Optimal Intra-annual Exploitation of the Maine Lobster Fishery

Daniel S. Holland

Abstract

I present a model of the Maine lobster fishery that incorporates a monthly demand model and an empirically estimated production function that accounts for seasonal variability in catchability, inseason depletion, and congestion effects. I compare optimal exploitation with observed exploitation and evaluate the extent to which profits under a conventional individual transferable quota (ITQ) system would be dissipated by congestion and in-season depletion externalities. The models show that profits could be substantially increased from the status quo through effort reductions and changes in the harvest schedule, but profits under an ITQ system may be reduced by as much as 30% by unresolved externalities. (JEL Q22)

I. Introduction

American lobster is the most valuable fishery in the northeastern United States and by far the most valuable fishery in Maine. Landings in Maine, which now account for over 80% of U.S. landings, have grown dramatically over the last two decades with record landings of over 75 million pounds in 2009 (Figure 1). Revenues for the fishery grew with landings, peaking at $318 million in 2005, but then declined by 30% to $221 million in 2009 despite the fact that landings in 2009 were nearly 10% greater than 2005 (Maine Department of Marine Resources 2010). Although the gross value of the fishery is high, the profitability of the fishery is low (Thunberg 2007) and is decreasing with increases in fuel and bait prices and recent declines in ex-vessel price.

There has long been a concern that landings from the fishery are too concentrated in the late summer and fall, when most of the lobster has recently molted, is of low quality, and fetches a lower price.1 It is also widely accepted by managers and also by the majority of license holders that there is excess effort in the fishery.2 There is interest from managers and the industry in implementing management actions that would shift production to other times of year, thereby increasing overall revenues. Cheng and Townsend (1993) suggested that revenue could be increased by up to 18% if landings were distributed more optimally; however, as I show here, shifting production too much may also increase costs and more than offset revenue gains.

I construct an intraseasonal bioeconomic model of the fishery that I use to evaluate how revenues, costs, income to fishermen, and profits might change with different effort levels and seasonal exploitation patterns. I create a retrospective model that I use to look back at prior years and, assuming the same level of biomass and recruitment to the fishery, determine the optimal allocation of effort and catch over the year. I then compare profits associated with optimal exploitation to the observed exploitation pattern.

I also evaluate the degree to which an individual transferable quota (ITQ) management system would be expected to optimize profits in the fishery. Because the fishery is subject to in-season depletion and congestion externalities that a standard ITQ system would not resolve, we would expect some reduction in profits relative to the optimum (Boyce 1992). Clark (1980) showed that congestion externalities could lead to rent dissipation in an ITQ fishery but discounted this as serious problem. The ambiguity of this concept has led to some disagreement over whether losses from congestion externalities are likely to be significant (Boyce 1992; Danielsson 2000; Boyce 2000). Boyce (1992) also showed with a theoretical model that in-season stock depletion could create rent dissipation in an ITQ fishery by promoting a race for fish. Bisack and Sutinen (2006) explored this question empirically with a study of the Challenger scallop fishery in New Zealand and estimated that an unresolved in-season stock externality in that ITQ fishery reduced firm profits by 9% to 20% below optimal management. I show that there would likely be similar levels of rent dissipation in an ITQ system for Maine lobster due to a combination of congestion and in-season stock externalities.

Figure 1

Maine Lobster Landings and Landed Value 1950-2009

II. The Model

I construct a numerical bioeconomic model that embeds an empirically estimated monthly demand model and an individual-level daily production function that together determine the catch, revenue, cost, and profits associated with alternative monthly effort levels, given specified biomass and input prices. Monthly effort levels are optimized to achieve alternative objectives and compared to observed profits.

