Technical efficiency effects of input controls: evidence from Australia's banana prawn fishery

Abstract

This paper provides the first ex post estimates of the effects of input controls on technical efficiency in a fishery. Using individual vessel data from the northern prawn fishery of Australia for the years 1990–1996 and 1994–2000, stochastic production frontiers are estimated to analyse the efficiency impacts of input controls on engine and vessel size. The results indicate that technical efficiency is increasing in a measure of vessel size and engine capacity that was controlled by the regulator from 1985 to 2001, and decreasing in an unregulated input, gear headrope length. The study shows that fishers have substituted from regulated to unregulated inputs over the period 1990–2000 and technical efficiency has declined coincident with increasing restrictions on vessel size and engine capacity. The decline in technical efficiency indicates that the goal of the regulator to increase economic efficiency has not been realized.

Acknowledgements

The authors would like to thank Steve Beare, Peter Gooday, Janet Bishop and Bill Venables for data access and helpful comments on this work and the nature of the Australian northern prawn fishery in general. We are also grateful for helpful comments by Kevin Fox and Dale Squires. Funding from the Fisheries Resources Research Fund is also gratefully acknowledged. This paper is dedicated to the memory of David Timcke, a colleague and friend.

Notes

Alam et al. (2002) analyse the issue of licence limitation, especially area licences, in the context of tropical fisheries.

Townsend (1990) provides a classic survey of the evidence.

Squires (1987) was the first to empirically measure input substitution in fisheries using a flexible functional form. Empirical studies that have assessed the impact of input controls have measured the level of input substitution and rent dissipation (Dupont, 1991), input substitution from regulated to unregulated inputs (Campbell, 1991; Del Valle et al., 2003) and the contribution of unmeasurable inputs to efficiency (Pascoe and Coglan, 2002).

We are prevented from estimating allocative efficiency in this fishery because of lack of data on input prices. We note that technical efficiency is necessary, but not sufficient, to ensure economic efficiency.

Green (1993) and Førsund et al. (1980) are useful surveys of the stochastic frontier approach. Kumbhakar and Knox Lovell (2000) is also highly recommended.

This is especially true for tiger prawns, but the tiger and banana prawn fisheries are treated as a single fishery for management purposes.

For the specifications in Section IV, likelihood ratio tests (not reported) reject a time trend in the technical inefficiency model, so the effect is ignored here.

See NPFAMP (2000). Fuel is related to engine power and to a lesser extent to vessel size. In the data set, boat fuel expenditures are available only as an aggregate over tiger and banana prawn output. The measure of fuel used for banana prawns is thus obtained by multiplying total fuel expenditures by effort days in banana prawn production as a fraction of total effort days in banana and tiger prawn production.

Results from the translog and the pre-test are available from the authors upon request. It should also be noted that although the translog specification allows more scope for substitution, the input restrictions used in the NPF make conventional measures of elasticities of substitution inappropriate (Dupont, 1991).

Equation 5 is comparable to the approaches used in Kirkley et al. (1995) and Sharma and Leung (1999). Given the available data, the first paper uses days at sea, stock abundance and labour to estimate sea scallop production in the mid-Atlantic and the second paper uses trip days, crew size and other inputs (fuel, bait, ice, etc.) to estimate output in the longline fishery in Hawaii. It should be mentioned that Kirkley et al. (1995) use a two-step procedure to estimate technical inefficiency, rather than estimating the stochastic production frontier and technical inefficiency effects directly in a single step. The latter provides more efficient estimates (see Battese and Coelli, 1995 and Kumbhakar et al., 1991) and, moreover, the two-step procedure is inconsistent with the assumption of identically and independently distributed technical inefficiency effects.

Fuel input quantities and prices are not available by vessel.

In the Carpentaria region, rainfall in the summer months (December to March) in the years 1990, 1992 and 1994, obtained from the CSIRO, was 2654, 3445, 3550 millimetres compared to average for all other years in the sample period of 4550 millimetres.

The specification given by Equation 5 assumes that weather effects, as normally distributed and unbounded random variables, are accounted for (with adjustments to coefficient values for inputs through the choice of relevant dummy variables) in the disturbance term v _it , rather than in the technical inefficiency model. Managers at AFMA also believe that there has been a downward trend in the stock of prawns over the 1990s, but do not have the data to test their hypothesis.

A-units and fuel expenditures (which are known to vary by engine size and power) are correlated and thus both variables are not included together. Nevertheless, replacing fuel with A-units gives comparable final estimates in all cases.

Diagnostic tests, available from the authors, also indicate that if the input variables are included in levels in Equations 7 and 8, there is no specification error.

Identical estimates were also obtained using GAUSS. The GAUSS program was also used to derive diagnostic tests.