abare.gov.au - research report 09.5

6	Case study: the northern prawn fishery

The case study for the northern prawn fishery includes the MEY analysis for the tiger prawn component of the fishery and the stochastic production frontier estimated for the banana prawn component of the fishery. An overview of the fishery is discussed in the first section.

Overview of the northern prawn fishery

The northern prawn fishery (NPF), first established in the late 1960s, is one of Australia’s most valuable fisheries. The fishery occupies an area of 771 000 square kilometres off Australia’s northern coast, extending from the low water mark to the outer edge of the Australian Fishing zone (AFZ) along approximately 6000 kilometres of coastline between Cape York in Queensland and Cape Londonderry in Western Australia (see map 1).

Although there are more than fifty species of prawn that inhabit Australia’s tropical northern coastline, only about nine species are caught. Three species (the white banana prawn Fenneropenaeus merguiensis, the brown tiger prawn Penaeus esculentus, and the grooved tiger prawn P. semisulcatus) account for almost 95 per cent of the total annual landed catch weight from the fishery (ABARE 2008a). Endeavour prawns (Metapenaeus endeavouri and Metapenaeus ensis) and the red-legged banana prawns (F. indicus) form most of the remainder of the catch. Other commercial catch includes the giant tiger prawn (P. monodon), western king prawn (Melicertus latisulcatus) and the red spot king prawn (Melicertus longistylus) (AFMA 2002).

The gross value of prawn production in the NPF in 2006-07 is estimated to be A$64 million with a total harvest of about 5100 tonnes. Nearly 90 per cent of all prawn output is exported to Japan and Asia (ABARE 2008a).

In 2007-08, 52 vessels actively participated in the NPF. All vessels are purpose built twin-gear otter trawls and generally range in size from 14 to 29 meters, with the most common boat size between 18 and 25 meters (AFMA 2008a). Most boats operate between 80 and 90 per cent of the time available for fishing, with breakdowns and unloading (to mother-ships) accounting for much of the remaining time. The fleet is technologically advanced, employing modern packing and freezing capabilities and sophisticated fishing aids such as echo sounders and satellite global positioning systems and plotters.

The banana prawn fishery is primarily located in the eastern waters of the Gulf of Carpentaria, in isolated grounds along the Arnhem Land coast and in Joseph Bonaparte Gulf. Annual catches since 1983 range from 2200 to 6600 tonnes per year (Caton and McLoughlin 2000 and ABARE 2008a). The white banana prawn accounts for more than 80 per cent of all banana prawn catch. The spawning of banana prawns generally occurs in offshore areas, while recruitment of prawns to the fishery usually takes place in late spring. Banana prawns form dense aggregations (boils), which are easily spotted, allowing for rapid harvesting. The fishing season (with mostly daytime catch) starts around April and lasts only a few weeks. Single aggregations of prawns usually contain four to 180 tonnes, but can be as high as 400 tonnes. Highest seasonal catches generally follow higher than average rainfall during the preceding summer (see Staples and Vance 1986). Given the ease in harvesting, trawls for banana prawns are typically of a short, 10 to 20 minute duration.

Total effort attributed to the fishery in 2006-07 was approximately 11 100 boat days, comprising 7400 and 3700 boat-days for tiger and banana prawns respectively (estimated from the NPF Surveys carried out by ABARE in 2008). Although it is clear that potential catch is highly dependent on weather patterns, the relationship between catch and future stock size for banana prawns is not. As yet, there is still no conclusive evidence that effort affects future stock abundance in this fishery (see Staples and Maliel 1994), although very recent catches below expectations have caused concern. In fact, the maximum sustainable yield for banana prawns is estimated to be 4000 tonnes, which is roughly equivalent to the average catch over the past decade (Taylor and Die 1999).

The fishery has historically been managed with input controls such as gear and vessel restrictions, limited entry, area closures and seasonal closures. A brief history of the management arrangements in the fishery is outlined in table 1. Since 2000, the main management tool has been input controls in the form of restrictions on the length of net headrope allowed to be towed in the fishery. Gear units allocated to each operator specify the length of headrope allowed and operators are free to buy, sell or lease these gear units.