Demand Model

A monthly inverse demand function for the Maine fishery is estimated using landings and revenue data from 1990 to 2007.3 Using a log-log specification I model the monthly average landed prices in Maine as a function of monthly landings in Maine, the U.S.-Canadian exchange rate, U.S. quarterly per capita personal income, and the percentage change in U.S. GDP from the prior year. I model a structural shift in the market in 1994, at which time both the slope and intercept of the demand function shift. I do not model supply and demand simultaneously because supply appears to be exogenously determined by lobster availability and the time of year rather than price. The fact that catches, and apparently effort, have increased in recent years as price declined provides support for that assumption. Though I expected Canadian and perhaps Massachusetts lobster landings to be important predictors of lobster price in Maine, these variables prove to be insignificant. The Canadian exchange rate is a factor because much of the Maine lobster is exported to Canada for processing; however, Maine lobster (at the ex-vessel market level) does not appear to compete in the same market with Canadian lobster, perhaps because the imported Maine lobster is being processed into frozen product rather than being sold live. The model specification also corrects for first-order autocorrelation.

Table 1

Inverse Demand Function for Maine Lobster Ex-Vessel Price

This quite parsimonious demand model provides an excellent fit to the data, explaining 84% of the variance with all of the explanatory variables significant at the 1% level (Table 1). As expected, monthly prices are inversely correlated with monthly Maine landings. The intercept is higher but the elasticity is greater after 1994, reflecting an increased share of product being exported to Canada for processing. It appears that since the lobster is not going into the live market (but rather is processed into frozen product) the price is lower, but more product can be absorbed with less impact on price. As expected, prices generally increase with per capita income but are also sensitive to economic growth. When the change in GDP from the prior year is higher, prices are relatively higher, reflecting the fact that lobster is a luxury good and demand tends to fall in an economic downturn.

Production Function

A production function is estimated using individual-level catch and effort data collected through a port sampling survey that has been running since 1966. The survey collects data on landings, traps hauled, average soak times, and, since 2001, bait use from a randomly selected sample of lobstermen each month. I use a translog specification to model daily catch as a function of the number of traps hauled, the soak time, bait per trap, the estimated total number of traps hauled in Maine that month, and the legal-size biomass of lobster as estimated by the current stock assessment model (Atlantic States Marine Fisheries Commission 2009). The total numbers of traps hauled in Maine is included as a proxy for congestion. There is anecdotal and experimental evidence that there is a substantial congestion effect in the fishery (e.g., Wilson 2007), and the significance of this variable confirms that. I also include monthly dummy variables to capture recurrent seasonal shifts in lobster catchability that relate to the migratory and molting patterns of lobster. As the model shows, catchability is considerably higher in the late summer and fall, peaking in November (controlling for other factors). The production function I use in the simulations is estimated with data from 2001 to 2007, when data on bait use was collected. I refer to this production function henceforth as the “current” production function (Table 2). I found that the production function estimated with only this more recent data is systematically different than the function fit with the longer data set (Table 3). As I explain below, the more recent data and production function better reflect the current fishery and how it would be affected by changes in seasonal effort distribution in the short to medium term. In the long run, if the lobster biomass declines to the levels in the 1980s or 1990s, the long-term production function may be more applicable, with important consequences on optimal intra-annual effort and catch distribution (Holland 2011).

Table 2

Current Translog Production Function of Individual Daily Lobster Landings in Maine Estimated with 2000-2007 Data

With the translog specification, the interpretation of the individual model coefficients is not straight forward; however, it is illustrative to consider the elasticities of catch with changes in daily effort, soak time, total effort in Maine (i.e., congestion), and monthly biomass (i.e., in-season depletion). Table 4 shows the elasticities from both the post-2000 model and the model estimated with the longer data set from 1982 to 2007, and Figures 2-4 illustrate these effects graphically. Both models show economies of scale, that is, the elasticity of catch to changes in traps hauled per day is greater than one, but elasticity is greater with the more recent production model. Both models also show congestion effects (e.g., elasticity is negative for total trap hauls in Maine), but the absolute value of the elasticity with the post-2000 model is smaller, suggesting a weaker congestion effect (Figure 2). Elasticity of catch to monthly biomass is positive but less than one for both models, suggesting catch rates decline as biomass is fished down but less than proportionately (Figure 3). However, the elasticity is smaller in recent years, suggesting catch rates are less affected by in-season depletion. Optimal soak time is affected by biomass and total effort levels (Figure 4). The more recent model suggests that, at the mean soak times in recent years, catch would actually be higher with shorter soak times, while the opposite is true with the long-term model. As Figure 4 shows, optimal soak time is higher when biomass and total effort are lower. Actual soak times used by lobstermen do in fact tend to be much higher in the winter and spring, when fishable biomass and total effort are low (as are lobster metabolism and activity), demonstrating that fishermen are aware of this effect and operate accordingly. The reasons for the differences in the “current” and longer-term production function are unclear, but anecdotal evidence suggests that the fishery has expanded spatially as the abundance of lobster has increased, so that it can effectively absorb more effort with less congestion. Lobstermen may also be following the lobster more as they migrate offshore in the late fall, keeping catch rates higher. Unfortunately there is insufficient data on the fishery to explore these hypotheses.

