Scholarly article on topic 'Decomposing productivity and efficiency changes in the Alaska head and gut factory trawl fleet'

Decomposing productivity and efficiency changes in the Alaska head and gut factory trawl fleet Academic research paper on "Economics and business"

CC BY-NC-ND
0
0
Share paper
Academic journal
Marine Policy
OECD Field of science
Keywords
{"Total factor productivity" / Efficiency / "Fishery cooperative" / Rationalization / "Productivity decomposition"}

Abstract of research paper on Economics and business, author of scientific article — Benjamin E. Fissel, Ronald G. Felthoven, Stephen Kasperski, Christopher O’Donnell

Abstract Fishing fleets are subject to numerous factors that affect economic performance, making identification and attribution of such impacts difficult. This paper separately identifies the effects of changing input and output prices, fishery management, and quota allocations on total factor productivity using a Lowe Index. Indices account for technical change and decompose productivity estimates into its technical, environmental, and scale-mix components. This results in measures that reflect shifts in the production frontier, and movements by vessels toward and around the frontier, to capture economies of scale and mix after a policy shift to a catch share program that includes fishing cooperatives and a limited access fishery. The difference between cooperative and limited access vessels is exploited to compare the changes in economic performance between the groups after the introduction of the shift to catch shares and cooperative management, which allowed the vessels to improve the timing and coordination across multi-species fisheries and to decrease incidental catch of quota-limited bycatch species that had closed the target fisheries prematurely in the past. Results indicate that total factor productivity increased significantly after the move to a catch share program, largely due to increases in technical change that shifted out the production frontier of the fishery.

Academic research paper on topic "Decomposing productivity and efficiency changes in the Alaska head and gut factory trawl fleet"

MARINE POLICY

Marine Policy I (I

Contents lists available at ScienceDirect

f^Wi Marine Policy

ELSEVIER journal homepage: www.elsevier.com/locate/marpol

Decomposing productivity and efficiency changes in the Alaska head and gut factory trawl fleet

Benjamin E. Fissela'*, Ronald G. Felthovena, Stephen Kasperskia, Christopher O'Donnellb

a NOAA's National Marine Fisheries Service, Alaska Fisheries Science Center in Seattle, USA1 b The University of Queensland, Australia

ARTICLE INFO

ABSTRACT

Article history: Received 16 June 2015 Accepted 17 June 2015

Keywords:

Total factor productivity Efficiency

Fishery cooperative Rationalization Productivity decomposition

Fishing fleets are subject to numerous factors that affect economic performance, making identification and attribution of such impacts difficult. This paper separately identifies the effects of changing input and output prices, fishery management, and quota allocations on total factor productivity using a Lowe Index. Indices account for technical change and decompose productivity estimates into its technical, environmental, and scale-mix components. This results in measures that reflect shifts in the production frontier, and movements by vessels toward and around the frontier, to capture economies of scale and mix after a policy shift to a catch share program that includes fishing cooperatives and a limited access fishery. The difference between cooperative and limited access vessels is exploited to compare the changes in economic performance between the groups after the introduction of the shift to catch shares and cooperative management, which allowed the vessels to improve the timing and coordination across multi-species fisheries and to decrease incidental catch of quota-limited bycatch species that had closed the target fisheries prematurely in the past. Results indicate that total factor productivity increased significantly after the move to a catch share program, largely due to increases in technical change that shifted out the production frontier of the fishery.

Published by Elsevier Ltd.

1. Introduction

Fishing fleets are subject to numerous factors that affect economic performance, including those under their control (e.g., input and output selection) and those exogenously determined (e.g., market prices, regulations, stock levels, and quota allocations). As such, analysts interested in monitoring the economic performance of fishing fleets, or identifying the impacts of particular policies, have the daunting task of identifying relevant and feasible metrics given available data, and decomposing the metrics to inform the questions of interest. Total factor productivity (TFP), the focus of this study, is a commonly employed metric to evaluate performance. TFP change is the quantity change component of profitability change, and it can be estimated under a wide range of data availability scenarios. For an introduction to the analysis of productivity and firm performance, see Grifell-Tatje and Lovell [1]. However, no fishery TFP studies to date have decomposed estimated TFP change into its components including technical change,

* Corresponding author. E-mail address: Ben.Fissel@noaa.gov (B.E. Fissel).

1 OMB disclaimer: The findings and conclusions in the paper are those of the authors and do not necessarily represent the views of the National Marine Fisheries Service.

http://dx.doi.org/10.1016/j.marpol.2015.06.018 0308-597X/Published by Elsevier Ltd.

environmental efficiency, technical efficiency and scale-mix efficiency. Measuring the specific components can be particularly informative, as effects of fishery management policies are likely to be lumped in with other factors in conventionally produced composite productivity residuals. These measures allow the identification of specific changes due to scale increases from vessel consolidation, changes in relative efficiency after the introduction of fishing cooperatives, and potential cost-saving input reallocation due to ending the race for fish.