Seasonal closures in the fishery create two distinct fishing seasons, a banana prawn season and a tiger prawn season. In 2006, the banana prawn season was open from 9 April to 21 May and the tiger prawn season was open from 1 August to 15 November. In recent years the fishery has been closed during August. However in 2005, AFMA agreed to include August in the tiger prawn season to minimise catches of tiger prawn in the banana prawn season.

Initially, management efforts were confined to limiting entry and imposing controls on boat replacement through the 1977 and 1980 three year plans. Adoption of A-units as the measure of capacity and B-units as the effective right for a boat to fish in 1984 was part of an attempt to control the increasing effort that resulted from replacement of old boats with new (AFMA 1999). In 1986, data compiled by CSIRO showed a serious decline in brown tiger prawn stocks in the western Gulf of Carpentaria. A Voluntary Adjustment Scheme, involving buy back of A-units, was developed largely in response to that finding and a consequent CSIRO proposal for an immediate 25 per cent cut in effort to protect pre-spawning tiger prawns (Pownall 1994). Initially the intent was to reduce total A-units in the fishery to 70 000 by the start of the 1990 season. Any shortfall would be met by a compulsory acquisition at the start of the 1990 season. However, industry opposition and eventual Senate rejection eliminated the compulsory element of the buy back. The voluntary element was extended to acquisition of B-units, so effectively buying out the right to operate a boat in the fishery.

The voluntary buy back scheme was refinanced and extended in 1990. An initial target of
50 000 A-units by the beginning of 1993 was set, later being amended to 53 844. If the target was not reached through voluntary buy back, the residual was to be met by a proportional surrender of A-units. The target was met by a combination of voluntary and compulsory acquisition, with 53 844 A-units and 132 B-units remaining in the fishery on April 1, 1993. Those units were rolled over as Class A and Class B Statutory Fishing Rights (AFMA 1999).

Throughout the period of the Voluntary Adjustment Scheme a series of other policy changes was implemented, in part in recognition of limited effectiveness of a slowly proceeding reduction in A-units. Those changes included the introduction of gear restrictions and both a daylight trawling ban and mid season closure for tiger prawns. Since 1993, two major changes in management have been implemented. In 1999 the basis for input constraint was changed from boat size and power (A-units) to headrope length (gear units), with a concurrent reduction in gear units of 15 per cent. In 2002 gear units were reduced by a further 25 per cent and the tiger prawn season length was further reduced. Further reductions followed in 2005. A new target level of catch of maximum economic yield (MEY) to replace maximum sustainable yield (MSY) was accepted by the AFMA Board in 2004 after being recommended by the Northern Prawn Fishery Management Advisory Committee (NORMAC) (AFMA 2004b, NORMAC 2004). This objective implies that the fishery be managed so that effort, catch and thus stock biomass are at levels that allow net economic returns to be maximised.

Several things are notable about the sequence of management regimes in the fishery. Each of the changes was made in recognition that the system it replaced had failed to constrain effective effort sufficiently to protect prawn stocks. Where effective effort was reduced by management change, the primary reduction appeared to be short lived. This effect, and one of the primary reasons for it, is illustrated in figure g. Fishing power, measured as the average catching ability of a boat in a day’s fishing has risen rapidly and consistently over time. The rise in fishing power is the result of continuous improvements in technology, input combinations and knowledge. The acquisition of improved scientific knowledge of the fishery, along with the observation of declining catches has made it increasingly clear over the past few years that prawn stocks are not being conserved and catches are not being controlled.

The combination of policy changes in 1999 and 2002 appears to have temporarily slowed the increase in fishing power as well as contributing to a rapid fall in total days fished. Total days fished in 2002 were 55 per cent of the 1998 level, probably as a result of a combination of policy change and significant falls in prawn prices since 1998. A return to the average annual rate of growth in fishing power that applied from 1970 to 1998, more than 6.7 per cent, would see effective fishing days return to the 1998 level in less than 9 years, even if no other adaptations were made. It is evident from figure g that the effects of input policy change on fishing power are never more than temporary.

Maximum economic yield analysis for the northern prawn fishery

This analysis constructs MEY estimates for the tiger prawn component of the northern prawn fishery, illustrating the importance of MEY and fleet size. The theoretical context is a bioeconomic model calibrated by specific fishery parameters. The exercise also nicely illustrates the construction of standard bioeconomic models of a fishery. Kompas and Che (2004) and Rose and Kompas (2004) provide a more detailed analysis.

