Skip to main content

Bioeconomic modelling of the fisheries of the English channel.

Deliverables

The methodological report (Pascoe, S. and Mardle, S. (Eds) 2003 'Efficiency analysis in EU fisheries: Stochastic Production Functions and Data Envelopment Analysis', CEMARE Report 60, CEMARE, University of Portsmouth, UK) includes a detailed overview of the theory underlying the estimation of stochastic production frontiers, comparisons of stochastic production frontiers with production functions, application of stochastic production frontiers in other industries, applications to the fishing industry, measurement of technical efficiency and inefficiency models, estimation and interpretation of scale elasticity and elasticity of substitution, the theory underlying DEA and applications to the fishing and other industries, and development of stochastic DEA models. The best practice for the alternative approaches is presented, along with a step-by-step guide to the use of the preferred software. As well as achieving the main objectives of the project, a number of innovations were developed as part of the project that was not originally anticipated. These included a method for estimating composite stock indexes using DEA and a method for separating changes in stock from pure technological change. These methods proved necessary for the estimation of several models for fisheries with limited stock information. The development of multi-output stochastic frontier models was also not originally anticipated, but was successfully applied to several of the fisheries examined. The models provided useful information not only on the level and distribution of efficiency in the fisheries, but also the degree to which fishers can target individual species within the catch. Groups of species can be identified with limited substitutability. This information is important when attempting to set compatible quotas. A number of different econometric software packages were examined. While several software packages can estimate stochastic production frontiers, only LIMDEP and FRONTIER can be used to estimate individual efficiency scores. Only FRONTIER is able to estimate inefficiency models directly and as such was the preferred software for the econometric analysis in the study. The existing literature in the Data Envelopment Analysis (DEA) field was reviewed in order to examine the different software for solving DEA models, and which software was most commonly used. While a wide range of DEA software was available, only a few products allow the researcher to add or exchange different constraints to the model. The GAMS software was found to be the most adaptable, as it allows you to build different DEA models formulated by the researcher. GAMS is not a specific DEA software, but is a generalised optimisation software. Methodological work was also undertaken in terms of software development. A windows based version of the econometric software used for the estimation of technical efficiency was developed for use by the partners (and the broader research community). The advantage of the revised software is that many of the standard tests and transformations are automated, simplifying its use. The revised software will make efficiency estimation easier for other research teams in Europe involved with fisheries or other industries.
The methodological report (Pascoe, S. and Mardle, S. (Eds) 2003 'Efficiency analysis in EU fisheries: Stochastic Production Functions and Data Envelopment Analysis', CEMARE Report 60, CEMARE, University of Portsmouth, UK) includes a detailed overview of the theory underlying the estimation of stochastic production frontiers, comparisons of stochastic production frontiers with production functions, application of stochastic production frontiers in other industries, applications to the fishing industry, measurement of technical efficiency and inefficiency models, estimation and interpretation of scale elasticity and elasticity of substitution, the theory underlying DEA and applications to the fishing and other industries, and development of stochastic DEA models. The best practice for the alternative approaches is presented, along with a step-by-step guide to the use of the preferred software. As well as achieving the main objectives of the project, a number of innovations were developed as part of the project that was not originally anticipated. These included a method for estimating composite stock indexes using DEA and a method for separating changes in stock from pure technological change. These methods proved necessary for the estimation of several models for fisheries with limited stock information. The development of multi-output stochastic frontier models was also not originally anticipated, but was successfully applied to several of the fisheries examined. The models provided useful information not only on the level and distribution of efficiency in the fisheries, but also the degree to which fishers can target individual species within the catch. Groups of species can be identified with limited substitutability. This information is important when attempting to set compatible quotas. A number of different econometric software packages were examined. While several software packages can estimate stochastic production frontiers, only LIMDEP and FRONTIER can be used to estimate individual efficiency scores. Only FRONTIER is able to estimate inefficiency models directly and as such was the preferred software for the econometric analysis in the study. The existing literature in the Data Envelopment Analysis (DEA) field was reviewed in order to examine the different software for solving DEA models, and which software was most commonly used. While a wide range of DEA software was available, only a few products allow the researcher to add or exchange different constraints to the model. The GAMS software was found to be the most adaptable, as it allows you to build different DEA models formulated by the researcher. GAMS is not a specific DEA software, but is a generalised optimisation software. Methodological work was also undertaken in terms of software development. A windows based version of the econometric software used for the estimation of technical efficiency was developed for use by the partners (and the broader research community). The advantage of the revised software is that many of the standard tests and transformations are automated, simplifying its use. The revised software will make efficiency estimation easier for other research teams in Europe involved with fisheries or other industries.
When the scores of single output SPF analysis and multiple-output DEA analysis are compared, the effect of random error is reduced and the TE scores are closer in magnitude than when single output SPF and DEA results are compared. This is due to the fact that the multiple-output measures provide several pieces of output information against which each vessels' activity can be ranked to determine where the efficient frontier lies and how each vessel compares to it, given its input level. In comparison, the single-output measures contain information on only one output. For example, in an analysis involving single-output measures, if one vessel caught very large amounts of one species whilst other vessels fishing the same gear in the same month caught 'normal' amounts of this species (and all vessels caught normal amounts of other species), the other vessels would be ranked much less efficient if a single-output measure is used as compared to a multi-output measure. In this situation, a multi-output measure would determine that the other vessels caught normal amounts of the other species, but were just not as lucky to catch so much of the high value species. Thus their DEA scores would be higher and more accurate under the multi-output measure analysis, as compared to the single-output measure. Consequently, the greater level of information available in the multi-output analyses enables a more