Spillovers in Regional Fisheries Management: Do Catch Shares Cause Leakage?

I. INTRODUCTION

U.S. commercial fisheries conducted in federal waters are managed by regional fishery management councils that represent collections of states. An individual fishery’s assignment to a particular management council generally reflects the species range, its proximity to the political boundaries of participating states, and home ports of participating fishermen. However, species ranges can span management borders and ultimately link together fishermen whose home ports are in different regions. The result is that individual fishermen can participate in multiple fisheries managed by different councils. This institutional arrangement raises the question of whether a policy change for a fishery in one region will affect outcomes for fisheries in other regions not directly involved in the policy process. Most policy evaluation focuses on the intended effect on the target fishery and fails to account for potential ancillary effects on other fisheries. In this paper, we test for the existence of policy-induced transboundary impacts on fishery outcomes and explore the implications for regional fisheries management. We examine whether a policy change in one region—from industry-wide quotas with effort restrictions to catch shares—affects adjacent fisheries in another region that are managed with industry-wide quotas.

An important motivation for our study is the trend toward using catch shares to manage fisheries in the United States and in many other parts of the world. Economists have focused their empirical attention on the economic and biological consequences of catch share programs for the target fisheries. Evidence for economic efficiency gains is strong—from fishery case studies, ex ante empirical analyses, and individual fishing quota market data (Weninger 1998; National Research Council 1999; Grafton, Squires, and Fox 2000; Newell, Sanchirico, and Kerr 2005; Lian, Singh, and Weninger 2009). There is still debate about whether catch shares improve biological outcomes in the target fishery, that is, whether apparent gains reflect other policy measures or population changes are plausibly attributable to catch shares (Costello, Gaines, and Lynham 2008; Bromley 2009; Nowlis and van Benthem 2012). As this debate continues, economists are beginning to examine the effects of catch share programs on employment (Abbott, Garber-Yonts, and Wilen 2010) and the implications of distributional consequences for policy design (Pereau et al. 2012).

Although understanding the direct effects of catch share programs is important, evidence of fishery spillovers from a range of policy contexts suggests that indirect effects are likely to matter. One area of research examines how effort for targeted species affects catch of nontargeted species. Scholars have found that overall fishing effort supply reflects target fishery entry and exit decisions (Bock-stael and Opaluch 1983) as well as finer-scale targeting decisions within multispecies fisheries (Zhang and Smith 2011). Effort also adjusts spatially to economic incentives on fine scales such as fishing grounds (Holland and Sutinen 2000; Smith 2002; Zhang and Smith 2011) and across large regions (Mistiaen and Strand 2000; Smith and Wilen 2004; Hicks and Schnier 2008). Other work examines how regulation of one target species or one fishing location changes the opportunity cost of fishing effort and thus affects outcomes for other species and locations. Quaas and Requate (2013) show theoretically that regulation of one species can lead to depletion of other stocks that otherwise would have been less depleted under open access. There is some evidence for this process occurring within regions in which individual fishing quotas (IFQs) affect the targeting of unregulated species (Asche, Gordon, and Jensen 2007; Hutniczak 2014). Others find evidence of this process in the effects of regulation of a target species on a bycatch species (Abbott and Wilen 2011). At a global scale, the hypothesis of “roving bandits” suggests that excess fishing capacity roams the globe to exploit profit opportunities where local institutions have failed to restrict access (Berkes et al. 2006). The evidence for this phenomenon is less direct than the empirical studies of fishing behavior, but the mechanism is plausible. All of this work highlights the potential for a policy change in one region to affect fishery outcomes in another.

Outside of fisheries, economists have paid considerable attention to policy spillovers generally, and leakage specifically. For example, if one country imposes restrictions on carbon emissions (e.g., a carbon tax or cap- and-trade system), carbon-intensive industries have an incentive to relocate to countries without these restrictions. The overall reduction in carbon is lower because of this “leakage” (Goulder and Parry 2008; Wiener 2007; Fowlie 2009). Carbon policies can also create leakage through natural resource markets. Carbon sequestered in conserved forests in one location may be offset by carbon emissions from deforestation in other areas, and there is a substantial potential for leakage in the U.S. forest sector (Murray, McCarl, and Lee 2004). Indirect evidence from trade in forest products is also consistent with a leakage mechanism; reforestation gains in seven developing countries are mostly offset by land use changes in other countries (Meyfroidt, Rudel, and Lambin 2010). Recent theoretical work suggests the possibility of negative leakage if a carbon tax changes the production mix and draws more productive resources into the newly regulated sector from the unregulated sector (Baylis, Fullerton, and Karney 2014).

