You are on page 1of 16

Journal of Productivity Analysis, 9, 145–160 (1998)

°
c 1998 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.

Characterizing Managerial Skill and Technical


Efficiency in a Fishery
JAMES KIRKLEY jkirkley@vims.edu
College of William and Mary, School of Marine Science, Virginia Institute of Marine Science,
Gloucester Point, Virginia 23062

DALE SQUIRES dsquires@its.ucsd.edu


National Marine Fisheries Service, Southwest Fisheries Science Center, P.O. Box 271, La Jolla, California 92038

IVAR E. STRAND
Department of Agricultural and Resource Economics, University of Maryland, College Park, Maryland 20742

Abstract

Researchers have long recognized that entrepreneurial or managerial skill is a major de-
terminant of productivity or reason why production among firms varies. Yet, except for
a few studies, differences in productivity and output levels are usually attributed to plant
configuration or scale. More important, there appears to have been few attempts to re-
late technical efficiency to managerial skill. Utilizing a stochastic production frontier, we
examine the relationship between technical efficiency and characteristics of skill such as
experience and education in a fishery. Although we can not determine threshold or essen-
tial levels of experience and education, substitution possibilities are found to exist between
years of experience and education levels. Additional analysis of efficiency for two captains
of the same background and experience reveals that additional characteristics need to be
considered in the examination of skipper skill or the “good-captain” hypothesis.
Keywords: Technical efficiency, managerial skill, and fisheries

1. Introduction

The notion that some fishing captains are better managers than other captains, and in
turn, consistently have higher production and earnings has long been recognized by fish-
ery researchers (Acheson, 1975; Bjorndal, 1987, 1989; Carlson, 1975; Rothschild, 1972;
Thorlindsson, 1987). Noetzel and Norton (1969), for example, attributed superior returns
of “highliners” in the Boston large trawl fleet in part to skipper skill. Hilborn and Ledbetter
(1985), using an analysis of variance model to explain catching power of purse seiners in the
British Columbia salmon fishery, concluded that skipper skill and motivation were major
determinants of catching power. Houghton (1977) also concluded that skipper ability was
an important factor in explaining catch rates or productivity.
There appears, however, to have been little research devoted to examining quantitatively
the characteristics which might define skipper skill. Two possible reasons for the lack of
146 KIRKLEY, SQUIRES AND STRAND

research on the “good captain” hypothesis are the difficulty of directly measuring skipper
skill and inadequacy of data necessary for evaluating characteristics of skill (Mundlak 1961;
Palsson and Durrenberger, 1982).
Skipper skill has been determined to be an important determinant of vessel productivity
in the limited research which has examined the relationship between harvest levels, vessel
productivity, and subjective or qualitative measurements of skipper skill. For example,
Comitini and Huang (1967) used a subjective assessment of skipper skills as average, good,
or excellent and concluded that good captains made differences in vessels’ productivity and
that the seeming presence of technical progress was largely accountable by the managerial
skills of the captain. Campbell (1991) similarly used subjective rankings of skill to assess
whether or not skill might be an important determinant of production in the Tasmanian rock
lobster fishery. Campbell concluded that skill and technology were not important in the
inshore fishery but were very important for explaining differences in the offshore fishery.
In this paper, we equate technical efficiency (TE) to skipper skill and then examine the
relationship between TE and skill characteristics of captains. TE is a necessary condition
for economic efficiency, and the degree to which it correlates with policy instruments such
as vocational training can be important to social efficiency. Previous empirical research has
neglected the obvious relationship between skipper skill and TE.
Using a data set for 10 Mid-Atlantic sea scallop, Placopecten magellanicus, dredge ves-
sels operating between 1987 and 1990, a stochastic translog production frontier is specified
and estimated at the trip level. TE per trip is subsequently estimated by the procedures of
Jondrow et al. (1982) and Battese and Coelli (1988). Using data on skipper characteristics
obtained from a 1990 survey of primary boat captains, estimates of average TE for each ves-
sel and primary captain during 1990 at the trip level are examined relative to characteristics
believed indicative of skipper skill. The analysis suggests that experience and education
are important components of skipper skill and may be substitutable within limits.
An additional analysis of TE for two other vessels is also conducted. The two vessels
are identical in configuration, size, and operating characteristics but considerably different
from the other 10 Mid-Atlantic dredge vessels. The vessels are captained by Mexican-
Americans of the same age, who grew up in the same home town of Texas, and were from
the same high school graduating class. Each captain was trained by the same individual
and had the same number of years of experience as fishermen and scallop captain. The
analysis reveals that one captain consistently is more technically efficient and has higher
earnings. The result leads to the conclusion that skipper skill can only be partly measured
by basic characteristics such as age, education level, and years of experience; additional
factors, such as motivation, must be defined and considered in the analysis of skipper skill.