This section provides the biological model, and in particular the relationship between spawning stock-recruitment and spawning stock-biomass as well as the effect of fishing harvest and mortality.

Biological model

The spawning stock-recruitment relationship is modelled according to Ricker’s equation (Ricker 1954) or:

where R_t is the total number of recruits produced in year t and

is the spawning stock of the previous year (estimated as the number of prawns), and where

and

are parameters that determine the relationship between recruitment and the spawning number of the previous year. The measure

reflects uncertainty or the stochastic behaviour of the spawning stock-recruitment relationship.

Spawning stock is taken as a proportion

of the total female stock, assuming that female prawns constitute half of the total stock of prawns, so that:

Following Penn et al. (1995) and Wang and Die (1996) the spawning stock

is the result of annual recruitment R_t and fishing effort, defined as:

where F_t is fishing mortality at year t and m is the annual natural mortality rate.

Following Wang and Die (1996) fishing mortality at year t is defined by:

where q is the ‘catchability coefficient’ and E_t is fishing effort at year t. Fishing effort is determined as total ‘standard’ boat days in the fishery, which is a multiple of total ‘standard’ boats (B_t ) and nominal fishing days in the season (N_t ). In this study one unit of fishing effort is defined as the daily effort of a ‘standard’ boat. A standard boat is used to avoid the problem of equating boat day units between large and small vessels. In practical terms, this capacity can be measured by boat engine power and a measure of hull units or the length or the weight of boat. In this study boat capacity is measured in terms of A-units, or a simple linear combination of a kilowatt of engine power and a cubic metre of hull. The measure of boat capacity used here is the same as that used to specify A-units in the NPF up until the introduction of gear units in July 2000. Define a standard boat size as A units so that the total standard boat numbers at year t is given by

where M is the number of boats in the fishery, and A_it is the capacity of boat i in year t. With technological change, fishing mortality at year t is simply given by:

where TEC_t is the variable that measures the change in technology at year t.

Following Wang and Die (1996) the annual catch h_t in tonnes is approximately defined as:

Catch increases asymptotically to a maximum of

as fishing effort tends to infinity (Wang and Die 1996).

Based on equation 6.7 the catch per unit of effort (CPUE) is given as

where

represents stochastic error in CPUE. For input controls random error in CPUE is generally captured by variance in fish harvest. For output controls the error is generally captured by variance in fishing effort (E_t ).

Bioeconomic model

Annual total revenue of the fishery is defined as the multiple of annual fish harvest and the annual (average) price of fish, so that

where p_h is the price of fish drawn from an inverse demand curve. Following Danielsson (2002) and Campbell, et al. (1993) this price is determined by

where e is the elasticity of demand for catch and P_o is the unit price of the catch when the volume of the catch is H_o.

Annual total cost is assumed to be the sum of labour, material, capital and other costs. Labour costs generally include a share of total fish revenue and packaging and gear maintenance expenditures directly correspond to total fish revenue. Capital costs are defined by the cost of capital calculated as a sum of depreciation cost and the annual opportunity cost of boat capital value. Capital costs and other costs (of which fuel is a major component) are assumed to depend on fishing effort so that total costs can be expressed as:

c_L and c_M is the share cost of labour and materials per each Australian dollar of output respectively; and c_K and c_o is the average capital and other costs per unit of effort respectively. The average capital cost of a unit of effort (c_K ) is estimated by dividing total capital costs by total effort. Average other costs (c_o ) per unit of effort is estimated by dividing total other costs by total fishing effort. The value of a represents a fixed cost component.

Annual fishery profit is defined by subtracting annual total cost from annual total revenue. From equations 6.9 and 6.11 annual profit is expressed as

Fishing effort is defined as total ‘standard’ boat-days with the number of ‘standard’ boats (B_t ) as computed in equation 6.5.

An objective of a fishery is to maximise aggregated profits over time. Although seasonal and area closures are important in almost every fishery, including the NPF, the main concern here is the choice between input and output controls. Under input controls the fishery authority targets overall effort levels through a combination of input restrictions, limits on technology and limitations on days fished. With output controls the authority sets a catch quota with vessels free to adjust their effort levels to meet total allowable catch. Assuming that effort levels are observable and enforceable, the problem for an input control regime is to maximise:

through a choice of or variations in effort. The choice of the control variable E_t is set by the length of time or nominal days fished. Introducing discounting and substituting 6.6, 6.7 and 6.1 into equation 6.13, gives

for

the discount rate and where

is the net present value of profit at year t.