We examine the potential for leakage due to regional-specific fisheries regulation. Our empirical analysis examines the impacts of the 2010 New England Groundfish Sector Program (Sector Program hereafter) on adjacent fisheries in the Mid-Atlantic (MA) region. Similar to how IFQs operate, the Sector Program allocates shares of groundfish total allowable catch (TAC) to cooperative entities called sectors. The economic logic is that these sectors will rationalize the deployment of fishing effort, lower costs, increase revenues through market timing, and profit share. The baseline policy prior to the Sector Program used effort controls (specifically days-at-sea restrictions) to enforce the TAC. Recent work using quasi-experimental methods finds evidence of economic benefits of sector management on target fisheries (Scheld, Anderson, and Uchida 2012; Scheld and Anderson 2014), and survey-based research finds evidence that sector formation is correlated with social capital (Holland et al. 2013). However, empirical work on the Sector Program to date does not address the potential impacts on other fisheries. We hypothesize that regulation in one region (New England) frees up fishing capital to exploit fisheries in another region (the MA) that has lower barriers to entry. In practical terms, many New England groundfish vessels had permits and established catch records for MA species before the Sector Program. This regulatory history, coupled with species ranges that in some cases span both regions, suggests that there were low barriers to entry; the Sector Program might have lowered the average opportunity cost of using capital from New England fisheries in adjacent MA fisheries. On the other hand, increased profitability of New England fisheries could draw high-skill New England effort into New England fisheries and away from MA fisheries. If low-skill New England effort operating in MA fisheries does not have an incentive to increase fishing in the MA, the net effect would be negative leakage. The central question of this paper is to what extent New England groundfish fisherman increased participation in the MA as a result of groundfish sector management that altered the deployment of fishing effort in New England.

In the next section, we describe the fisheries and the regulatory systems before and after the policy change. This background motivates the development of a simple theoretical model in the following section. The model shows that no leakage, positive leakage, and negative leakage are all possibilities, and hence, understanding the extent of leakage is an empirical question. We approach this empirical question by analyzing the creation of the Sector Program as a natural experiment. The control group includes MA fishermen who did not fish for New England groundfish (sector) species before the program. We use their fishing outcomes (catches and revenues) before and after the Sector Program to estimate our counterfactual. We use a differencein-differences (DID) estimator to isolate the treatment effect. We estimate models with both pooled and disaggregated MA species and perform falsification tests. We find evidence of positive leakage in the aggregate data, but the effects are heterogeneous across individual fisheries.

II. BACKGROUND

The New England groundfish complex includes a diverse set of commercially caught demersal species. Historically, these fisheries were among the highest-volume and highest-value fisheries in the United States, with Atlantic cod (Gadus morhua) being the most notable species. Table 1 contains landings data for each groundfish species in 2009, the year prior to Sector Program implementation. Although all of these species are marketed as varieties of whitefish, species-specific ex-vessel price per pound ranges from $0.48 to $1.66. These prices reflect the continuum of market categories, including inputs into processed products like fish sticks, fillets for fish and chips, and higher-end plate fish. Nevertheless, compared to typical high-end white-fleshed plate fish such as red snapper, grouper, and halibut, the groundfish prices are low. For context, the 2009 red snapper ex-vessel price per pound averaged across all U.S. fisheries was $3.24. Table 1 also breaks down landings by gear type. Otter trawl is the dominant gear type (61%), with substantial catches from gillnets (34%). The vast array of other fishing gear types accounts for only 5% of landings for which gear information is available. Across the fisheries, there is some variation in the shares of different gears. Most are heavily dominated by otter trawl, but gillnets contribute roughly half for cod and pollock.

A number of factors could facilitate spillover from the New England sector (NES) to MA fisheries. Home port states of NES fishermen¹ include all of the New England states with marine coastlines but also include several states from the adjacent MA region. New Jersey, for instance, is listed as a home port for fishermen in seven of the sectors. Other MA states in the list include New York, North Carolina, and Virginia. Beyond geography, there are possibilities of market substitutability, gear substitutability, and institutional barriers. Market substitutability would matter if fishermen have an easier time identifying buyers for products that are close market substitutes than for products that are qualitatively different. If a fisherman can use the same gear to catch a different species, then the fisherman does not need to incur the costs of changing gear and potentially learning a new fishing method. If the Sector Program indeed frees up capital, it frees up vessel time and the gear on that vessel. Lastly, institutional barriers could prevent spillovers by increasing the cost of entry. Our data do not allow us to explore the relevance of institutional barriers because MA fisheries with qualitatively different regulatory institutions—namely, catch share-based management that would require entrants to purchase quota—also use very different types of gear from those used to catch New England groundfish.

Table 2 contains landings data for species managed by the Mid-Atlantic Fishery Management Council (MAFMC). Here, all species are included to consider all possibilities for leakage from New England into the adjacent region. There are four species that we categorize as whitefish: summer flounder, scup, black sea bass, and tilefish. Summer flounder plausibly competes with the flounder species in the New England groundfish complex (winter flounder, yellowtail flounder, and plaice) despite having a higher ex-vessel price. Scup similarly is a plausible market competitor for the lower-end whitefish species in New England (hake, pollock, and redfish). The other two species, black sea bass and tilefish, are higher-end plate fish, have ex-vessel prices that are substantially above those of the highest-priced groundfish, and thus are less likely to compete with groundfish directly in the market.

