Professional Documents
Culture Documents
x
ij
+
i
w
i
J
j=1
e
x
ij
+
i
w
i
(2)
Where x
ij
are those individual-specic attributes (characteristics
that depend on the individual) and w
ij
are those choice-
specic attributes (characteristics that depend on the choice). See
McFadden, 1973for amoredetaileddescriptionof themethodology.
5. Variables
The deterministic part of the utility function is explained by the
variables obtained from the qualitative analysis and some other
relevant variables presented in Table 2.
The dataset contains trip level information for the period
19962002. Data have been obtained fromlogbooks and sale forms
3
See for example Kobayashi (1993) for an explanation of this distribution.
of trawl vessels based on ports in the region of the Basque country
(Spain).
The description of the variables is provided as follows. The
variables are classied in ve different groups: vessel, trip, man-
agement, economic and behaviour characteristics.
5.1. Characteristics of the vessel
- Fishing length: The average vessel length is 31m, a maximum of
39m and a minimum of 23m.
- Vessel base harbour: Similar patterns from vessels of the same
harbour are checked (base, departure or landing harbour).
- Fishing gear: If the gear chosen has any signicant effect on the
selected area.
5.2. Characteristics of the trip
- Distance to the shing area: The distance in miles fromthe depar-
ture harbour to the shing ground (distance to the shing area)
and the distance from this last shing ground to the landing har-
bour (distance to the landing harbour).
- Duration of the shing trips: The number of days each trip lasts.
- Season: Season of the year in which the trip was made
5.3. Management characteristics
- Fishing rights: Fishing rights, in terms of allowable shing days in
each shing area, at the beginning of each year.
28 R. Prellezo et al. / Fisheries Research 97 (2009) 2431
Table 3
Parameter estimates from the multinomial logit (unordered) model on shing area choices.
Variables log(PVII /P
VIII a, b, d
) S.E.VII ZVII Log(PVI /P
VII I, ab, d
) S.E.VI ZVI
Intercept 4.70
***
0.88 5.30 21.08
***
1.27 16.52
Vessel characteristics
Vessel length 0.31
***
0.022 13.97 0.48
***
0.03 15.43
Trip characteristics
Duration 0.15
***
0.069 2.19 3.21
***
0.25 12.48
Managerial characteristics
Fishing rights VI 0.0031 0.03 0.87 0.09
***
0.005 17.68
Fishing rights VII 0.0017 0.0012 1.36 0.013
***
0.001 7.86
Fishing rights VIII a, b, d 0.005
***
0.0008 6.18 0.0009 0.001 0.845
TAC Hake 0.0003
***
0.00001 23.87 0.0002
***
0.00001 15.55
Economical characteristics
Fuel cost 0.0035
***
0.25 6.063 0.03
***
0.094 14.81
Production function 0.00044 0.00002 17.746 0.00000 0.000038 0.178
Behavioural characteristics
Risk 10.39
***
0.37 27.7 4.293
***
0.29 14.55
Inertia 1.25
**
0.18 6.91 1.64
***
0.47 3.45
N=10502 Pseudo-
2
=0.679 Prob[ChiSqd>value] =.0000000 Log-likelihood=2176.
Statistical signicance at 10% level.
**
Statistical signicance at 5% level.
***
Statistical signicance at 1% level.
- Fishing rights utilisation: Remaining shing rights in terms of
shing days by area and by vessel.
- TAC: The TAC established for the main three species captured
by these eets (Hake, Anglersh and Megrim) by area
4
(ACFM,
2003).
- TAC utilisation: Remaining quota for the overall eet by trip.
5.4. Economic characteristics
- Fuel cost: Fuel expenditure per trip in Euros. This is calculated
from the ratio between the total expenditure in fuel each year
and the total shing days, and then multiplied by the average trip
duration of each eet.
- Turnover: Landings value (in Euros) obtained in each trip
5
.
- Turnover by effort: Landings value (in Euros) divided by shing
days.
- Production function: Used as a proxy of the expected turnover,
by vessel and by area. In order to consider different expectations,
two different alternatives have been tested: Lagging the turnover
to the previous trip or to the previous year.
5.5. Behaviour characteristics
- Inertia: A dummy variable describing whether the current trip is
inthe same area as the previous one. Whenthat is the case, inertia
is equal to 1 and 0 otherwise. This variable captures the tendency
of vessels to returnto areas where experience gives a comparative
advantage.
- Risk: It stands for the variability of the turnover of each choice,
measured in terms of standard deviation relative to the average
turnover. Several alternatives have beentested(by year, by season
and by season weighted by the distance).
4
For the case of the TAC of Hake, Sub-areas VI and VII have a common TAC (which
alsoincludes DivisionVb, andSub-areas XII andXIV) or the case of the TACof Megrim
the Sub-area VI also includes Division Vb and Sub-areas XII and XIV.
