectrum of quantile regressions estimated on four-month subsets of the data with vessel and month fixed effects confirm these patterns (Figure A3, available at http://le.uwpress.org).
↵15 As with halibut, these descriptive findings are borne out by seasonal quantile regressions with vessel and seasonal controls (Figure A4, available at http://le.uwpress.org).
↵16 Our treatment of the multiple margins of behavioral adjustment is inherently selective. We limit our focus to the margins that were highlighted in interviews with industry personnel as salient, could be investigated using existing data sources, and were found to be empirically important upon closer investigation.
↵17 Black (white) areas indicate areas of greatest (lowest) density of fishing. The third panel is the difference of these densities, where black (white) indicates areas with increases (decreases) in relative effort in 2008–2010 and medium gray indicates areas with no change. These kernel densities are calculated using the Spatial Analyst Toolbox in ESRI ArcGIS (ESRI 2010). A quadratic kernel function with a bandwidth of 5 km was selected with results saved to a raster with a pixel size of 2.5 km2.
↵18 Comparisons for the noncooperative fleet are not possible for this part of the season as most noncooperative vessels are pursuing Atka mackerel in the Aleutians at this time.
↵19 White (black) shading indicates areas of low (high) cod or halibut STC, while gray areas indicate moderate cod or halibut STC. The use of gray backgrounds is designed to emphasize visualization of high and low values within the constraints of grayscale figures. Densities are calculated with the Spatial Analyst Toolbox in ESRI ArcGIS (ESRI 2010) using an inverse distance weighting (IDW) technique with a power of two and a search radius of 5 km, with results saved to a raster with a pixel size of 2.5 km2.
↵20 There is also a notable shift of effort post-A80 away from the extreme northerly and northwesterly grounds. While some of these grounds are notable for high cod catch rates, they are generally low in halibut. The primary driver of this shift in effort, however, seems to be driven less by bycatch avoidance and more by the unusually thick and early ice cover in these northern grounds in recent years (Stabeno et al. 2012), combined with the elongation of the southern rock sole season due to lower halibut bycatch rates.
↵21 Available at http://le.uwpress.org.
↵22 The data in this section are cleaned so that measures of distance are between truly subsequent fishing hauls within Bering Sea trips, thus eliminating hauls interrupted by a nonfishing event (weather delay, mechanical problem, crew injuries, etc.), trips to port, or cases in which subsequent trawls are split between the Bering Sea and Aleutian fisheries. We also eliminate observations with unusually long time lags between trawls.
↵23 Available at http://le.uwpress.org.
↵24 Available at http://le.uwpress.org.
↵25 Bearings are calculated using the arctangent of the change in the deployment and retrieval longitude and latitude coordinates, thereby assuming that vessels follow paths that are the shortest distance on a plane.
↵26 These include monthly measures of fuel prices, vessel and month fixed effects, total catch per hour and catch per hour of yellowfin sole and rock sole (the dominant target species) in the previous haul, and an indicator of whether the previous haul was conducted at night. We also include area-specific (STAT6) fixed effects for the previous haul. The STAT6 zones are one-half degree of longitude by one-half degree of latitude (roughly 60 km on each side) and have a regular grid shape except for around coastal areas where the area and topology of zones is less regular.
↵27 The distribution of halibut concentration in a haul has a strongly positive skew (Table 1), with a significant proportion of very large values despite a median bycatch rate of well below 1% in pre-A80 years. Indeed, 10% of hauls account for 52% and 58% of total halibut bycatch preand post-A80, respectively.
↵28 Available at http://le.uwpress.org.
↵29 Available at http://le.uwpress.org.
↵30 Available at http://le.uwpress.org.
↵31 Available at http://le.uwpress.org.
↵32 Statements about the statistical significance of reactions to the more extreme bycatch events must balance the fact that such events are often quite rare within a given fourmonth interval, yielding imprecise estimates.
↵33 These reductions are on top of the reductions that would have naturally occurred from merely staying put. The intercept term in the “no controls” model provides an estimate of this effect (the analog for the “full controls” specification has no such simple interpretation). The negative sign indicates that high bycatch events are “rare events,” even conditional on a point in space and time, so that a high bycatch event is likely to be followed by a period of reversion to the mean.
↵34 Our definition of night is after “civil twilight,” when the center of the sun is six degrees below the horizon, a value that seems to have some credibility in its linkage to the behavior of fish (Steve Barbeaux, NOAA fisheries biologist, personal communication, May 4, 2011).
↵35 Natural day length varies slightly during a given week and also because vessels operate at different locations. We calculate the proportion of daylight hours in a 24-hour period for all observed hauls in 2008 and then present the weekly mean.
↵36 This estimator has a number of advantages. First, unlike linear models, it predicts strictly inside the [0,1] interval. Second, it handles proportions that are either zero or one, unlike techniques that model the log-odds. Third, it allows us to precisely control for seasonal variations in night length by using night length directly as an “offset” variable (so that a doubling of nighttime hours doubles the expected proportion of nighttime hauls). Finally, as a member of the linear exponential family, the maximum likelihood estimates of the parameters of the conditional mean are consistent regardless of failure of the underlying distribution (Gourieroux, Monfort, and Trognon 1984). Since our estimates are quasi-maximum likelihood estimates, we replace the usual maximum likelihood estimate standard errors with “robust” estimates clustered on the intersection of vessel, year, and week to control for serial correlation and heteroskedasticity within these clusters.
↵37 Note that if a =1 then trawl duration is an “exposure” variable and functions like “effort” in a Schaefer production function.
↵38 Pacific cod, as a gadid, is likely less available to bottom trawl gear at night due to vertical migration (Ryer, Rose, and Iseri 2010). The nighttime reductions in catch rates for flatfish are consistent with evidence from fish behavioral studies and experimental trawls using real-world flatfish trawl gear and are likely driven by a reduced nighttime “herding” behavioral response to the long sweeps frequently used in flatfish trawl gear (Ryer and Barnett 2006; Ryer, Rose, and Iseri 2010). See Rose, Gauvin, and Hammond (2010) for an illustration of flatfish trawl gear.
↵39 The reason for this pattern is unclear, but it might lie in the heterogeneous seasonal spatial distribution of different age/size classes of halibut in the Bering Sea (Adlerstein and Trumble 1998) and variations in availability and catchability of halibut according to age and size (Ryer and Barnett 2006; Ryer, Rose, and Iseri 2010).
↵40 Technical change has not been limited to adoption of existing technologies. More recently, cooperative research has led to the testing and widespread adoption in 2011 of trawl gear to minimize contact with the seafloor, minimizing habitat damage (Rose, Gauvin, and Hammond 2010). Whether this R&D and rapid adoption would have been forthcoming outside of the post-A80 incentives faced by the industry is unclear.
↵41 In 2013, NPFMC passed a “Bering Sea Flatfish Harvest Specifications Flexibility” action that will allocate the difference between acceptable biological catch (ABC) and TAC for flathead sole, rock sole, and/or yellowfin sole among A80 entities. This flexibility measure will allow A80 entities to exchange quota share of the TAC of one species for an equivalent amount of another species from the ABC surplus as vessels encounter more or less of different species with different degrees of bycatch. The action is intended to increase the opportunity of maximizing the harvest of and revenue from these three species while staying below their prescribed ABCs (NPFMC 2013a).