The species categorized as “other seafood” in Table 2 are all qualitatively different from groundfish. Monkfish, though white fleshed, is a unique product; its tail is consumed and its texture is similar to lobster. Atlantic mackerel is a low-end, oily pelagic fish with an exvessel price per pound that is less than half of the lowest-priced groundfish species. Butterfish is a small, bony fish that was once marketed to Asia but now has no direct target fishery and is caught as bycatch. Long-finned squid, surfclam, and ocean quahog produce seafood products that are very different from whitefish. Bluefish is a low-end oily fish that spoils easily and thus does not compete directly with groundfish, and spiny dogfish is a small shark that is often smoked. Despite all of these differences, participants in the sector program that substitute into MA species needed to have some established catch records for those species and, in the process, may have developed relationships with buyers. Considering market substitutability overall, spillovers are most likely for scup and summer flounder but cannot be ruled out for any of the MA species.

Analyzing MA landings by gear reinforces some of the market relationships but offers some other possible substitutes. Both summer flounder and scup are caught primarily by otter trawl, which is the dominant gear type for groundfish. As such, these species appear even more likely as spillover candidates. Black sea bass and golden tilefish are caught primarily by other gears (pots for bass and long line for tilefish). Combined with their market differentiation, spillovers for these species seem very unlikely. The same is true for surfclam and ocean quahog, which are caught with hydraulic dredges. In contrast, Atlantic mackerel, butterfish, and long-finned squid are all dominated by otter trawl, making these species strong candidates for spillovers for groundfish fishermen using otter trawls. Monkfish is a reasonable candidate on the basis of gear, with a substantial percentage of otter trawl, but a higher percentage of gillnets. Spiny dogfish seems less likely with only a quarter of landings coming from otter trawl. Bluefish is dominated by gillnet gear, so the likelihood of spillover hinges on whether gill-net groundfish fishermen are likely to spillover. Overall, given the dominance of otter trawl in groundfish, the best candidates for spillover on the basis of gear are summer flounder, Atlantic mackerel, butterfish, and long-finned squid, and secondary candidates include monkfish and scup.

The baseline regulatory environment prior to the formation of the Sector Program was regulated restricted access and fleet-wide quotas for both NES and MA fisheries. Each fishery required a permit to participate, the numbers of permits were limited, and, like most U.S. federally managed fisheries, TACs for the fleet as a whole were set on the basis of biological stock assessments. In NES fisheries, days-at-sea (DAS) restrictions limited the amount of effort, to prevent the fleet from exceeding the TACs. MA fisheries, in contrast, relied on season closures to enforce TACs; as in regulated open access (Homans and Wilen 1997), MA fisheries are shut down when they reach the TACs.²

After the Sector Program, MA fisheries continued to be managed in the same way, whereas NES fisheries kept some of the baseline regulations and changed others. In particular, the biologically based TACs and limited entry programs were maintained for NES fisheries, but DAS restrictions were eliminated for Sector Program participants. Instead of DAS restrictions, shares of each TAC were allocated to cooperatives called sectors; the allocation to each sector was proportional to the combined historical catches of the fishermen that joined the sector. Sectors could then coordinate fishing effort in order to maximize profits. The presumption was that sectors could catch the same amount of fish using less effort and at lower cost than under the DAS restrictions. For more details on the regulatory history of NES fisheries, see Holland and Wiersma (2010) and Cunningham (2012).

III. THEORETICAL MODEL

Our general hypothesis is that sector management, by promoting the efficient use of capital, frees up capital to target species in nonsector fisheries, including ones in the MA region. To examine sharper predictions about causal effects of the Sector Program, we develop a simple theoretical model. In the model, there are five fishermen, each representing a segment of the composite fleet that fishes for MA and for NES species. All NES fishermen in the model are assumed to have permits to fish in MA fisheries, but MA fishermen have permits only for MA fisheries (our empirical specification below restricts the data set accordingly). These segments (and corresponding nomenclature for fishing effort) are as follows: MA fishermen fishing only for MA species (E^MM), low-type NES fishermen fishing for MA species (E^LM), high-type NES fishermen fishing for MA species (E^HM), low-type NES fishermen fishing for NES species (E^LS), and high-type NES fishermen fishing for NES species (E^HS). High (low) type indicates relatively high (low) fishing skill.³

Fishing effort is a composite of amount of gear deployed, fish-hold capacities of the vessels, time to find and catch fish, and time to identify a buyer when the fish are landed. In the latter sense, the model incorporates market substitutability on the input side. We assume that there are no output-side market considerations such that fishermen are price takers (discussed further below). To keep the model simple, we assume that fishermen who are high (low) type in sectors are also high (low) in the MA. The corresponding harvest functions are as follows: , , , and , where t denotes time, is the stock of MA species, and is the stock of sector species. We assume that the production functions capture fishermen heterogeneity and that all fishermen face a constant marginal cost of effort, c.