Table 3

Long-Term Translog Production Function of Individual Daily Lobster Landings in Maine Estimated with 1982-2007 Data

Table 4

Elasticity at Data Means of Catch per Trip to Trip Haul, Soak Days, Monthly Total Trap Hauls in Maine, and Monthly Fishable Biomass

Figure 2

Simulated Catch per Trap Haul as a Function of Total Monthly Trap Hauls with Post-2000 Data versus 1982- 20û7 Data Models, Ceteris Paribus with Other Variable Set at October 2006 Levels

Figure 3

Simulated Catch per Trap Haul as a Function of Monthly Fishable Biomass with Post-2000 Data versus 1982-2007 Data Models, Ceteris Paribus with Other Variable Set at October 2006 Levels

Figure 4

Simulated Catch per Trap Haul as a Function of Soak Days for Production Function with Post-2000 Data and 1982-2007 Data Model, Ceteris Paribus with Other Variables Set at October 2006 Levels for Fall and with Biomass and Total Trap Hauls Set at One-Third October Levels to Represent Winter Conditions

Optimization Model

For the scenarios where either economic profits or net revenues are maximized the model maximizes the objective subject to a constraint that total catch is less than or equal to the observed catch for that year. The control variable for the optimization is the number of vessels fishing each month and the soak time. The number of trap hauls per vessel per fishing day is set at 250.4 The number of fishing days per month is determined by the soak time, the trap hauls per day, and the total number of traps in the water per individual, which is fixed at 800 traps (the current regulatory limit).

The model optimizes effort and catch over a one-year period, running July through June, since the new recruits (called shedders) on which the fishery is mostly based typically begin molting to legal size and thus into the fishery sometime in July. The model starts with the biomass of legal-size lobster that was estimated to have been available to the fishery at that time according to the stock assessment model. The biomass is then adjusted by taking out the simulated catch and adding in additional recruitment5 for the month (calculated by comparing the estimated legal-size biomass from the stock assessment and adjusting for the difference in actual vs. simulated catch).

Vessels are modeled as homogeneous, due to a lack of cost data to differentiate costs by scale of vessel or operation. I use estimates of average fuel use per trap haul from Driscoll (2008), bait use is specified as a function of soak time estimated from the port sampling survey, and estimates of fixed costs are drawn from a fixed-cost survey carried out by the National Marine Fisheries Service (2008). Fixed costs are set at $35,000—the average fixed costs for lobster vessels under 40 feet, which includes the majority of vessels in the Maine lobster fishery. Total fixed costs are determined by multiplying the fixed costs per vessel by the maximum number of vessels fishing during the year (i.e., the fixed costs are incurred for a vessel whether or not it fishes the entire year). I consider alternative “low” and “high” prices for bait and fuel. Low prices are representative of prices in recent years, and high prices reflect a 50% increase in price. Crew costs are calculated as 20% of revenues after subtracting fuel and bait costs. This is the common method of crew remuneration for operations with a single sternman, which make up the majority of the active fleet. Economic profit is calculated by subtracting from gross revenues the variable and fixed cost and crew share and also the opportunity cost of the captain’s time. I set the opportunity cost of the captain’s time at $15 per hour, which was the median hourly wage in Maine in 2009. This is probably a low estimate of opportunity cost, given that many captains are highly experienced and skilled fishermen; however, most do not have post-high school educations, and in many areas there are few job opportunities other than lobstering (Gulf of Maine Research Institute 2008).6