For output controls the problem is to maximize

through a choice of harvest (h_t ) and where

is the net present value of profit at
year t.

A solution requires substituting from equations using equations 6.1, 6.6 and 6.7 to ensure that equation 6.15 is a function of catch or harvest only. Larger stock values clearly lower the costs of fishing or the amount of effort required to meet a catch quota. Solving equation 6.15 also requires that spawning stock at the period o

be known. In all cases the appropriate transversality condition at time t=T is that the value of profits is zero (Clark 1990).

Optimal solutions and maximum economic yield analysis

Parameters used in the model are indicated in Kompas and Che (2005), for base year 2000. Details of these parameters and their sources are discussed below. The initial spawning stock

for brown and grooved tiger prawns in the year 2000 is estimated from spawning stock indexes provided by CSIRO (2002a). The conversion from indexes to millions of prawns was done by using a coefficient linking spawning stock in terms of indexes at 1993 (CSIRO 2002a) with the spawning stock in terms of millions of prawns indicated by Wang and Die (1996). The recruitment of prawns is estimated from the initial spawning stocks following equation 6.1.

Parameters

and

for the biological relationship described in equations 6.1 and 6.3 are provided by Wang and Die (1996). The annual mortality rate (m) follows Wang (1999) and Wang and Die (1996). The sex (male and female ratio) is 1:1, following Wang (1999), Wang and Die (1996) and Dichmont et al. (2003b). The value of

for brown tiger and grooved tiger prawns is computed as the average of the monthly percentage of female spawning over the ‘biological year’, from August to March (Crocos 1987a and 1987b).

Parameters for the catch, stock and effort relationship (

and

) as defined in equation 6.7 are provided by Wang and Die (1996). The coefficient of effort distribution between brown and grooved tiger prawn fishing follows research by Dichmont et al. (2003b). The actual number of vessels operating in the year 2000 was 120 (AFMA 2002). Standard vessel fishing capacity is 400 A-units, the average obtained from CSIRO (2002b). The catchability rate of one unit of fishing effort is obtained from Wang (1999). This catchability rate is given for 1993 with a natural mortality rate of 0.045. For the year 2000 this number is adjusted to account for effort creep and technological change, based on Dichmont et al. (2003a). For the cases of ‘basic high’ and ‘spatial high’ it is estimated that fishing power has increased at around 2 to 2.4 per cent per annum. Therefore in this study the adjustment to the catchability parameter at 2000 is 19 per cent higher than in 1993. For years subsequent to 2000 it is assumed that there is no effort creep.

The initial price of tiger prawns is computed from ABARE (2008a). The initial catch and price of tiger prawns or H(o) and p(o) is based on values at year 2000. Based on statistics for catches and prices (ABARE 2008a), and given that 90 per cent of tiger prawns are exported to Japan, the coefficient of flexibility between supply and price is estimated at 15. This number is based on the empirical relation between prawn supply and demand in Australia (based on ABARE 2008a). The proportion of revenue share by labour, materials and other costs are based on data collected in economic surveys of the NPF carried out by ABARE.

The average capital cost per unit of fishing effort is computed from total capital cost and total fishing effort. The value of vessel capital is the market value at year 2000 of vessel, hull, engine and onboard equipment as of July during the survey years. Capital costs are defined by the user cost of capital calculated as a sum of depreciation cost, the annual opportunity cost of the total capital value and the difference in boat value between season opening and closing time in a given year. Vessel depreciation is based on the ‘discrete diminishing value’ approach. The opportunity cost for vessel capital was derived as the multiple of the nominal interest rate and vessel capital.

The average of ‘other costs’ per unit fishing effort is computed from total other costs (mainly fuel costs) and total fishing effort. Both average capital and other costs per unit of fishing effort are measured in real prices, base year 2000. The Consumer Price Index was obtained from the DOL (2008).