5
It is important toremarkthat the price of the same species varies withthe shing
area where they have been caught. When possible (for some years this data cannot
be obtained) this is the approach taken into account.
6. Empirical procedure
The variables described in the previous section are by con-
structionindividual specic. All tripandbehavioural characteristics
of the eet are individual (trip) specic and hence, the model
to use is a multinomial (unordered) logit
6
. Eq. (2) can then be
rewritten as:
Prob(Y
i
= j) =
e
x
ij
1 +
J1
j=1
e
x
ij
(3)
The deterministic component of the indirect utility function (3)
in the multinomial (unordered) logit model has empirically been
specied as:
V
j
= + Vessel characteristics + Trip characteristics
+ Managerial characteristics
+ Economical characteristics
+ Behavioural characteristics (4)
Where , , , , are vectors of the coefcients of each category.
The utility function (4) is estimated iteratively using a multinomial
logit model and the results converge to the maximum likelihood.
Independence of irrelevant alternatives (IIA)
7
is assumed. In order
to avoid endogeneity issues in the estimation, highly correlated
variables have been dropped out.
The individual statistical signicance of each variable has been
tested. The results are presented in Table 3:
V
j
= + Vessel length + Duration +
1
F Rights VI
+
2
F Rights VII +
3
F Rights VIII +
4
TAC Hake
+
1
Fuel Cost +
2
Production function
+
1
Risk +
2
Inertia (5)
6
Furthermore for J alternatives in a multinomial logit model, only J-1 distinct
parameter vectors may be identied and the probability of the reference category
has to be computed taking into account that the summation of the probabilities
should be equal to one.
7
The statistical package used has been Limdep.
R. Prellezo et al. / Fisheries Research 97 (2009) 2431 29
The estimated model is presented in Table 3. The pseudo-
2
indicates that 68% of the variation of the area selection behaviour
has been explained. The log-likelihood ratio and chi-square value
are also signicant. Finally, residuals are tested to conrm that a
Weibull distribution is followed.
One of the categorical variables, Division VIII a, b, d, is dropped
in order to obtain full rank. All the results are presented relative to
this alternative.
In Table 3 individual signicances are presented. Most of the
variables are statistically signicant. Looking at the vessel charac-
teristics, we can observe that the odds of choosing Sub-areas VI
and VII relative to choosing Division VIII a, b, d increase with vessel
length. This implies that larger vessels have higher probability of
harvesting in Sub-areas VI and VII.
Trip characteristics are important to determine area choice. As
the expected duration of the trip increases, the likelihood of choos-
ingSub-area VII relative tochoosingthe Bayof Biscaydecreases. The
probability of choosing Sub-area VI however decreases. The reason
is that most of the vessels shing inSub-area VI landinshing ports
close to the shing area, in Irish and British ports.
The variables that describe management are also statistically
signicant. This is expected as regulatory measures vary (in quan-
tity not in quality) with the shing area. The effect of regulation in
each of the shing areas varies considerably. The number of days
allowed to sh in the Bay of Biscay affects negatively the likeli-
hoodof choosing Sub-area VII relative to the Bay of Biscay. Likewise,
the higher number of shing rights in Sub-area VI, the probabil-
ity of selecting this area relative to shing in the Bay of Biscay is
higher.
Output regulation, measured by the mean of TACs by area for
the three main species (Hake, Anglersh and Megrim), is statisti-
cally signicant. An increase in Hake quota will also increase the
probability of shing in Divisions VIII a, b, d. An increase in Hake
TAC is understood by shermen as a lower constraint as well as
increase (or non decrease) in the relative abundance of Hake, mak-
ing vessels go to closer shing areas. It is important to remark that
theother twoTACs (AnglershandMegrim) donot affect areaselec-
tion. There are two main reasons for this result. First, the other two
TACs have not been as restrictive as the one applied to Hake (the
evolution of these two TACs have been stable), and, second, hake is
the target species of 90% of the trips for pair trawlers and for 50%
of the total trips (including Baka trawlers).
Economic characteristics affect shing area choice. Fuel cost
affects the probability of choosing Sub-areas VI and VII relative
to Division VIII a, b, d. Furthermore, expected turnover affects the
probability of choosing Sub-area VII relative to the Bay of Biscay.
Behavioural characteristics are also signicant. In Table 2 we
observe that the lowest risk (or uncertainty) is foundinSub-area VII
and the highest risk depends on the season. Risk reduces the proba-
bility of choosing Sub-areas VI and VII. Finally, inertia increases the
likelihood of shing in Sub-areas VII or VI when the previous trip
Table 4
Area choices explained using model estimates.