Catch per vessel fishing day is determined by the production function estimated with post-2000 data described earlier. Price is determined endogenously by the demand model, which uses the simulated catch and the actual observed U.S.-Canadian exchange rate, U.S. quarterly per capita personal income, and the percentage change in U.S. GDP from the prior year to determine the ex-vessel price. For purposes of comparing the simulated optimized revenues and profits to the observed levels, I use the observed catches and effort but use the demand model to simulate prices in both cases so as to make the comparison fair.

Under the profit maximization scenario, net revenue (revenue less variable costs) per pound of catch is substantially higher in the summer and fall than in the winter and spring. For example, under low cost assumptions, net revenues per pound in November are 20% to 30% higher than in May over the three years simulated. Under an ITQ system we would expect effort to be distributed so as to equalize net revenues per pound over the year , since otherwise an individual quota holder could increase his profits per unit of quota by shifting production to a month with higher net revenues per pound. I simulate the ITQ scenario by setting monthly effort to maximize net revenue subject to the constraint on total catch being less than or equal to the observed level, but also use an additional constraint that net revenues per pound are equal across months. Actual net revenues might of course vary across months for a variety of reasons including lack of perfect information, but this provides an approximation of rent dissipation due to the in-season depletion and congestion externalities.

Table 5

Simulated Annual Economic Profit (July-June) for July 2004-June 2007 for Low and High Cost Assumptions

III. Results and Discussion

The model suggests that annual economic profits7 with observed effort and catch levels during the three-year period from July 2004 through June 2007 ranged from 42% to 57% of what could have been achieved with profit maximizing monthly effort and catch levels (Table 5). If input costs were 50% higher, then gains from optimal management would also be higher, profit with observed effort and catch being only 22% to 46% of the profit maximizing level. The profit maximizing scenario decreases the fleet size and total effort and moves some effort from the late summer and fall into the winter and spring (Figure 5). However, catch remains fairly concentrated in the late summer and fall (Figure 6), because catchability is higher and biomass is higher because the legal-size fish stock (which is comprised mostly of lobster that molted to legal size in late summer) has not been fished down as much as it has later in the year or the following year prior to the next molt. The number of vessels fishing remains relatively constant, but the total trap hauls still decline in winter as a result of longer soak times and fewer fishing days per month. Under the profit maximization scenario, total production and total revenues actually decline relative to the observed catch and effort scenario. There are some gains in average price resulting from shifting some production into the winter, and thus smoothing production, but the increase in profits under the profit maximization scenario results primarily from decreased fixed costs and, in some years, a decrease in variable costs as well.

Under the ITQ scenario, effort and catch are much more concentrated in the fall than under the profit maximization scenario, though some effort and catch is shifted to the winter and spring relative to the observed levels (Figures 5 and 6). Profit under the ITQ management scenario ranges from 90% to 96% of maximum profit under low cost assumptions, but only 70% to 86% of maximum profits under high cost assumptions (Table 5). The ITQ scenario actually has higher revenues than the profit maximization scenario, but higher fixed costs and higher variable costs (Tables 6 and 7). With higher input prices the difference in variable costs and fixed costs increases, widening the margin between optimal and ITQ management. It is no-table that the costs for labor (i.e., crew share and opportunity cost of captain’s time) are much higher under the ITQ scenario and the observed catch and effort levels in comparison to the profit maximization scenario. While economists would generally consider payments for crew labor and the opportunity cost of captains’ labor a cost, fishery managers, politicians, and the industry tend to see them as income for the industry and may be as or more interested in increasing incomes to the fishermen as in maximizing economic profits. This is currently an owner-operated fishery, so profits actually go directly to the license holders, but this might not be the case under an ITQ system if quota leasing was allowed.