2. Technical Efficiency

Technical efficiency (TE) for a fishery is a measure of the ability to produce the maximum
output possible from a given set of inputs subject to the production technology, resource
levels, weather conditions, and other technological constraints. TE is a relative concept
since each vessel’s production performance is compared to a best-practice input-output
relationship and measured as the deviation of individual vessels from this best-practice
CHARACTERIZING MANAGERIAL SKILL AND TECHNICAL EFFICIENCY IN A FISHERY 147

frontier.
We base our assessment of TE on results obtained from estimating a stochastic production
frontier. Exploitation of renewable resources appears to be inherently stochastic. Weather
conditions and biological abundance and availability appear to be extremely variable, and
captains must often make quick changes in response to changing resource and weather
conditions. Alternatively, production or output levels are likely to be significantly affected
by unanticipated weather and resource conditions (e.g., a sudden squall may limit the ability
of the vessel and crewto fish or reduce resource availability and density).
The stochastic frontier is specified with two error terms (Aigner, Lovell, and Schmidt,
1977; Meeusen and van den Broeck, 1977). One term, vi , is assumed to be independently
and identically distributed (iid) as N (0, σv2 ) and captures exogenous shocks beyond the con-
trol of firms. Given the inherent stochasticity of harvesting fish, capturing these stochastic,
exogenous shocks seems particularly important. A one-sided, non-negative, error term,
u i , is introduced to represent technical inefficiency (TI); we consider the truncated normal
distribution u ∼ N (µ, σu2 ).1 If u i = 0, production lies on the stochastic frontier and produc-
tion is technically efficient; if u i > 0, production lies below the frontier and is technically
inefficient.
The stochastic production frontier permits output (Y ) to be specified as a function of
inputs (X ) and a disturbance term:

Yit = h(X 1it , X 2it , . . . , X N it ; A)e²it (1)

where Yi is output of the i th producing unit, t is time, X ji is the j th of N inputs, A represents


a vector of parameters, e is the exponential operator, and ²it = vit − u it is the disturbance
term.
TE for the individual firm, TEit = e−ui , or as shown by Coelli (1994), equals
E(Yi∗ | Ui , Xi )
TEi = (2)
E(Yi∗ | Ui = 0, Xi )

where E is the expectations operator, Yi∗ is the production of the i th firm, and equals Yi
when the dependent variable is in original units and exp(Yi ) when production is in logs. The
actual calculation of TE requires deriving the conditional expectation of u it conditional on
²it or vit − u it . Battese and Coelli (1992, 1993) provide the relevant expressions, which are
generalizations of the results of Jondrow et al. (1982) and Kalirajan and Flinn (1983), for
obtaining the conditional expectation of u it given ²it .
In general form, Jondrow et al. (1982) show that the conditional expected value of u i
equals
· ¸
σλ φ(µ∗ ) ∗
E[u i | ²i ] = −µ (3)
(1 + λ2 ) {1 − 8(µ∗ )}

where µ∗ = ²λ σ
+ σµλ , λ = σσuv , and σ = (σv2 + σu2 )1/2 . The value and statistical significance
has been shown by Schmidt and Lin (1984) to be quite important in determining the existence
of a stochastic frontier; rejection of the null hypothesis, H0 : λ = 0, implies the existence
of a stochastic frontier. Also, as the value of λ > 1, production may be said to be dominated
148 KIRKLEY, SQUIRES AND STRAND

by technical inefficiency. Coelli (1995), however, suggests that the Wald t-test of λ = 0
vs. λ > 0 may be an inappropriate test for the existence of a frontier. Coelli recommends
that the preferred test should be a one-sided likelihood ratio test of γ = 0 vs. γ > 0 where
γ = σu2 /(σv2 + σu2 ); we thus consider Coelli’s test for the existence of a stochastic frontier.
We believe that the TE estimate is a suitable indicator of skipper skill. It measures how
closely a particular captain on a particular trip came to achieving the production frontier.
Although technical efficiency is not sufficient for profit maximization, it is necessary and
certainly superior to subjective measures used in other studies to examine productivity and
skipper skill. Subjective measures may have the captains’ personalities and other traits
embedded in them.

3. The Stochastic Production Frontier and Data

3.1. A Translog Specification of the Production Frontier

The translog flexible functional form is chosen for the stochastic frontier. It is relatively
easy to estimate, permits a limited determination of the underlying technology, and easily
accommodates the inclusion of a one-sided error term to allow estimation of TE for each
observation. The stochastic frontier is specified at the trip-level for 10 Mid-Atlantic sea
scallop dredge vessels operating between 1987 and 1990:

ln Yit = β0 + β1 ln D Ait + β2 ln L it + β3 ln Sit


+β11 (ln D Ait )2 + β22 (ln L it )2 + β33 (ln Sit )2 (4)
+β12 ln D Ait ln L it + β13 ln D Ait ln Sit + β14 ln L it ln Sit + ²it

where i indexes individual vessels and t indexes fishing trips. The variables are scallop
meat weight landed (Yit ) by the i th vessel on the t th fishing trip, days at sea per trip (D Ait ),
crew size (L it ), and resource stock size (Sit ).
Battese and Coelli (1992) have a made a strong argument that when using panel data
there is a likelihood that firm level effects may vary systematically. Alternatively, there
is a need to consider time-varying efficiencies. It is unlikely, however, that technical
efficiency in a fishery systematically varies over years. TE may exhibit, however, some
systematic variation relative to months or seasons of a year. Unfortunately, we are not able
to readily adopt Battese and Coelli’s time-varying approach to estimate TE in the presence
of systematic temporal variation.2
Instead, we consider Battese and Coelli’s (1993, 1995) one-stage specification which
allows TE to vary systematically with months of year and inconsistently over years and
relative to each of the ten vessels examined in this study. We are thus able to examine a
considerably more flexible specification for examining the possible temporal patterns of TE.
Following Battese and Coelli’s (1993, 1995) specification, we specify inefficiency, u i , to be
a function of a series of dummy variables representing the four years, twelve months, dredge
size, and each of the ten vessels. More formally, u i are non-negative random variables and
are assumed to account for technical inefficiency in production and are further assumed
to be independently distributed as truncations at zero of the N (µit , σu2 ) distribution and
CHARACTERIZING MANAGERIAL SKILL AND TECHNICAL EFFICIENCY IN A FISHERY 149