Because all species in both regions were managed with fleet-wide TACs before and after the policy change, and these TACs were determined by biological science alone, we assume that the Sector Program did not have any effects on biological stocks.⁴ This allows us to suppress the stock argument from each production function. We assume all production functions are strictly concave, and that for any arbitrary level of effort, marginal product of the high type is higher than that of the low type. Also because of tight biological controls, we suppress the time subscript, t, in the modeling below and distinguish only between the periods before and after the policy change (denoted by before and after subscripts where necessary). We further assume that prices of MA (P^M) and sector species (P^S) are exogenous and unchanged by the policy. The rationale again flows from tight biological control; the policy change does not affect the total landed quantity of each species as long as catch limits bind both before and after (a reasonable approximation for nearly all MA and NES sector species, though our model can accommodate nonbinding quotas). Even in the absence of binding catch limits, exogenous price may be a reasonable assumption given that most of the species in our empirical work below compete in a global whitefish market and constitute a tiny fraction of that market. General equilibrium bioeconomic models of seafood markets (e.g., Quaas and Requate 2013) are more relevant for open access or management settings in which there are no binding biological controls on catches.⁵

Let Q^M and Q^S denote the TAC for, respectively, MA and NES species. The TACs are assumed to be the same before and after the policy change. The TAC constraints can thus be written as [1] and [2]

Denote the fishermen-specific shadow values (MA, low-type, high-type) for [1] as γ^MM, γ^LM, and γ^HM. Similarly, denote the fisher-men-specific shadow values (low type, high type) for [2] as γ^LS and γ^HS. All fishermen have constraints on the maximum amount of effort that they can exert in a season: Ē^M, Ē^L, and Ē^H. MA fishermen exert effort only in the MA, whereas NES fishermen exert effort in both NES and MA fisheries. The constraints with corresponding shadow values in square brackets are [3] [4] [5]

These constraints reflect a combination of time available for fishing and finding buyers, sizes of vessels, and available gear. Prior to the policy change to sectors, there were also DAS regulations that limited the amount of effort sector fishermen could apply to sector species. The limit is the same for both high and low types, Ē^DAS, leading to the following constraints with corresponding shadow values in square brackets: [6] [7]

All 10 shadow values are strictly nonnegative. We assume that for all sector fishermen, DAS limits are below their total effort capacities, Ē^DAS≤Ē^L and Ē^DAS≤Ē^H.

We assume that the MA fishermen take NES fishermen’s effort in the MA fisheries as given. As such, their optimization problem has the same structure before and after the policy change; the only difference is how much NES fishermen are fishing in the MA. Maximizing profits subject to constraints [1] and [3] yields the following Kuhn-Tucker conditions for MA fishermen: [8] [9]

Maximizing profits for low-type NES fishermen subject to constraints [1], [2], [4], and [6] yields two sets of Kuhn-Tucker conditions both before and after the policy change. Before the policy change, they are [10] [11] [12] [13]

For high-type fishermen, whose constraints are [1], [2], [5], and [7] before the policy change, the Kuhn-Tucker conditions are [14] [15] [16] [17]

After formation of sectors, individual fishing quotas are assigned to low- and high-type fishermen that add up to the TAC: [18]

This policy change brings into alignment the shadow value of individual NES fishermen for the NES species, which we denote as λ to distinguish it from the period before the policy change. In equilibrium, λ is also the quota price in each fisherman’s optimization. After the policy change, there are no longer DAS restrictions on participants in the Sector Program, so constraints [6] and [7] disappear. The Kuhn-Tucker conditions for the MA fisheries are unchanged for both MA and for NES fishermen, but the shadow values themselves can be influenced. For low-type NES fishermen after the policy change, the Kuhn-Tucker conditions for MA fisheries are the same, and for NES fisheries they are [19] [20]

For high-type NES fishermen after the policy change, the Kuhn-Tucker conditions for the MA fisheries are the same, and for NES fisheries they are [21] [22]

IV. DATA AND METHODS

We are interested in testing for a causal relationship between the expansion of New England’s groundfish sector management program and altered levels of participation in MA commercial fisheries. To this end, we employ a DID estimator that relies on defining treatment and control groups and measuring outcomes before and after the policy change. In this case, treatment groups are collections of the individual fishermen who may or may not have altered their level of fishing behavior in MA fisheries after the implementation of the NEFMC Northeast Multispecies Fisheries Management Plan Amendment 16 on May 1, 2010. These are the NES fishermen fishing for MA species in the theoretical model. For statistical comparison, control groups are defined to include the individual fishermen who participated in MA fisheries during the same study years but were not affected by policy change in New England (from the theoretical model, MA fishermen fishing for MA species). The empirical model estimates the level of a measured outcome for the average individual fishermen, so implicitly we average over low-type and high-type NES fishermen.