Figure 5

Monthly Effort Trajectories for Simulations from July 2004 to June 2007 for Alternative Objectives

Figure 6

Monthly Catch Trajectories for Simulations from July 2004 to June 2007 for Alternative Objectives

Table 6

Breakdown of Total Annual Revenues of Simulations with Varying Objectives and Low Cost Assumptions (millions of dollars)

Table 7

Breakdown of Total Annual Revenues of Simulations with Varying Objectives and High Cost Assumptions (millions of dollars)

The profit maximizing number of vessels fishing each month is quite sensitive to input price and fixed cost assumptions. As noted earlier, the information on fixed cost is poor and does not enable us to model the true heterogeneity of the fleet. It is likely that many of the vessels that are smaller and fish only in the summer and fall may have lower fixed costs than the vessels that fish most of the year. If this is the case, the optimal catch and effort trajectory may be closer to the status quo or the ITQ scenario than this analysis suggests, and the optimal fleet size in the fall might be considerable larger. This is an important question to resolve because the current number of participants in the fishery is far greater than the optimal number suggested by this analysis, and reducing participation is likely to meet strong opposition for social and political reasons—particularly in the “downeast” (northeastern) part of the state and in the islands where many communities are heavily dependent on the lobster fishery, and there are few other sources of employment.

The simulations presented here embedded the production function estimated with post- 2000 data. The production function estimated with the longer time series of data indicated a much stronger congestion effect and in-season depletion effect. This production function may be more applicable if lobster abundance declines to the levels in the 1980s and early 1990s. In this case optimal effort levels are likely to be much lower and the optimal effort and catch trajectory would also likely by smoother, that is, with relatively more catch and effort pushed into the winter and spring. The relative losses in economic profits under an ITQ relative to the profit maximization scenario might also be substantially higher due to stronger in-season depletion and congestion externalities.

The inverse demand function used to simulate price in these simulations reflects the fishery that has existed over the last 15 years in which most of the product is landed “soft” in the late summer and fall and is processed for meat or frozen in the shell. If more production were to be consistently shifted into the winter and spring when lobster quality is higher, the Maine industry would likely begin to develop new and larger markets for live lobster and might be able to maintain higher prices in those months than this model suggests. Thus over time the advantages of shifting production into the winter and spring might be expected to increase.

IV. Conclusions

This analysis suggests that the profitability of the Maine lobster fishery could be substantially increased by reducing fishing effort and perhaps even catch and by shifting some effort and catch from the late summer and fall into the winter and spring. However, because of higher catchability in the late summer and fall it still makes sense to concentrate production in those months despite the fact that the product is of lower quality and the concentration of landings drives down price. Some of the loss in revenues associated with concentrated landings might also be offset by “pounding” lobsters, that is, storing and feeding live lobsters in enclosures and selling them at times of year when landings are low and prices higher. Nevertheless, an optimal effort and harvest schedule almost certainly involves considerably fewer vessels.

It might be that profitability of the fishery could be increased simply by reducing the number of licenses and trap limits. Sufficient reductions would tend to shift some production into the winter and spring as the analysis suggests is optimal. However, fishermen might respond by decreasing soak times, and there is also substantial latent effort in the fishery now that might simply enter and dissipate any increased rents. Based on individual reporting that just became mandatory in 2008, Maine Department of Marine Resources estimated that out 6,492 license holders with trap tags, only 68% landed any lobster and only 42% landed more than 1,000 pounds (personal communication, Carl Wilson, Maine Department of Marine Resources). The majority of license holders use fewer than the regulatory maximum of traps much of the year (Gulf of Maine Research Institute 2008). Tradable trap systems, which have been implemented in a number of lobster fisheries elsewhere, including southern New England, can, with sufficient reductions in traps, eliminate latent effort, but they may also lead to various forms of input stuffing including shorter than optimal soak times. ITQs or perhaps allocations to harvest cooperatives might provide greater incentives to maximize profits; however, as noted above, ITQs may fail to incentivize the profit maximizing intraseasonal harvest schedule due to unresolved externalities. Because of the seasonality of the fishery and the congestion externality, it may be necessary to explicitly constrain the amount or proportion of catch by season as well as year to achieve maximum profitability. Even this may fail to resolve congestion externalities that relate to the fine-scale spatial heterogeneity of the fishery, suggesting some type of territorial uses rights, perhaps in conjunction with catch rights, may be an important facet of optimal management.