µit = Z it δ (Coelli, 1994). Z it is a 1 × p vector of dummy variables which are thought to


possibly influence the efficiency of the fishing vessels and δ is a p × 1 vector of parameters
to be estimated using the one-stage routine of Battese and Coelli (1993, 1995).
Omitting the reference dummy variables and including an intercept or constant term for
the one-stage estimation, we have 9 dummy variables for vessels, 11 dummy variables for
months, 3 dummy variables for years, and one dummy variable for dredge size (dredge size
equals 1 for a 13 foot dredge and 0 for a 15 foot dredge which are the only two sizes used).
We thus have 25 parameters for our one-stage specification. To test for the existence of a
stochastic frontier, we have to test 26 restrictions—all 25 δ parameters and γ equal 0.0.
Direct estimation of Eq. (4) raises the issue of endogenous explanatory variables. That
is, if producers maximize profits or minimize costs, direct estimation of the production
function leads to simultaneous equation bias. Zellner, Kmenta, and Dreze (1966), though,
argued that where output is stochastic due to “acts of nature,” such as weather, firms can
be assumed to choose inputs to maximize expected profit. Choice of inputs to maximize
expected profit can be assumed to be subject to “human errors,” which are uncorrelated with
the stochastic error term in the production function. Also as noted by Campbell (1991) and
Bjorndal (1989), under this error specification, which would seem to fit the case of a fishery,
direct estimation of the production function rather than as a part of a system of equations
(including share equations) yields consistent parameter estimates.
The input variables are not conventional economic inputs. Days at sea embodies capital,
energy, materials, and some labor. It is not known whether the choice of days at sea is
the first stage of a non-separable two-stage technology as in Pollak and Wales (1987) or
a composite input formed by a separable technology as in Squires (1987) and Hannesson
(1983). Crew size is the stock and not the flow of labor services; labor services, however,
are proportional to total crew size and days at sea since crew members usually work 12
hours per day.
Despite the limitations caused by the specification of inputs, Eq. (4) offers a convenient
framework for examining TE in a commercial fishery. In addition, the framework is con-
sistent with bio-economic specifications typically used to examine production in a fishery.
Bjorndal (1987, 1989), Campbell (1991), and Strand, Kirkley, and McConnell (1981), also
faced with data limitations, adopted similar specifications for fishery production functions.

3.2. Data

Data on output and input levels were obtained from trip settlement summaries or trip-level
costs-and-earnings statements for 10 Mid-Atlantic scallop dredge vessels operating between
1987 and 1990. One of the vessels exited the fishery in 1988 so that data for this vessel
were included only for 1987. The ten vessels were relatively homogeneous. All were
between 79 and 82 feet, steel-hulled, and constructed between 1981 and 1984. Three of
the ten vessels had engines with smaller horsepower than the engines of the other vessels,
and hence, dragged smaller dredges (13 foot instead of 15 foot dredges). All of the vessels
had two radars, Loran C, and plotters. It was not known if there were differences in other
electronics and other vessel equipment (e.g. gear-matics mounted on bridge or whether
vessels had remote rudder control). The total number of observations was 584.
150 KIRKLEY, SQUIRES AND STRAND

A problem posed by the data and the reason for using days at sea rather than other inputs
(e.g., fuel) was “split trips”. Vessels frequently made two to three trips but did not settle
until completion of the last trip. The settlement sheets indicated days at sea and landings
for each trip or part of trip, but not the usage of inputs such as fuel for each leg of a split
trip. Split trips usually occur when vessels migrate north or south and do not land at their
home port. They also occur because of storms and emergencies.
The 10 vessels selected for study were not representative of the fleet. Vessel size for the
fleet ranged between 5 and 250 gross registered tons; the vessels included in our panel data
set were between 120 and 143 gross registered tons. The age of all vessels in the fleet was
between 3 and 60 years old. Hulls were constructed of steel, wood, steel and wood, and
fiberglass; all 10 vessels in our study panel were steel hulled. The reason we selected these
10 vessels was because they were relatively homogeneous in their characteristics. Using
data on other vessels with substantially different characteristics would require the analysis of
TE to consider differences in vessel characteristics (e.g., vessel size, gear configuration, and
hull material). It may be extremely difficult to separate the influence of resource conditions,
input levels, and skipper skill from the influences of vessel characteristics.
Stock size information was obtained from data routinely collected as part of a stock
monitoring program; Kirkley and DuPaul (1989) provide a detailed discussion of the data
collection program. The program began in January 1987 and obtained scallop samples and
detailed information from the last tow of approximately fifty dredge vessels per year. The
last tow was made strictly for the purposes of providing information on resource conditions,
product yields, and reproductive patterns. Detailed information obtained included number
of baskets of scallops harvested, geographic area of tow, time gear in water, water depth,
and size of dredge. Approximately 300 samples were obtained each year.
Stock size was calculated in this study for a given vessel and trip as the geometric mean
of baskets of scallops caught per hour from the last tow of all vessels participating in the
stock monitoring program and fishing the same area during the same period of time and
using the same dredge size. Catch per unit effort (CPUE) and landings per unit of effort
(LPUE) have been widely criticized as inadequate indicators of stock size (Richards and
Schnute, 1986). These data, however, are the only data available as an indicator of stock
size. In addition, they should be free of many of the biases associated with commercial
data since they were collected only for the purpose of monitoring stock conditions and the
reproductive cycle. Moreover, Dickie (1955) suggested that average catch per vessel or per
unit effort per trip is likely to be a good indicator of resource abundance of sea scallops; it
is at least likely to offer a viable indicator of resource density.
Data on captains were obtained from a 1990 survey of primary captain characteristics.
Captains change vessels during a year for a variety of reasons, including vacations, injuries,
and retirement, and therefore, the survey was limited to primary captains or those who took
the most trips aboard each vessel during 1990. Information obtained included race, age,
years of formal education, years of experience in fisheries, and years of experience in the
sea scallop fishery.
Mundlak (1961) noted that characteristics such as age and education may not be the best
indicators of managerial skill. They certainly may not reflect motivation which is likely to
be a critical characteristic of skill. Determining a measure of motivation, however, is likely
CHARACTERIZING MANAGERIAL SKILL AND TECHNICAL EFFICIENCY IN A FISHERY 151