Although the ideal outcome of interest is fishing effort, we lack a direct measure of effort and instead use two alternative outcome variables: landed volume and revenue. For these variables, the National Marine Fisheries Service (NMFS) collects complete data from both fishermen and fish-dealers, so the commercial fisheries database contains some cross checks. For this study, we have access to this database, which includes every individual fishing vessel’s landed quantity and gross revenue by species for every day from January 1, 2006, through November 30, 2011. Data on raw landings by individual vessel permits were reshaped to form a balanced panel around the sharp Sector Program implementation date of May 1, 2010.

Treatment groups are defined to include those individual fishermen whom one might expect to be changing their MA fishery participation in response to the expansion of groundfish sectors. We consider three different formulations of treatment: Group 1. All fishermen with landings of MA species who also reported landing any species man-aged under the New England Northeast Multispecies (groundfish) Fisheries Management Plan, between 2006 and 2011Group 2. All Group 1 fishermen who enrolled in one of the 17 New England groundfish sectorsGroup 3. All Group 2 fishermen who enrolled in one of the 17 New England groundfish sectors, but who were reported by their sector manager as “inactive” in that year’s Sector Operations Plan submitted for NMFS approval prior to the May 1 start of the fishing year

Group 1 provides the broadest definition of treatment. Group 2 is a subset of Group 1, and Group 3 is a subset of Group 2. Running a model with the Group 1 treatment definition would indicate the effect of the Sector Program, if any, on the level of MA fishery participation for any individual whose landings history links him or her to the New England groundfish fishery. Using Treatment Group 1 takes a wide view of who comprises the potential source of leaked effort. Finding a significant DID estimator for a Group 1 model run could be taken to indicate a watershed change in how the fishing population targets NES and MA species as a result of the Sector Program.

Group 2 fishermen are a subset of Group 1; Group 2 includes only those individuals who enrolled in a sector, thereby contributing their catch history to the sector’s collectively managed allocation. Total enrollment across 17 sectors in 2011 was 819 individuals. In keeping with the idea that sector management creates opportunities to expand fishing activity beyond New England groundfish—by virtue of increased flexibility to catch groundfish over the entirety of the fishing season—one would expect to see a stronger positive increase in MA fishery participation in Group 2 models relative to Group 1 models. If DID estimates are significant for Group 2 but not for Group 1, one could infer that the Sector Program altered MA fishing effort for participating individuals, but not the entirety of the fishing community that lands NES species.

Group 3 fishermen are a subset of Group 2, including those individuals who reap the financial benefits of their sector’s collective catch but do not actively fish their own allocated share. The number of inactive sector members was 448 in 2011. Idled in New England fisheries by their sector managers, inactive sector members might have the greatest opportunity to redirect effort into MA fisheries. Some of these fishermen may have been idled because their vessels are less efficient or because they had a lesser desire to continue fishing. This possibility cuts against the likelihood of a positive treatment effect. Nevertheless, holding fishing skill and vessel efficiency constant, Treatment Group 3 should show the strongest, most positive increase in MA fishery participation. That said, the many ways that substitution can unfold leads to some ambiguity in predictions, and the lack of a single definitive theoretical basis for defining treatment is another reason to consider multiple definitions of treatment.

In an ideal experiment, treatment and control groups would be the same in every way except for their levels in the outcome variable of interest. In such a case, the difference in the mean outcome level between the treatment and control groups would provide a confident measure of the effect of being in the treatment group. The multitude of unobservable factors affecting pounds and value of landings makes it impossible to control for every factor that explains the difference in outcome levels between treatment and control. Weather, local product demand, and the strength of the overall economy as it impacts the opportunity cost of a fisherman’s time are just a few examples of these unobservable factors that are not reported in commercial fisheries data. DID estimation is an appropriate strategy for this case because it compares the relationship between trends in outcome levels over time for treatment and control groups. The identifying assumption of this strategy is that, in the absence of treatment, the difference in the treatment and control groups’ outcome levels— explained by omitted unobserved factors— remains the same over time. This first-stage difference describes the relationship between the underlying trends in treatment and control group fishing effort prior to policy implementation.

We use an ordinary least squares (OLS) regression–based DID estimator with fixed effects. A regression model is superior to the mean differencing estimation in its ability to control for variation in outcomes that are attributable to time period effects that are independent of the observational unit and time-independent effects. Effects that are invariant across time periods would include the essential differences between fishermen, such as relative access to capital, skill level, relationships with buyers, or experience. Fixed-effects models provide valid estimates of outcomes even if the unobserved, time-independent individual effects are correlated with other regressors. As with other OLS models, conditional exogeneity is required. This means that, after controlling for observable and unobserved factors that differentiate one individual in the sample from another, the expected value of the residual error is zero (Hayashi 2000, 7).