There is very little information currently available on input use, costs, and spatial dynamics of the lobster fleet in Maine. There is definitely substantial heterogeneity among license holders in terms of the size of vessels, annual landings, and, almost certainly, cost structure. There is also substantial variation in employment along the coast that impacts opportunity cost of labor, and opportunity cost of labor likely varies over the course of the year as well. More detailed information that describes this heterogeneity would almost certainly affect the “optimal” effort and catch distribution both over the year and along the coast. It may well be optimal to have a large part-time fleet fishing only in the late summer and fall if these individuals have low fixed costs and other employment opportunities at other times of the year. There are plans to collect more detailed economic data from the industry over the next few years that should shed light on these questions.

Despite these caveats, it seems quite clear that substantial reductions in effort (in terms of total traps fished and trap hauls) would increase profitability. This analysis suggest that it may be more profitable to harvest less lobster, even without accounting for the fact that leaving more lobster uncaught should increase catch rates and perhaps recruitment in future years. There are also environmental reasons to reduce effort. Interactions with marine mammals, particularly whales, would be reduced by reducing gear in the water, and this may in fact be required in future to protect endangered right whales.

The temporal and spatial heterogeneity in this fishery and the heterogeneity of the fishing fleet itself create both challenges and opportunities for increasing the value of the fishery, and this is, no doubt, true of most fisheries to varying degrees. Models of fisheries that ignore this heterogeneity may often fail to explain how changes in management will actually impact the value of the fishery and what steps may need to be taken to resolve externalities that simple controls on aggregate effort or catch (including ITQs) will not resolve. Furthermore it may often become apparent that heterogeneity of fishing activity in space and time may be an important characteristic of optimal exploitation, which should be explicitly addressed by the management system either through regulatory controls, incentives, or institutions such as local comanagement.

Acknowledgments

This work was done primarily while I was a research scientist with the Gulf of Maine Research Institute. Funding for this research came from the National Science Foundation, Coupled Natural and Human Systems Program, Award 0709527. I would like to thank Rich Ryan who contributed to an early version of this analysis, and Yong Chen, Jui-han Chang and Carl Wilson for providing the data necessary to estimate the production function.

Footnotes

  • 1 The lobsters have not filled out their shells so meat content is lower, and they are also more prone to mortality when shipping so that much of the product goes into processing rather than the live market.

  • 2 A survey of Maine lobster license holders by the Maine Department of Marine Resources in 2008 found that 56% of respondents saw the need to reduce the number of traps being fished in their area, and 51% favored an across the board cut in the trap limit (see www.maine.gov/dmr/rm/lobster/effortquest7-17-08.pdf).

  • 3 While data from before 1990 is available and we estimated models with data as far back as 1994, the market appeared to change substantially in the 1990s as the landings began to increase. We found that a model with post-1989 data provided the best predictive power for recent years.

  • 4 The number of traps hauled by individual per day varies widely from under 50 to as many as 400, but 250 is a common objective. Most lobstermen tend to try to haul about the same number of traps each day though they may adjust soak time and the number of days fished per month during the year.

  • 5 Recruitment refers to the growth of lobsters (through molting) to legal size. The recruitment added in the simulations is more accurately defined as net growth of the legal- size lobster stock, net of fishing but including natural mortality.

  • 6 A reviewer correctly noted that opportunity cost likely varies over the course of the year. Also since most lobstermen use the late winter and early spring when the fishery is at a low ebb to do vessel and gear maintenance, total effort in the fishery may not be quite as temporally concentrated as fishing effort.

  • 7 We refer to these as economic profits rather than rents, since optimization is over a year and the optimization does not account for effects on stock size in future years.

References