to be extremely complicated. One possible measure would be the value of debt and assets;
for example, many of the “best” captains (those individuals with highest earnings per year)
have substantial investments (e.g., 1 or more rental properties and 1 or more luxury cars).
The captains intend to keep their investments or possessions, and thus, work extremely
hard. It is also possible that motivation allowed the captains to obtain the investments.
Unfortunately, it was not possible to obtain data on captains’ debt, assets, or wealth. We
therefore restricted our characterization of skipper skill to the relationship between TE and
primary captain attributes.
Data were also obtained on two other scallop dredge vessels operating in 1988. These
two vessels, however, were quite different from the 10 vessels. The hulls were wood and
steel and both vessels were 87 feet long. The two vessels were initially shrimp trawl vessels
constructed in 1981; they were converted to scallop dredge vessels in 1984. Relative to
the 10 dredge vessels, the 2 vessels had different configurations of gear and equipment.
Except for possible differences in crew amenities, the two vessels were identical in vessel
characteristics. The total number of observations on these two vessels was 30, or 15 per
vessel, and reflected fishing activities for the two captains in 1988.

4. Empirical Results

4.1. Parameter Estimates and Statistical Results

The translog stochastic production frontier, Eq. (4), was estimated by maximum likelihood
procedures available in Frontier 4.1 (Coelli, 1994). Frontier 4.1 utilizes a three-step pro-
cedure for estimating the maximum likelihood estimates. Initially, ordinary-least-squares
are used to obtain initial values for the parameters and variances. Next, a two-phase grid
search of γ is conducted with all β parameters, except β0 , set to the OLS values; the β0 and
σ 2 are adjusted according to the corrected ordinary least squares formula in Coelli (1994).
All other parameters, µ and δ, are set to zero during the two-phase grid search. Last, values
selected in the grid search are used as starting values in an iterative procedure to obtain the
final maximum likelihood estimates.
Many parameter estimates were statistically different from zero at the five percent level
of significance (LOS) (Table 1).3 Kim (1992), however, argues that interpretation of the
individual parameters of a translog may not be particularly meaningful. For the purpose of
assessing the fit of the estimated frontier, we report the adjusted R 2 from the ordinary-least-
squares regression of the model (R 2 = 0.829). Likelihood-ratio tests of different structures
rejected the Cobb Douglas form and the lack of interaction between crew and days at sea,
crew and stock size, and days at sea and stock size (χ62 = 90.13 vs critical χ6,0.05
2
= 12.60
for Cobb-Douglas and χ3 = 34.56 vs critical χ3,0.05 = 7.815 for no interaction terms).
2 2

Based on results of Coelli’s preferred one-sided likelihood ratio test, we concluded that
deviations from the frontier were due to technical inefficiency. The calculated χ26 2
= 290.55
and the one-sided critical value at the 5% level of significance is 35.60. We consider 26
restrictions because we impose zero restrictions on all δ parameters and on γ . The finding
of technical inefficiency suggests some scope for expanding production by raising TE.
152 KIRKLEY, SQUIRES AND STRAND

Table 1. Parameter estimates of stochastic production frontiera .

Variableb Estimatesc
Intercept 4.70∗
(1.76)
Days at sea 2.87∗
(1.43)
Labor −1.33∗
(0.23)
Stock abundance −0.54
(0.49)
Days at sea squared −0.16∗
(0.05)
Labor squared 0.44
(0.48)
Stock abundance squared 0.17∗
(0.03)
Days at sea∗ labor −0.23∗
(0.09)
Days at sea∗ stock abundance −0.25∗
(0.05)
Labor*stock abundance 0.51∗
(0.21)
σ2 0.17∗
(0.03)
γ 0.78∗
(0.06)
Constant −0.75∗
(0.34)
Dredge 0.66
(0.52)
1988 0.29∗
(0.13)
1989 0.75∗
(0.15)
1990 0.53∗
(0.14)
February −0.42∗
(0.20)
March −1.30∗
(0.52)
April −1.79∗
(0.64)
May −0.54∗
(0.20)
June −0.88∗
(0.33)
July −0.32∗
(0.12)
August −0.38∗
(0.19)
September −0.04
(0.17)
CHARACTERIZING MANAGERIAL SKILL AND TECHNICAL EFFICIENCY IN A FISHERY 153

Table 1. Continued.