Predicted
VI VII VIII a, b, d
Actual VI 537 (70%) 153 81
VII 89 1088 (77%) 235
VIII a, b, d 44 134 8141 (98%)
Table 5
Marginal average effects.
Variables dPVI /dxK dPVII /dxK dP
VIII a, b, d
/dxK
Vessel length 0.0085 0.0066 0.015
Duration 0.0918 0.0575 0.0343
Fishing rights VI 0.0027 0.0014 0.0013
Fishing rights VII 0.0004 0.0003 0.0001
Fishing rights VIII a, b, d 0.0001 0.0002 0.0001
Fuel cost 0.0008 0.0003 0.0005
TAC Hake 0.0000 0.0000 0.0000
Production 0.0000 0.0000 0.0000
Risk 0.0445 0.3959 0.3519
Inertia 0.0656 0.0821 0.0165
was located in those areas.
The predictive power of the model is presented in Table 4. Real
observations are compared to model predictions. The diagonal of
Table 4 indicates that 70% of the trips made in Sub-area VI, 77% of
the trips in Sub-area VII, and 98% of the trips in the Bay of Biscay
are well predicted.
7. Comparative static
The estimated parameters presented in Table 3 are interpreted
once the marginal effects (dP
j
/dx
K
) are computed (Powers and Xie,
2000).
The marginal average effects presented inTable 5 (area averaged
over individuals) give an interesting point for the decision making
process. Any eet modernization policy that increases the average
length of this eet will increase the probability of shing in Sub-
areas VI and VII. If the expected duration of the trip increases, the
probability of shing in Sub-area VI reduces.
All these probabilities are constrained by the current manage-
ment scenario. Anincrease inthe shingrights of one area increases
the probability of shing in that particular location. There are two
main implications of this result. First, if any shing right is bought
from any area, the actual shing area distribution will be moved
towards this area. Second a raise in fuel costs increases the number
of trips in Sub-area VI.
Finally, behavioural aspects are also important. Higher risk (in
terms of variability of the shing trips turnover), increases the
probability of shing in relatively safe trips increasing so the prob-
ability of shing in the Bay of Biscay.
Fig. 2. Probability of choosing Sub-area VI, VII or Divisions VIII a, b, d with vessel overall length.
30 R. Prellezo et al. / Fisheries Research 97 (2009) 2431
Fig. 3. Probability of choosing Sub-area VI, VII or Divisions VIII a, b, d with fuel cost.
Fig. 4. Probability of choosing Sub-area VI, VII or Divisions VIII a, b, d with Production function.
Fig. 5. Probability of choosing Sub-area VI, VII or Divisions VIII a, b, d with Risk Perception.
The probability of shing area selection has been simulated
under different scenarios of changes in the overall vessels lengths,
fuel cost, expected turnover and perception of risk. Results are pre-
sented in Figs. 25.
The current average vessel size is 31m. If a modernization pol-
icy is established, increasing the vessel size 510%, by the sheries
manager or vessels owners (shing rms), changes in the probabil-
ity of choosing the shing area will be small (Fig. 2). A policy that
increases the length (beyond 3m in average terms) will however
signicantly change the shing area selection, increasing the prob-
ability of choosing Sub-areas VI and VII, and decreasing the number
of trips made in the Bay of Biscay. The same argument applies to
a decommission scheme that drops the smallest vessels out of the
shery.
The short terminelasticity of fuel price (Lazkano, 2008), implies
that increment of it, does not affect, in the short term, the quantity
demanded
8
. This result can be seen in Fig. 3. An increase of 25%
in the price of fuel does not change the probability. Higher than
25% increase changes some of the trips from the Bay of Biscay to
Sub-area VII, looking for a greater RPUE
9
.
8
Note that this result can be biased due to the price receives a subsidy.
9
In the short term the stock size is invariant, while in the long run it can clearly
inuence the results obtained.
When the expected turnover increases, both sides of the Y axis
of Fig. 4 are especially relevant. Turnover could increase or decrease
due to technological, natural or shing strategic changes
10
. In any
case, the probability of selecting the Bay of Biscay as the shing
ground will decrease, increasing the probability of selecting Sub-
area VII, while the probability of selecting Sub-area VI does not
change.
Finally, in Fig. 5 changes in risk perception have been simulated.
As the risk increases the probability of selecting the relative safety
of the Bay of Biscay rises and, therefore, the probability of choosing
Sub-area VI goes down.
8. Discussion and concluding remarks
In this paper shermen behaviour in relation to shing area
selection in the north-east Atlantic trawl shery during 199602
has been analysed. The Basque trawlers eet has been used as a
case study.
10
The reader could think that the expected turnover can change due to many
other reasons, but in this case, we only include those that have not been taken into
consideration when dening the utility function of each shing area alternative.