Using a genetic algorithm the optimal solutions described in equations 6.14 and 6.15 are obtained and reported in table 2. The choice variable for output controls is harvest or catch, for input controls, effort. Several models are solved. The first is a base model, assuming no uncertainty in effort or catch, setting the variance in stock and CPUE equal to zero. The second model obtains results based on the actual variance in stock and CPUE in the NPF, using the variance in the residuals from the estimates above. Finally, the case of discounting is considered.

The time horizon for the optimal process used in this study is 50 years, long enough to guarantee that optimal results are sufficiently close to their steady state values before diverting to meet a terminal condition in year 50. The terminal condition is such that the value of profits at year 50 goes to zero. The issue of what discount rate should be used in a Commonwealth fishery is contentious. Firms in the industry would prefer a rate that reflects the opportunity cost of investment in vessels and fishing capacity. The fishery manager would likely prefer a more ‘conservationist’ approach, or even a zero discount rate. For a Commonwealth resource some rates in between may be the most appropriate. In this study the case of zero discounting and a 3 per cent discount rate is used and compared.

The optimal solutions for the case with a discount rate of 3 per cent are reported in table 2. Both cases indicate more catch earlier in the planning horizon and consequently smaller ‘near’ steady state stocks than in cases without discounting.

In the case of stochastic recruitment and CPUE, optimal results show output controls dominate effort controls in the NPF for tiger prawns (there is no difference in a deterministic setting). Fishery profits are larger under output controls and the variance in profits is considerably smaller. Since there is less variance in stock relative to CPUE it is easier to control stocks by targeting catch, maintaining stock size, lowering its variance relative to the use of effort controls and thus decreasing the overall costs of fishing. The difference in profits between output and input controls is of course smaller under discounting since future gains from stock recovery and control in the future are worth less today. The issue of the appropriate rate of discount (if any) in a Commonwealth fishery is a subject for future research.

The comparison between MSY and MEY is most important. Using a constant number of boats at 120, shows that the ratio of stock at MEY to stock at MSY is roughly 1.42, indicating substantial overfishing in the NPF. Optimal effort levels in table 2 are roughly 60 per cent of current levels.

Finally, it should be noted that all results are obtained under the assumption of no effort creep or losses from high-grading. In cases where effort controls are used, effort creep can be considerable resulting in falls in fishery profits and the need to periodically measure ‘true’ fishing effort and thus adjust optimal target effort levels. This also involves costs in fleet restructuring and administrative and negotiating cost with new management arrangements. In cases where output controls are used, high-grading is often a concern since it may pay to discard low value catch, although in the NPF, with relatively homogenous stocks and catch this may be less of a problem.

A stochastic production frontier analysis for the northern banana prawn fishery

This second case study summarises the results of a stochastic production frontier analysis for the NPF, specifically for the banana prawn fishery, by estimating equations comparable to 4.3 and 4.4. This approach allows for vessel level measures of economic efficiency and gives an assessment of the efficiency implications of the use of input controls for the fishery as a whole. A more elaborate description of the model and results is contained in Kompas et al. (2004).

Data sources and variables

The unbalanced panel data used to estimate the stochastic frontiers for the NPF comes from two different data sets. Data on a larger set of variables is available for an unbalanced panel of 853 observations for 138 vessels over the period 1990 to 2000. The vessels included in the data harvested almost 40 per cent of the total catch of banana prawns each year and are drawn from surveys and statistics for the NPF fleet carried out and compiled by ABARE and the CSIRO. The data includes measures of output by species (banana, brown and grooved prawn), crew size, revenue, boat variable costs (not available by species), capital costs, nominal fishing days for banana prawns and vessel characteristics (hull units, engine power, A-units, gear length, boat size). The vessel characteristics, landings of banana prawns and nominal fishing days for banana prawns are provided from the CSIRO surveys for the fishery.

Generalised likelihood ratio tests are used to help confirm the functional form and specification of the estimated models. The correct critical values for the test statistic come from a mixed chi-squared distribution (at the 5 per cent level of significance). A translog specification was initially estimated, but a pre-test with the null hypothesis of the Cobb–Douglas as the correct functional form could not be rejected (Kompas et al. 2004).
Maximum likelihood estimates of the model were obtained, following a three-step procedure. OLS estimates are first obtained, followed by a grid search that evaluates a likelihood function for values of gamma between zero and one, with adjustments to OLS estimates. All other values of beta are restricted to be zero in this step. Finally, the best likelihood values selected in the grid search are used as starting values in a quasi-Newton iterative procedure to form maximum likelihood estimates at a point where the likelihood function obtains its global maximum.