Each calendar year was subdivided into three seasonal periods, or trimesters (January–April; May–August; September–December), creating 18 time periods within the time series. Including a dummy variable as a regressor for each portion of the year absorbs the effect of time-specific aggregate landings shocks that would be correlated across individuals in a certain period, regardless of the year. Controlling for seasonal variation is particularly important when analyzing fisheries, because harvest patterns vary significantly in response to stock movements, spawning seasons, weather, regulatory restrictions, market conditions, and intense derby periods in regulated open access fisheries. A vector of year dummy variables is also included to control for variation in outcomes that are unique to a given year but affect all individuals equally. We also include a linear time trend for the trimester-year combination. This time trend absorbs any monotonic changes in catch or value over the entire time period.⁷ The fixed effects DID models used are of the form [23] where i denotes an individual fisherman, p denotes a time period within the year, y denotes a year, Y_ipy is the outcome level for individual i in time period p and year y, D_ipy takes the value of 1 for treated individuals in time period p and year y and 0 for control, T_py is a linear time trend for the period-year combination, γ_p are time period dummies, δ_y are year dummies, α_i are individual fixed-effects, and ε_ipy is the error term, which is assumed to be independent across different fishermen, but allowed to be correlated for different period observations for the same fisherman. The estimated value of β is interpreted as the DID estimator. The models were run with both POUNDS_ipy and VALUE_ipy as the dependent variable.

The critical identifying assumption in a DID model is that in the absence of treatment, trends in outcomes between the treated and control groups would have been parallel. To test the validity of this identifying assumption, we conduct falsification tests for all outcomes that have statistically significant treatment effects. To do this we drop all posttreatment data (years 2010 and beyond) and redefine 2008 as the proxy “treatment” year. Thus, data from 2006 and 2007 are considered “pretreatment,” and data from 2008 and 2009 are considered “posttreatment.” The exact same DID models are run on these data. Since there was actually no treatment during this entire period, if the identifying assumption holds, we should observe no statistically significant treatment effect.

Defining the scope of MA fisheries is critical to conducting policy-relevant model runs. The MAFMC directly manages 13 marine species under seven fishery management plans. Twelve of these species are listed in Table 2, and the last is short-finned squid, which does not appear in the aggregate NMFS database for landings by gear (so was excluded from the table). Short-finned squid landings do appear in our microlevel database and can be used in the analysis. The broadest possible definition of MA fisheries thus includes all 13 species. We pare down the full roster of MA species to eight highly susceptible species based on the following reasoning. New England groundfish species are primarily targeted with bottom and midwater trawl gear. Landings data on surfclams and ocean quahogs were excluded, given the specialized nature of the shellfish dredging fishery, the very different product (not substitutable for whitefish characteristic of New England groundfish), and the fact that they are managed with catch shares. Tilefish data were excluded due to the small number of active fishermen, gear differences, and low entry pressure that has created a de facto individual transferable quota fishery. Bluefish data were excluded because the fishery is low value and is mainly prosecuted with hook-and-line gear. Black sea bass data were excluded because over 40% of landings are taken with pot fishing gear, and commercial landings are heavily outweighed by recreational fishing (National Oceanic and Atmospheric Administration 2012). Using each of the three treatment groups in turn, the model was run to determine effects on aggregated pounds and landed value outcomes for the eight remaining species: summer flounder, scup, spiny dogfish, long-finned squid, short-finned squid, Atlantic mackerel, butterfish, and monkfish. These models provide a region-wide look at altered activity patterns in the MA. This high-susceptibility aggregate-level analysis is the main test of our hypothesis that the Sector Program induced spillover into the MA.

We also consider model runs for individual species or combinations of species. We base these runs on identifying likely candidates for spillovers. The number of individual fishermen included in a single model run is always large, with a minimum of 519 and a maximum of 2,380. As discussed above, the appeal of fishing for MA species to New England fishermen could stem from geographical stock distributions that straddle management regions, the amount of overlap in fishing seasons, ease of permit access for MA species, similarity of the product market, or the similarity between the type of gear used on New England groundfish and identified MA species. In addition to using information in Tables 1 and 3, these attributes were assessed based on information available from the NMFS and the Northeast Fisheries Science Center (National Oceanic and Atmospheric Administration 2012; NEFSC 2006), as well as personal communication with MAFMC staff.⁸