Variableb Estimatesc
October 0.27∗
(0.10)
November 0.52∗
(0.17)
December 0.63∗
(0.17)
Boat #2 0.74
(0.54)
Boat #3 −0.01
(0.51)
Boat #4 0.22
(0.17)
Boat #5 −0.07
(0.51)
Boat #6 0.31∗
(0.17)
Boat #7 0.33
(0.18)
Boat #8 0.47∗
(0.17)
Boat #9 0.47∗
(0.17)
Boat #10 0.56∗
(0.17)
a Number of observations equals 581.
b All variables except intercept and dummy variables are
in natural logarithms.
c Numbers in parentheses are asymptotic standard errors.

Estimates for dredge through boat #10 correspond to the one-stage


estimation routine of Battese and Coelli (1993,1995); variables for dredge,
years, months, and boats are dummy variables.
∗ Statistically significant at the 5% LOS.

We also find a clear, temporal pattern in TE which is consistent with expectations about TE
for the fishery. Our estimated TE values on a per trip basis were lowest between September
and January and highest between February and August. The low periods coincide with
adverse weather, reproductive activities, and minimal availability of plankton or food for
scallops. Between September and December, scallops are undergoing gametogenesis in
which energy is being transferred from muscle development to reproductive activities.
Spawning typically occurs between October and late December. There also are frequent
offshore storms between September and November which limit the ability of vessels to fish.
Extremely cold and dangerous weather between November and January frequently limits
the ability of vessels and crew to fish. Last, plankton abundance is typically quite low
between October and January which limits food available to scallops. In contrast, plankton
is typically quite high between February and June. There are also fewer offshore storms.
Scallops do spawn between April and June; in fact, the spring spawn is the dominant spawn.
Given the apparent high abundance of plankton, however, gametogenesis does not appear
154 KIRKLEY, SQUIRES AND STRAND

Table 2. Technical efficiency, primary skipper characteristics, and economic per-


formance, 1990.

Captain Years Experience Captain Net


Efficiency Age Racea Fishing Scalloping Education Returns per day
0.875 63 C 50 45 3.0 196.2
0.863 42 C 20 10 14.5 181.0
0.833 42 C 15 15 16.0 165.8
0.832 39 C 22 20 14.0 182.0
0.804 31 H 11 11 12.5 159.9
0.768 35 C 15 14 12.0 183.6
0.766 37 C 14 14 9.0 158.9
0.743 31 C 14 14 8.0 149.9
0.709 44 C 30 9 6.0 182.0
aC indicates caucasian and H indicates hispanic.

to limit yields. It is also between February and August when the scallop resource is highly
abundant and dense. Recruitment occurs between March and July and spring is a major
growth season for scallops. Overall, our estimates of TE closely follow expected temporal
patterns, and thus, support the one-stage specification of Battese and Coelli (1993, 1995).
It is not too surprising that the results indicated that technical inefficiency dominated
uncontrollable random shocks. The nature of fishing may be hit or miss, but captains often
have a good idea of resource abundance, availability, and spatial distribution. Moreover, the
scallop is not a very mobile animal, and thus, the frontier may not be as variable as it would
be for finfish. There are several other possible reasons for technical inefficiency. Crew size
may be too large or small given resource conditions. Trips may be too long given resource
conditions; Mid-Atlantic vessels typically take long trips because of location of fishing
grounds. Vessels often spend two days steaming to and from the fishing grounds each trip.
Differences in expected catch rates motivate this behavior. Last, inclement weather and
mechanical or medical emergencies may force vessels back to dock earlier than planned;
inputs may thus appear to be excessive given output levels.

4.2. Technical Efficiency and Attributes of Captains

Detailed information on characteristics of the captains, unfortunately, was only available


for primary captains in 1990 (Table 2). We define the primary captain as the skipper that
took the most trips on a vessel in 1990. In all instances, the primary captain had over 75%
of the trips a vessel made. The captain having the highest efficiency rating was also the
oldest and the one with the greatest experience. He had a third-grade education. The 3
captains with the lowest efficiency ratings were individuals having an education level of
9th grade or less. The second highest average efficiency was for a captain having 2 years
of college and a technical school education but with less than the sample’s average years of
experience.
Although the number of observations for assessing the relationship between TE and
captains’ characteristics was extremely limited (9 observations), there appeared to be a
CHARACTERIZING MANAGERIAL SKILL AND TECHNICAL EFFICIENCY IN A FISHERY 155

Table 3. Parameter estimates and statistical results of


truncated regression model of TE vs. skipper charac-
teristics, 1990a .

Variable Parameter Estimate Standard Error

Constant 0.59954 0.0372b

Years of 0.01073 0.0024


Education

Years of 0.00512 0.0009


Experience
as Scalloper

σ 0.02456 0.0058
a Ordinary-least-squares adjusted R2 = 0.786.
b All parameters statistically different than zero
at the 1% level of significance.

possible relationship between average TE per captain per trip and education, age, and years
experience. Since TE is censored at 0.0 and 1.0, but unobserved at the limits, we utilize a
double-truncated model to examine the relationship between TE and education and years
experience in the scallop fishery.4 The one-stage approach of Battese and Coelli (1993)
would be the preferred approach, but the available data were not amenable to utilizing
the one-stage approach. Age was not included because it was highly collinear with years
experience.
Results of the truncated estimation suggested a statistically significant and positive rela-
tionship between TE and both education and years experience (Table 3). The adjusted R 2
from the ordinary-least-squares regression was 0.79. Although the number of observations
were too few to draw broad conclusions about the relationship between TE, years expe-
rience, and education, the results suggested possible substitution between education and
years of experience. Calculation of the marginal rate of substitution suggested that 1.0 year
of education could be substituted for approximately 2.1 years of experience.

