Essays on Dynamic Behavior of Harvesters in Fisheries
Dynamics is a key aspect of analyses on renewable natural resources, as it replenishes over time. While long-run dynamic aspect of harvester behavior has been traditionally explored, microeconomic short-run harvester behavior analyses have been attracting more attention recently. One of the main reasons of this attention is due to the demand on analyzing and predicting harvesters’ reactions to regulations and management policies. A management policy could lead to an unwanted result because of undermined or mis-specified harvesters’ incentives. Particularly, their forward-looking behavior in short-run, such as a trip or a season, is important to consider because their behavior can be largely changed by a policy. In this dissertation, three different types of models are constructed to analyze harvesters’ short-run dynamic behavior. In the first chapter, harvesters’ choice of trip duration is analyzed using dynamic discrete choice model. Production function, or harvest function in natural resource economics, is often estimated with temporally aggregated data. However, such estimations can be misleading if variable inputs are dynamically determined within a time period of the data, because the variation within a period is not taken into account. In this study, using a data from a longline fishery, I demonstrate a case that cruise-level production is determined not only by use of quasi-fixed inputs, but rather by dynamic consideration of the rate of daily harvest, balancing the quantity and quality of harvest to maximize their cruise level revenue. This response is modeled as a daily optimal stopping problem, with the state variables representing the decreasing freshness of fish caught on each previous day of the cruise. I estimate trip duration decisions based on unusually detailed daily logbook data on a Japanese longline fleet. The dynamic discrete choice problem is modeled with a conditional choice probability (CCP) estimator, which estimates the reduced form of CCP and transition probabilities in the first step to calculate the continuation value, and estimate the structural parameter using the calculated continuation value in the second step. The predictability is improved avoiding over-fitting in flexible logit to estimate CCP in the first step with elastic-net logit, a machine learning method. The results show harvesters are particularly sensitive to freshness deterioration of fish on board for more days, and are more likely to terminate their fishing cruise when more fish is caught around 15 or more days ago. This suggests that catching power defined by quasi-fixed inputs is not fully utilized due to a dynamic consideration of fish quality, and that a management strategy based solely on technical efficiency will systematically over-predict actual catches. The second chapter discuss the harvesters’ decision on weekly target choices in a horizon of a season. This study develops an empirical model of the harvesters’ dynamic individual fishing quota (IFQ) use considering avoidance of bycatch, an incidental catch of non-targeted species, with the insight from a theoretical model. While theoretically shown, empirical analysis on allocation of IFQ over a season by harvesters have not been well-studied due to complexity of the seasonal dynamic problem. In this paper, we focus on participation and target species as harvester’s margin, which are flexibly chosen under IFQ. To explore the incentive, we theoretically modeled the harvesters’ behavior which purposes to maximize seasonal profit under constraints of the regulations. The solution motivates us to incorporate the dynamic quota use in a simple discrete choice model to estimate the harvesters’ choice. To link the shadow cost of quota in our theoretical model and harvester behavior in the data, we construct a variable which captures harvesters forward-looking decision. The application of this empirical model is implemented with an offshore fleet in Alaskan groundfish fishery, which has appropriate feature to apply the model such as multiple target choices and individual quota. The result indicates that the dynamic variable adjusts the incentive to catch pollock for revenue, and the estimates of the coefficients on bycatch rate supports the dynamic avoidance behavior of bycatch. The third chapter extends the model developed in the second chapter by incorporating spatial choice of harvesters. Because the main target and bycatch species are spatially heterogeneously distributed, the harvesters are possible to avoid bycatch while targeting the main species provided the participation decision in a given week. The harvesters decisions are modeled as two-stage process: target decision in the first stage, and location as the second. We adopt nested logit model to estimate this two-stage process. As a result, spatial choice is an important margin to avoid bycatch dynamically, but it is not in B season as the estimation result is not consistent with utility maximization theory in B season. In A season, the timing of high bycatch rate and the timing of matured pollock roe that has high value overlap, hence the harvesters cannot flexibly change the timing. Instead, use the spatial margin to avoid bycatch while maximizing the revenue from pollock. In B season, there is more flexibility in timing choice. The harvesters rely on timing choice rather than spatial choice in the B season. Both avoidance behavior is driven by the forward-looking expectations since the interactions with dynamic variables are key factors to explain the harvesters’ behavior.
- Economics