Noting that susceptibility to redirected effort varies across species according to the factors mentioned above, the three-treatment approach was applied to landings data for each of six MA species in isolation. The model runs used data on summer flounder, scup, long-finned squid, Atlantic mackerel, butterfish, and monkfish. Spiny dogfish and short-finned squid were dropped, having been identified as primarily incidental bycatch species, and we combined summer flounder and scup to create a single whitefish category. Table 3 summarizes the total number of fishermen targeting different groups of MA species and how many vessels are in treatment and control groups. Across models, the control group remains unchanged. This group corresponds to MA fishermen who never (in sample) have landed any of the sector species. Moving from Treatment Group 1 (all vessels with catch records of sector species) to Treatment Group 2 (enrolled sector members) raises questions about whether some of what we count in Treatment Group 1 and exclude from the control group are, in fact, MA fishermen who have had some incidental landings of sector species. In other words, we can be confident that all fishing vessels in Treatment Groups 2 and 3 are treated with the Sector Program, but it is possible that some fishermen in Treatment Group 1 should actually be in the control group. By excluding these fishermen altogether in model runs with Treatment Groups 2 and 3, we believe that our experimental design errs on the side of being conservative.

V. RESULTS

We estimate models for each of the three treatment group definitions (3 groups), both pounds landed and dollar value (2 groups), and all eight vulnerable species plus five individual species or combinations of species (6 groups), for a total of 36 models (3 × 2 × 6). All statistical significance is based on standard errors that are clustered at the fisherman level. Table 4 summarizes the treatment effects for all model runs, significance, and results of falsification tests.⁹

For all eight of the highly susceptible species combined, we find evidence of quantity spillovers in Treatment Groups 2 and 3 that pass the falsification test. Our estimates suggest that fisherman who were members of the Sector Program (Treatment Group 2) increased landings of MA species by 8,978 pounds relative to MA fishermen. The estimate for those fisherman who were inactive members of a sector (Treatment Group 3) is similar at 7,096 pounds. However, estimates of the treatment effect for the dollar value of landings was not statistically significant for any treatment definition.

Turning to individual species, the results are more varied. Atlantic mackerel is statistically significant for pounds and value in all three treatment definitions. However, only the results for dollar value and Treatment Group 1 pass a falsification test. Mackerel is caught almost exclusively with otter trawl and thus is a good substitute for groundfish species on the basis of gear, but it is a relatively weak market substitute. The passing of the falsification test for Treatment Group 1 but not for the other definitions raises questions about the treatment definitions; eliminating fishermen who have caught groundfish but who did not enroll in a sector leads to the weaker results. One possibility is that the formation of sectors caused groundfish fishermen in the common pool to intensify fishing for other species. This is not implausible given that mackerel is a high-volume and low-value fishery. It might be that fishing vessels best equipped to compete in a high-volume environment were less interested in joining a sector.

A similar mechanism could explain statistically significant Treatment Group 1 results that pass falsification tests for long-finned squid and butterfish. Like mackerel, squid is a high-volume and low-value species caught predominantly with otter trawl. Butterfish, which is also of low value, is caught mostly as bycatch in the squid fishery. In the sample period, there was no directed butterfish fishery. So, it is not surprising that its treatment effects track mostly with those of squid, though it is clear from Table 2 that butterfish is also caught with some other gears and thus as bycatch in other fisheries.

The results for monkfish are all negative, but only two results are statistically significant, and none pass falsification tests. In other words, there is no evidence that the Sector Program caused decreases in monkfish landings for New England fishermen relative to MA fishermen.

The results for combined summer flounder and scup (the whitefish category most similar to New England groundfish) are statistically significant for pounds and dollars in Treatment Groups 1 and 2, and the Treatment Group 2 results pass falsification tests. Both of these species are strong candidates for spillover on the basis of gear and market substitutability. As such, it is not surprising that the Sector Program caused spillovers for enrolled sector members. The question is why the statistical results do not show the same for inactive sector members (Treatment Group 3). The results are positive and of similar magnitude to the Treatment Group 2 results but lack statistical significance. One possibility is simply that inactive sector members have less profitable vessels, and it is more profitable for them to idle their vessels and allow active sector members to catch both the allocation of groundfish and some of the MA spillover. We do not observe ownership structure, but it is certainly a theoretical possibility that an individual owning two vessels with different efficiencies could choose to idle one to maximize profits across the two.

Overall, there is evidence that the Sector Program causes some spillovers into the MA, but the evidence is not consistent across different ways of defining the treatment group, pounds versus dollars, and species groups.

VI. DISCUSSION

Our theoretical model predicts that positive, negative, or zero effort leakage are all possible results of the transition to the Sector Program, depending on whether various constraints bind before and after the policy change. Effort leakage, in turn, may or may not translate into changes in catches for NES and MA fishermen. The complexities of our setting echo themes in the public economics literature on multiple market distortions. When there are multiple distortions, a policy change toward the first best in one market may not influence the other market in ways one would predict from first-best reasoning (Fullerton and Metcalf 2001). Moreover, the presence of multiple distortions in a second-best world is one possible explanation for our mixed empirical results. Some fisheries appear to have positive spillovers, while others do not.