4.3. The Exception: Technical Efficiency and Characteristics

Although basic characteristics such as years of education and fishing experience appear to
provide “crude” indicators of managerial skill, they as well as many other characteristics
may be inadequate indicators of skill. In this section, we explore TE for two relatively
homogenous or identical “highliners.” The two skippers are Mexican-American, both were
born in 1953, raised in the same town in Texas, graduated from the same high school in
1971, and trained by the same captain.5 Relative to 1988, the observation year used to
examine TE for these two skippers, the two skippers had 13 years experience in the scallop
fishery and 4 years as captain.
In 1988, the two individuals captained nearly identical vessels. The vessels were con-
156 KIRKLEY, SQUIRES AND STRAND

Table 4. Parameter estimates of stochastic production frontier of 2


homogeneous dredge vessels operating in 1988a .

Variable Estimates
Constant 2.79
[7.04]b

Days-at-sea 1.04
[8.79]

Crew Size 1.22


[7.49]

Stock Abundance 0.61


[4.74]

γ 0.99
[31.22]

Adjusted R 2 0.90
a Functional form of production function is Cobb-Douglas.
b Numbers in brackets are t-statistics.
c The one-sided likelihood ratio test of Coelli (1995) supports
the existence of a stochastic frontier;
χ12 = 5.299 vs critical one-sided value, at 5% LOS, of 2.71.

structed in 1981, built of wood and steel, had the same electronics and gear release/retrieval
systems, and the same engine horsepower. Nonetheless, one of the skippers consistently
outperformed the other and received higher net earnings. Both skippers are recognized as
highliners throughout the fishery, but the one captain is viewed as superior by other captains
and crew. In fact, the one captain often has a waiting list of potential crew and the turnover
of crew is quite low.
A stochastic frontier was estimated using 1988 data pooled over the two skippers (Table 4).
Given the limited number of observations (30 or 15 per vessel), the technology was restricted
to a Cobb-Douglas specification. The time-varying, error components specification of
Battese and Coelli (1992) was also considered. Battese and Coelli (1992) allow technical
inefficiency to systematically vary in accordance with the following specification: Uit =
(Ui exp(−η(t−T )) ), where Ui are non-negative random variables assumed to account for
inefficiency in production and to be iid as truncations at zero of the N (µ, σU2 ) distribution
and η is a parameter to be estimated. We were unable to support the time-varying model.
The estimated value of η was −.08406 and the corresponding asymptotic t-statistic was
1.02. We thus specified the time-invariant model with a half-normal distribution for the
inefficiency term.
Results support the view that technical efficiencies of the two captains were different; a
Kruskal-Wallis test rejected the equality of TE between the two captains (χ12 = 9.07 vs.
a critical value of 3.84 at the 5% LOS). Average TE per trip of one captain (captain 1)
equalled 107% of the average TE per trip of the other captain (captain 2). In 1988, captain
CHARACTERIZING MANAGERIAL SKILL AND TECHNICAL EFFICIENCY IN A FISHERY 157

1 produced $115,000 and 31,147 pounds more than captain 2. Captain 1 fished a total of
257 days while captain 2 was at sea 263 days in 1988. Profit before taxes for captain 1 was
11.3% higher than the before taxes profit realized by captain 2.
Captain one clearly outperformed captain 2. What are some possible factors that might
be responsible for the differences? For one thing, captain 1 appeared to make more frequent
changes in crew size and days at sea. The coefficient of variation (cv) for crew size per trip
was 19.2 for captain 1 and 10.02 for captain 2; the cv for days at sea per trip was 27.20 for
captain 1 and 23.60 for captain 2. If the relative cv is an indication of willingness to make
changes, then skipper flexibility may be considered an important characteristic of skipper
skill.
However, willingness to make changes in fishing strategy may indicate some other char-
acteristic of skipper skill (e.g., risk-taking behavior). Based on personal observations by
Kirkley on trips with the captains in 1988 and 1990, notable differences in the way the two
captains managed fishing activities were observed. The second captain is more relaxed and
spends considerable time in the wheel house. The first captain inspects the catch after every
tow and tends to make tows in soft-bottomed, sandy areas; the soft-bottomed areas typically
have large concentrations of scallops but require crew to pick through piles of gravel and
sand. The second captain spends less time on deck and generally avoids the soft-bottomed,
sandy areas. The first captain is more aggressive with the crew and frequently disciplines
and fines the crew.
The fact that there are differences in TE for the two captains fishing nearly identical vessels
and having nearly identical experience and backgrounds suggests there are likely to be other
possible indicators of managerial skill. Motivation which we can only subjectively assess
based on personal observation of the two captains is likely to be a very important component
of skill. Differences in the abilities of the first-mates could also explain differences among
the two captains; vessels fish 24 hours per day with the captain and mate each working
a 12 hour shift each day. We are left to conclude that while managerial efficiency does
depend on training and experience, it also depends on additional factors which we cannot
determine.