R. Prellezo et al. / Fisheries Research 97 (2009) 2431 31
First, a qualitative approach based on on-boards interviews
made to skippers is presented to obtain a general understanding of
the main variables that determine area selection. The results show
that shing location is mainly inuenced by experience, regula-
tions, expected harvest, external communications and fuel costs in
addition to a set of undetermined reasons.
The results of the qualitative analysis are usedas input ina quan-
titative analysis. A quantitative analysis can help determine the
exact effect of different variables in the selection of shing area.
A random utility model, a multinomial unordered logit model is
employed.
The variables have beendividedindifferent categories. The most
important result is that all categories are statistically signicant. It
gives a rst scope of the complexity involved in the shing area
choice. The most important variables are vessels characteristics
(vessel length), managerial constraints (shing rights), behaviour
of shermen (i.e., inertia), variable costs (fuel cost) and the risk in
each area (in terms of standard deviation of the average turnover
of each area).
The quantitative analysis provides the exact effect of the vari-
ables on the probability of selecting different shing areas as well
as the effect of simulations. In that sense, vessel length changes,
fuel cost, expected turnover and different uncertainty (risk) sce-
narios have been implemented. These scenarios are important to
determine the effect of changes in the economic conditions on the
distribution of shing area. The most important economic condi-
tions of the Basque trawl eet in the last decade are the oil crisis of
2000 and 2004, as well as the one in the rst part of 2008, a change
in the current limits of the shing rights transferability, and eet
renovation policies. The effect of those changes in the selection of
shing area has been studied through simulations in Section 7.
EU sheries management is mowing towards a shery-based
concept rather than a single stock basis. Fisheries are however, as
shown in this paper, dynamic in the sense that they are affected
by eet behaviour and, in particular, by the shing area choice.
Future management plans applied to a shery may change the sh-
ery itself and the restrictions in one area may affect other areas
through the re-allocation of shing effort. Furthermore, changing
the shing rights transferability conditions, for example by creat-
ing an individual transferable quota system can cause unexpected
shing effort re-allocations. Some areas can also be affected by
the economic framework changing so shing area locations when
fuel costs change. The analysis of shing location has proven to be
important both for shermen and for sheries managers. The prof-
its of shermen are determined by shing location and the success
of management policies depend on the effect of those policies on
shing area selection.
Acknowledgements
This work has been funded through the TECTAC project by the
European Union (DG Fisheries, study no. QLRT-2001-01291), the
CAFE project of the European Union (DG-Fish, contract no. 022644)
and by the Basque Country Government (Agriculture and Fisheries
Department). We would also like to acknowledge the work done by
Estanis Mugerza, Luis Arregi andJonRuiz inthe qualitative analysis.
All errors are our own responsibility.
References
Gordon, H.S., 1954. The economic theory of a common property resource the shery.
Journal of Political Economy 62, 124142.
Greene, W., 2000. Econometric Analysis, 4th ed. Prentice-Hall, New Jersey.
Holland, D.S., Sutinen, J.G., 1999. An empirical model of eet dynamics in New Eng-
land trawl sheries. Canadian Journal of Fisheries and Aquatic Sciences 56 (2),
253264.
Holland, D.S., Sutinen, J.G., 2000. Location choice in NewEngland trawl sheries: old
habits die hard. Land Economics. 76 (1), 133149.
ICES, 2003. Report of the study group on the development of shery-based forecasts.
Boulogne, France, ICES: 37.
Kobayashi, A. (Ed.), 1993. Handbook on Experimental Mechanics. VCH/SEM, New
York.
Laxe, G.F., 2006. Transferability of shing rights: the spanish case. Marine Policy 30,
379388.
Lazkano, I., 2008. Cost structure and capacity utilisation in multi-product indus-
tries: an application to the Basque trawl industry. Environmental and Resource
Economics 2 (2), 189207.
McFadden, D., 1973. Conditional logit analysis of qualitative choice behaviour.
In: Zarembka, P. (Ed.), Frontiers in Econometrics. Academic Press, New York,
p. 1974.
Mistianen, J., Strand, I.E., 2000. Location choice of commercial shermen with het-
erogeneous risk preferences. American Journal of Agricultural Economics 82 (5),
11841190.
Powers, D.A., Xie, Y., 2000. Statistical Methods for Categorical Data Analysis. Aca-
demic Press, London.
Pradham, N., Leung, P., 2004. Modelling trip choice behaviour of the longline shers
in Hawaii. Fisheries Research 68, 209224.
Sanchirico, J.N., Wilen, James E., 1999. Bioeconomics of spatial exploitation in a
patchy environment. Journal of Environmental Economics and Management,
Elsevier 37 (2), 129150.
Wilen, J., 2000. Incorporating space into sheries models: comment. American Jour-
nal of Agricultural Economics, American Agricultural Economics Association 82
(5), 12101212.