Our model does not reveal whether catch spillovers are costly to efficiency. We do not test the resulting effects on technical efficiency, and the theoretical model reveals potential outcomes that cut either way. If NES fishermen displace MA fishermen in MA fisheries, technical efficiency could increase. There would be equity considerations, especially because MA fishermen were not involved in the NES fishery policy process. On the other hand, because the Sector Program likely draws more efficient effort into NES fisheries and away from MA fisheries, NES vessels increasing participation in MA fisheries are likely less efficient and could lower the overall technical efficiency of MA fisheries.

There are at least four possible implications of our empirical findings that the Sector Program caused some spillovers into MA fisheries. First, potential efficiency gains from the Sector Program in New England may not reflect the full changes in efficiency; they may have been offset at least partially by efficiency losses in the MA or could have contributed to increased efficiency in the MA. It is important to note that we did not test this hypothesis directly, but our results are consistent with there being some unintended consequence. It could be that efficiency changes from spillovers are not large enough to warrant the costs of controlling them. We did not explore this question, which suggests a direction for future research.

Second, there may be opportunities for regions like the MA to address spillovers proactively before a policy change in an adjacent region. The use of sideboard limits in Alaska essentially attempts to deal with spillovers before they occur. In several instances, the North Pacific Fishery Management Council has restricted groups of vessels that participate in one catch share-managed fishery from expanding their aggregate level of retained catch in another fishery beyond the proportion that they harvested during a qualifying period.¹⁰ Vessels without sideboard exemptions are prohibited from directed fishing in the spillover fishery once they reach their collective participation limit. The MA chose not to pursue such an approach despite full knowledge of the coming sector program in New England. A reasonable question is whether the efficiency gains would have justified the cost of administering such a program in the MA.

Third, a more radical approach would be full and complete creation of catch shares for all federally managed fisheries that do not already have catch share programs. Such programs might be a mix of individual transferable quotas, sectors, other forms of cooperatives, and territorial use rights in fisheries, but would aim to undo residual open-access incentives in U.S. fisheries. Although changes of this sort would almost certainly be politically infeasible in the current policy climate, theoretically they would create high barriers to entry for fishing effort that might spillover from one region to another. As such, a radical policy change like this might avoid regional spillovers within the United States. The sentiment echoes that of calls to design climate policy that is all encompassing, that is, to “think globally, act globally” (Wiener 2007). Still, broad formation of catch shares in U.S. fisheries could produce unintended consequences. Barriers to domestic commercial fishing entry would undoubtedly rise. Excess fishing capital might enter unregulated high-seas fisheries, be sold and exported to countries with less tightly controlled fisheries, or create more pressure from the domestic recreational sector by expanding the supply of charter vessels. Moreover, a policy change like this in developing countries raises questions about the impacts of fisheries reform on the poor that rely on open-access conditions as part of a livelihood strategy that blends fishing and nonfishing income (Liese, Smith, and Kramer 2007; Wilen 2013).

Lastly, spillovers may reflect complex livelihood strategies within fisheries that single-species management fails to acknowledge. In this sense, spillovers from a policy change may be no different from effort redirection in response to fluctuations in fish stocks or market conditions. Managers would need to know not just if spillovers were occurring, but whether they had negative efficiency consequences. Supposing that there were not efficiency losses, the management implications are very different from those described above. Managers in this case might have a stake in facilitating, not limiting, opportunities to redirect effort. Single-species TACs might be replaced by value-based IFQs (Turner 1996) or some form of portfolio management (Sanchirico, Smith, and Lipton 2008). These ideas, though discussed in the economics literature, have yet to emerge as practical policy proposals.

The implications discussed above are not perfectly overlapping. The possibility that spillovers are just part of a larger economic context for fishing behavior suggests that no action is necessary, but it also begs the question of why some spillover effects actually passed falsification tests. Supposing that spillovers are enough of a problem to warrant a policy action, one type of action—for example, a complete allocation of catch shares— would negate the need for another type of action—for example, a network of sideboards.

The notion of spillovers as livelihood raises philosophical questions about how to define the treatment and control group in our context. Already, we chose to exclude many fishermen from the control group because they had caught some sector species even if these catches were truly incidental (Table 3). Scheld and Anderson (2014), focusing on a very different question from ours, find aggregate gains of $30 million from the Sector Program. However, they actually include gains in MA fisheries in their tally (increased revenue in the MA for Sector Program fishermen relative to counterfactual MA revenues for Sector Program fishermen). It could be that some of what Scheld and Anderson (2014) count as gains are, in fact, transfers from the MA fishermen, whereas some of what we imply are efficiency losses are, in fact, gains from better market timing that the Sector Program fishermen are able to exploit. Overall, our results are not an indictment of catch shares, but they suggest that there are possible ways to improve performance through some coordination across regions and taking a broader view of fisheries than the single-species focus of most current management institutions.