5. Concluding Remarks

Skipper skill is an important determinant of vessel productivity and TE. The better captains
or those with better managerial skills tend to have higher earnings, production, and TE. In
this study, two possible measures or indicators of skill—years experience in fishery and
education level—were found to explain a substantial proportion of estimated variation in TE.
Analysis also revealed some substitution possibilities between experience and education.
We could not, however, determine the minimum education level and years of experience
required of a captain (i.e., the thresholds or essential levels of education and experience).
Our results also suggest, however, that other characteristics need to be considered in
evaluating managerial skill. An analysis of TE derived from the stochastic frontier for two
captains of the same age, race, education, and experience and operating nearly identical
vessels found that one skipper was consistently more productive than the other skipper.
We documented the behavior of the better captain in terms of fishing practices and inter-
158 KIRKLEY, SQUIRES AND STRAND

actions with crew, but we could not derive the underlying characteristics of skill to explain
differences between the two captains.

Acknowledgments

We are especially appreciative of the assistance and advice offered by Finn Førsund, Knox
Lovell, and two anonymous reviewers. We are indebted to Tim Coelli for what amounted to
an e-mail, crash course on the use of his Frontier 4.1 estimation package and substantial help
with specifying an appropriate model. Ivar Strand is grateful to the Southwest Fisheries
Science Center for partial support while a Visiting Faculty Fellow. J. Kirkley is grateful
to the Virginia Sea Grant Program which partly funded this work. Any errors are the sole
responsibility of the authors. Results do not necessarily reflect the opinions of the National
Marine Fisheries Service, College of William and Mary, or the University of Maryland.

Notes

1. Bauer (1990) notes that the imposition of a specific one-sided error distribution is necessary to obtain
observation-specific estimates of TE. Lee (1983), however, suggests that a Lagrangean Multiplier test should
be used to examine the error distribution to avoid misspecification error. Schmidt and Lin (1984) argue that
the Lagrangean test as well as other tests are not particularly robust and considerably more research needs to
be done on determining the one-sided distribution. Bhattacharyya et al. (1995) recently offered, however, a
multistep estimation procedure for selecting the distributional form of the one-sided error closest to the data
generating process.
2. An important aspect of the analysis of TE is to access whether or not fishing trips of different lengths of time
(e.g., 16 vs. 23 days at sea) are related to different levels of efficiency. In our analysis, vessels frequently make
more than one trip per month. Estimation packages which permit estimation of time-varying efficiency do not
appear to be able handle multiple entries for the same firm and same time period.
3. Heteroscedasticity was examined by regressing vi2 against selected variables of the production function and their
squared values. Conventional F-tests rejected heteroscedasticity. Studentized residuals indicated three outliers
which coincided with extremely high stock values. Analyses were subsequently based on 581 observations.
4. The double-truncated and double-censored or Tobit model and the OLS parameter estimates are identical
since the limits are unobserved; the values of the standard errors from the tobit and double truncated models,
however, are not equal to those obtained by OLS.
5. Information about these two captains was obtained from 4 at-sea experiments with each captain in 1988 and
1990 for a total of 8 trips during the two years.

References

Acheson, J. (1975). “The Lobster Fiefs: Economic and Ecological Effects of Territoriality in the Maine Lobster
Industry.” Human Ecology 3, 183–207.
Aigner, D. J., C. A. K. Lovell, and P. Schmidt. (1977). “Formulation and Estimation of Stochastic Frontier
Production Function Models.” Journal of Econometrics 6, 21–37.
Anderson, L. G. (1989). “Conceptual Constructs for Practical ITQ Management Polices.” In Neher, P. A. Arnason,
and N. Mollett (eds.), Rights Based Fishing. Dordrecht: Kluwer Academic Publishers.
Battese, G. E., and T. J. Coelli. (1988). “Prediction of Firm-Level Technical Efficiencies with a Generalized
Frontier Production function and Panel Data.” Journal of Econometrics 38, 387–399.
Battese, G. E., and T. J. Coelli. (1992). “Frontier Production Functions, Technical Efficiency, and Panel Data.”
Journal of Productivity Analysis 3, 153–169.
CHARACTERIZING MANAGERIAL SKILL AND TECHNICAL EFFICIENCY IN A FISHERY 159