accurate estimation of efficiency scores. Another possible explanation for the higher estimates with multi-outputs may be related to the estimation process. With a single output measure, the boat's performance is compared to the boat that has the highest overall level of output (catch) irrespective of the composition of the catch. When the multiple outputs are considered, boats are compared with other boats that have a similar catch composition. As a result, the boats appear to be more efficient when compared with other boats with similar catch compositions (which are undertaking similar activities) than when compared to the boat with the greatest overall catch irrespective of catch composition (and may be undertaking quite different activities). The number of observations available largely limits the number of outputs that could potentially be included in the analyses, as too few degrees of freedom in the data results in inflated TE scores. A number of new techniques were also applied. These included an analysis of mix efficiency using DEA, a methodology for deriving output specific measures of technical change in multi-output industries such as fishing, and the use of multi-output stochastic frontier models. As well as providing information on the distribution of efficiency in the fisheries, the results of the analyses provided interesting insights into the operation of the fisheries examined, including distortions in input use as a result of input controls, and the potential for output substitution in multi-species fisheries (i.e. the degree to which fishers can alter their output mix). Multi-output analyses were undertaken by all project partners. The multi-output DEA results were compared with the single output efficiency estimates from SPF and DEA and the results are described in Pascoe, S., Tingley, D. and Mardle, S. (Eds) 2003 'Single output measures of technical efficiency in EU fisheries', CEMARE Report 61, CEMARE, University of Portsmouth, UK.
Efficiency measurement and related analysis can be of significant benefit to fisheries managers in helping them to formulate policy measures. For example, knowledge of the extent of inefficiency present in a fleet, and its progression over time, will help managers to predict the true impact, and required magnitude, of proposed input controls or a fleet reduction program. Efficiency analysis can also play an important role in tuning the catch function used in biological models to express fishing-related mortality. For most fleet segments analysed in the study, the average level of efficiency was between 0.7 and 0.8, although average efficiency scores as low as 0.48 were observed (a core of 1.0 indicates no inefficiency). There was no apparent trend in terms of one fleet segment being more or less efficient than others on average. If the existing set of fleet inputs were used at their full efficient level, then average output could increase by between 20 and 40 per cent in most of the fleet segments examined. In output-controlled fisheries, improvements in efficiency will result in lower effort levels and, consequently, decrease capacity utilisation. This, in turn, creates incentives to increase the level of non-recorded (black market) landings. In input-controlled fisheries, managers will need to respond by further restricting the level of inputs, preferably through fleet reduction. From the analysis of the factors affecting efficiency, a number of factors could be identified that may lead to efficiency increases. These factors appeared to differ substantially from country to country even when common information was available to be tested. For example, the introduction of bigger boats was found to increase average efficiency for the Danish seine fleet, but for many other fleet segments, larger vessels were generally less efficient than smaller vessels. Newer vessels were more efficient in the Danish seine and Greek inshore fleet segments, but not in the English Channel fleet segments. Skipper age was generally negatively associated with efficiency, which has implications for the success of early retirement programs. Fishers from families with a fishing history tended to be more efficient than first generation fishers. This implies that key fishing skills are passed down through families. Consequently, there appears to be validity in the arguments for policies to preserve fishing communities to ensure the persistence of fishing skills. Knowledge of the extent of inefficiency present in fleet segments is crucial in determining the form and impact of national decommissioning programmes. Assuming that the least efficient vessels are the first to exit, then a greater than proportional decrease in fleet size would be required to achieve a given reduction in output. For example, for the Greek inshore fleet, the fleet size would need to be reduced by almost 40 per cent to achieve a 20 per cent reduction in output. Similarly, a 20 per cent reduction in the output of the North Sea nephrops fleet would require an almost 30 per cent reduction in fleet size. The effectiveness of 'days at sea' restrictions is dependent on the elasticity associated with the variable input levels derived from the stochastic production frontiers. In most cases, the elasticities for the industrial fleets were around 1, indicating that the days at sea restrictions would result in a more or less proportional decrease in output, as intended. However, the impact of the days at sea restrictions may, in fact, be 33 per cent greater for English North Sea seiners than for Danish seniers for an equal reduction in days fished. The study produced a number of other measures that can provide useful information for fishery management. These include production elasticities, elasticities of substitution - both input and output - as well as measures relating to the scale of operation.
The estimation of technical efficiency is based on the observed, or recorded, landings. In many of the fleets examined, incentives exist, either as a result of management or through constraints on the available hold space, to discard part of the catch. In some cases, incentives may exist to land part of the catch illegally, particularly if the probability of detection is low. A common feature of both discarded or illegally landed catch is that it is not recorded. As a result, the observed catch does not necessarily equate to the actual catch. Provided sufficient observations exist in which all catch taken is landed and recorded, the production frontier underlying both the DEA and SPF approaches should represent the true relationship between inputs and outputs. However, deviations from this frontier will be attributed to inefficiency when, in fact, they may be due in full or in part to non-recorded catch. Non-recorded catch is effectively a third error term in the production process, the first being random error and the second inefficiency. A difficulty in separating out this third term is that its magnitude is not independent of both the level of inputs or the other two error terms. A highly efficient vessels using a high level of inputs in a 'lucky' year is more likely to exceed his or her quota or hold capacity, and therefore either discard some of the catch or land it illegally, than a less efficient vessel using fewer inputs. In this study, a number of different approaches were examined in order to try and estimate the possible extent of non-recorded catch using DEA. The results indicated that 'problem' species could be identified, and also that the magnitude of the problem could be estimated. However, the level of precision was not particularly high. One approach, however, appeared to be reasonably reliable in identifying the worst offenders. The analyses undertaken on non-reported catch demonstrated potential that could be further developed. Such further analyses, if successful, could potentially reduce surveillance costs in fisheries by ensuring that resources were targeted in the right areas.