Estimated results and efficiency analysis

Results for the model are reported in table 4, and a description of inputs in table 3. All input variables in the stochastic frontier production function are significant, except crew number, as are time trend and year-dummy variables. Estimates also show that inputs for banana prawn output in order of importance are fishing effort (boat days), fuel (as a proxy for engine size and power), headrope gear length and crew number. All input share coefficients sum to 0.75. Results of OLS estimates are also reported and as expected vary from frontier estimates for all input variables.

Results for the technical inefficiency model indicate that A-units and gear length are both significant. A-units have a significant negative effect on technical inefficiency (hence a positive effect on technical efficiency) and gear length has a positive effect on inefficiency. The hire of a skipper estimates as non-significant but has the expected sign. Incentive effects for owner-operated boats should likely result in an increase in technical efficiency relative to a hired skipper. The estimated results of the average technical efficiency are reported, showing the decreasing trend during the study period (see figure j).

Although banana prawn catch is highly dependent on seasonal weather effects, the relationship between catch and future stock abundance, as mentioned earlier, is not clear. In fact, it is argued that future stock size seems to be largely independent of the amount of fishing effort on adult stock, with the escape of spawners highly resilient to recruitment overfishing (Staples and Maliel 1994). Nevertheless, catches below expectations have generated concern that stock size may be falling.

Table 4 shows two important results of input controls in the northern prawn fishery. First, controls on A-units (hull and engine size) by the regulator has had the net effect of reducing technical efficiency (or raising technical inefficiency) because the estimated coefficient in the technical inefficiency model is statistically significant and negative. In other words, for the average vessel an increase in A-units lowers technical inefficiency (raises technical efficiency). Second, Kompas et al. (2004) found that because of controls on A-units in the 1980s fishers have tended to substitute to increased headrope length so as to increase their fishing power. Unfortunately, the technical inefficiency model indicates that such input substitution has raised technical inefficiency (lowered technical efficiency) because its estimated coefficient is positive and statistically significant. Table 4 shows this effect in terms of the average measure of technical efficiency over time in the NPF. It falls considerably. The reason is clear. On average the ratio of A-units to gear in the fishery falls over time. Given the estimates this must imply efficiency falls as well.

The efficiency analysis indicates that input controls on hull size and engine power and the substitution to unregulated inputs, such as headrope length, have reduced technical efficiency in the NPF. Such an outcome runs counter to the stated objective of the fishery regulator to both maximise economic efficiency and ensure the sustainability of the resource.

3	Description of inputs and vessel specific variables in the northern prawn fishery (138 vessels for the period 1994-2000)

variables

description

Output

Output of banana prawns (kg)

Fishing effort

Nominal fishing days for banana prawns (days)

Hull units

Under deck volume (m³)

Engine power

Registered engine power (kW)

Vessel A-unit

The sum of one A-unit for every cubic meter of hull
volume and one A-unit for each kilowatt of engine power

Gear length

Headrope length of gear (meters)

Boat size

Vessel length (meters)

Sources: Statistics from CSIRO (2003).

4	Estimated results of the NPF frontier analysis (19902000)

coefficient

asymptotic T-ratio

Stochastic production frontier

Constant

4.65

(0.43)

10.79

Effort

0.72

(0.03)

23.74

Engine power

0.38

(0.09)

4.01

Head rope length of gear

0.44

(0.14)

3.19

Year 1994

–0.62

(0.04)

17.89

Year 2000

–0.38

(0.04)

9.7

Technical inefficiency model

Constant

(2.81)

3.91

Head rope length of gear

2.2

(1.16)

1.88

Hull units

–1.59

(0.61)

2.52

Fishing effort

–1.27

(0.37)

3.42

Engine power

–1.15

(0.49)

2.37

Sigma-squared

0.68

(0.21)

3.27

Gamma

0.91

(0.03)

33.63

Ln (likelihood)

348.54

Mean Technical Efficiency (19902000)

74.40%

a Denotes statistical significance at the 0.05 level.
b Denotes statistical significance at the 0.01 level.
Note: Numbers in parentheses are asymptotic standard errors.