Battese, G. E., and T. J. Coelli. (1993). “A Stochastic Frontier Production Function Incorporating a Model for
Technical Inefficiency Effects.” Working Papers in Econometrics and Applied Statistics, No. 69, Department of
Econometrics, University of New England, Armidale, pp. 22.
Battese, G. E., and T. J. Coelli. (1995). “A Model for Technical Inefficiency Effects in a Stochastic Frontier
Production Function for Panel Data.” Empirical Economics 20, 325–332.
Bauer, P. W. (1990). “Recent Developments in the Econometric Estimation of Frontiers.” Journal of Econometrics
12, 24–42.
Bhattacharyya, A., T. R. Harris, R. Narayana, and K. Raffiee. (1995). “Technical Efficiency of Rural Water
Utilities.” Journal of Agricultural and Resource Economics 20(2), 373–391.
Bjorndal, T. (1987). “Production Economics and Optimal Stock Size in a North Atlantic Fishery.” Scandinavian
Journal of Economics 89, 145–164.
Bjorndal, T. (1989). “Production in a Schooling Fishery: The Case of the North Sea Herring Fishery.” Land
Economics 65, 49–56.
Campbell, H. F. (1991). “Estimating the Elasticity of Substitution between Restricted and Unrestricted Inputs in
a Regulated Fishery: A Probit Approach.” Journal of Environmental Economics and Management, 262–274.
Carlson, E. (1975). “The Measurement of Relative Fishing Power Using Cross-Section Production Functions.”
In J. Pope (ed.), Rapports et Proces-Vervaux des Reunions du Conseil International pour l’Exploration de la
Mer 168, 99–102.
Coelli, T. (1994). “A Guide to Frontier Version 4.1: A Computer Program for Stochastic Frontier Production and
Cost Function Estimation.” Department of Econometrics, University of New England, Australia.
Coelli, T. (1995). “Estimators and Hypothesis Tests for a Stochastic Frontier Function: A Monte Carlo Analysis.”
Journal of Productivity Analysis 6, 247–268.
Comitini, S., and D. S. Huang. (1967). “A Study of Production and Factor Shares in the Halibut Fishing Industry.”
Journal of Political Economy 75, 366–372.
Dickie, L. M. (1955). “Fluctuations in Abundance of the Giant Sea Scallop, Placopecten magellanicus (Gmelin),
in the Digby Area of the Bay of Fundy.” Journal of Fisheries Research Board of Canada 12, 797–854.
Hannesson, R. (1983). “Bioeconomic Production Function in Fisheries: Theoretical and Empirical Analysis.”
Canadian Journal of Fisheries and Aquatic Sciences 40, 968–982.
Hilborn, R., and M. Ledbetter. (1985). “Determinants of Catching Power in the British Columbia Salmon Purse
Seine Fleet.” Canadian Journal of Fisheries and Aquatic Sciences 40, 968–982.
Houghton, R. (1977). “The Fishing Power of Trawlers in the Western English Channel between 1965 and 1968.”
Journal du Conseil and the Exploration of the Sea 37, 130–136.
Huppert, D. D. (1989). Comments on R. Bruce Rettig’s “Is Fishery Management at a Turning Point? Reflections
on the Evolution of Rights Based Fishing.” In P. A. Neher, R. Arnason, and N. Mollett (eds.), Rights Based
Fishing. Dordrecht: Kluwer Academic Publishers.
Jondrow, J., K. Lovell, I. Materov, and P. Schmidt. (1982). “On the Estimation of Technical Inefficiency in the
Stochastic Frontier Production Model.” Journal of Econometrics 19(2/3), 233–238.
Kalirajan, K., and J. Flinn. (1983). “The Measurement of Farm-Specific Technical Efficiency.” Pakistan Journal
of Applied Economics 2, 167–180.
Kim, H. Y. (1992). “The Translog Production Function and Variable Returns to Scale.” Review of Economics and
Statistics 74, 546–552.
Kirkley, J. E., and W. D. DuPaul. (1989). “Science, Commercial Practices, and the Determination of Fishery
Regulations.” Journal of Shellfish Research 8, 139–149.
Lee, L. F. (1983). “A Test for Distributional Assumptions for the Stochastic Frontier Functions.” Journal of
Econometrics 22, 245–268.
Meeusen, W., and van den Broeck. (1977). “Efficiency Estimation from Cobb-Douglas Production Functions
with Composed Error.” International Economic Review? 18, 435–444. 109–122.
Mundlak, Y. (1961). “Empirical Production Functions Free of Management Bias.” Journal of Farm Economics
43, 44–56.
Noetzel, B. G., and V. J. Norton. (1969). Costs and Earnings in the Boston Large-Trawler Fleet. Bulletin 400,
University of Rhode Island.
Palsson, G., and P. Durrenberger. (1982). “To Dream of Fish: The Causes of Icelandic Skippers’ Fishing Success.”
Journal of Anthropological Research 38, 227–242.
Pollak, R. A., and T. J. Wales. (1987). “Specification and Estimation of Nonseparable Two-Stage Technologies:the
Leontief CES and the Cobb-Douglas CES.” Journal of Political Economy 95, 311–333.
160 KIRKLEY, SQUIRES AND STRAND

Richards, L. J., and J. T. Schnute. (1986). “An Experimental and Statistical Approach to the Question: Is CPUE
an Index of Abundance?” Canadian Journal of Fisheries and Aquatic Sciences 43, 1214–1227.
Rothschild, B. (1972). “An Exposition on the Definition of Fishing Effort.” Fishery Bulletin 70, 671–679.
Schmidt, P., and T-F Lin. (1984). “Simple Tests of Alternative Specifications in Stochastic Frontier Models.”
Journal of Econometrics 24, 349–361.
Squires, D. (1987). “Fishing Effort: Its Testing, Specification and Internal Structure in Fisheries Economics and
Management.” Journal of Environmental Economics and Management 14, 268–282.
Strand, I., J. Kirkley, and K. McConnell. (1981). “Economic Analysis and the Management of Atlantic Surf
Clams.” In L. G. Anderson (ed.), Economic Analysis for Fisheries Management Plans. Ann Arbor Science.
Thorlindsson, T. (1987). “The Skipper Effect in the Icelandic Herring Fishery.” Human Organization 47, 199–212.
Waldman, D. (1982). “A Stationary Point for the Stochastic Frontier Likelihood.” Journal of Econometrics 18,
275–279.
Zellner, A. J., J. Kmenta, and J. Dreze. (1966). “Specification and Estimation of Cobb-Douglas Production
Function Models.” Econometrica 34, 784–795.

You might also like