This paper analyzes changes in TFP and its components for one of the more prominent fishing fleets operating in the North Pacific: the Alaska head and gut factory trawl fleet, which will be referred to in this paper as the Amendment 80 (A80) fleet. With the implementation of A80 to the Bering Sea and Aleutian Islands (BSAI) Fishery Management Plan (FMP) in 2008, this fleet transitioned from a common-pool fishery in which vessels competed for a share of the allowable catch to a catch share fishery comprised of cooperatives who were given an allocation of fish; there is also a

2 B.E. Fissel et al. / Marine Policy I (l

limited access fishery for those who do not join cooperatives that operates under slightly different rules. Forming a cooperative has several advantages including having an exclusive harvesting privilege based on the catch history of member vessels, receiving an allocation of incidentally caught prohibited species catch (PSC),2 and having to meet increasingly restrictive groundfish retention standards at the co-op level rather than at the vessel level. One cooperative formed in 2008 included 16 of the 24 eligible vessels and the remaining 8 vessels fished in the limited access fishery until 2011 when they formed their own cooperative and eliminated the limited access fishery.

The cooperative structure has allowed the vessels to improve the timing and coordination across multi-species fisheries and to decrease incidental catch of quota-limited PSC species that had closed the target fisheries prematurely in the past [2]. In fact, it was the historically high levels of PSC and discards of target species in the past that led the North Pacific Fishery Management Council (Council) to simultaneously implement heighted catch retention standards and the option for a quota-based cooperative structure which was believed to provide the flexibility to meet the new standards [3]. Catch data from the fishery indicate that the program has been a success in improving retention, as nearly all of the vessels exceeded the standard [4]. This paper seeks to investigate whether other improvements in productivity or efficiency resulted from the management change and whether these improvements differed between the fleet as a whole and those who participated in cooperatives.

Changes in economic performance are examined using a Lowe TFP index, described in more detail in a following section, to separately identify the effects of changing input and output prices, catch share program implementation, and quota allocations on TFP. Specifically, Lowe TFP indices can be decomposed into measures of technical change and measures of technical, environmental, and scale-mix efficiency change. This decomposition results in measures that reflect movements in the production frontier, movements by vessels towards the frontier, and movements by vessels around the frontier to capture economies of scale and scope through changes in the catch portfolio after the policy shift to catch shares. Because of the differential adoption of the cooperative structure over time the analysis is able to distinctly evaluate the economic performance of those participating in the limited access fishery from those who are fishing in cooperatives.

The next section provides an overview of the literature on estimating productivity in fisheries, followed by a detailed exposition of the chosen methodological approach. The following section gives details on the fishery under study and data used in the analysis. This is followed by a description of the econometric model and results. The last section includes a discussion of the findings and some conclusions from the analysis.

2. Relevant literature

Productivity in fisheries has a considerable history of interest to both fishery managers and researchers, but there have been varied and somewhat inconsistent approaches for tracking performance by analysts. This is likely in part due to data availability dictating the approach, as fisheries are notoriously data scarce, but also due to competing methodologies with varying degrees of rigor and/or restrictive assumptions.

Numerous studies in the fisheries productivity literature

2 PSC is a special category of bycatch of Pacific salmon, Pacific halibut, and king and Tanner crab in the groundfish fisheries that cannot be retained or sold by the vessel. See 50 CFR Part 679 available at: http://alaskafisheries.noaa.gov/regs/ part679_all.pdf, for more detail.

provide context for this analysis. Walden et al. [5] provides an extensive list of such studies, starting with the work by Comitini and Huang [6] which used a parametric approach to estimate a Cobb-Douglas production function representing halibut fishing vessels in the North Pacific. Jones et al. [7] used data from a sample of South Carolina's shrimp trawlers to analyze resource productivity and profitability in the fishery from 1971 to 1975. Norton, Miller and Kenney [8] created an Economic Health Index using data from several U.S. fisheries to estimate productivity across fisheries. These early studies suffered from identification issues because they failed to account for the influence of the resource stock in productivity growth. Squires [9-11] broadened the literature by employing index number and growth accounting theory and using the resource stock to disentangle changes in productivity and resource abundance in the Pacific Coast Trawl Fishery. Weninger [12] generalized the fisheries TFP index by using a non-parametric, directional distance function to examine changes in productivity for surfclam vessels. Jin et al. [13] used a growth accounting approach to conduct a broad total factor productivity study for the entire total New England groundfish fishery covering the years 1964-2003. Hannesson [16] examined different specifications of TFP change in Norwegian fisheries, emphasizing the importance of accounting for resource abundance in productivity growth. Felthoven and Paul [14] reviewed past productivity studies and suggested ways that the methodological approaches commonly employed could be improved to reflect many of the idiosyncrasies of fisheries settings.

Building on this foundational research, empirical fisheries studies routinely employ the productivity framework to examine the state of the industry and relationship between policy and productivity. After a license buyback in Australian fisheries, Fox et al. [15] examined the resulting changes in productivity, capacity, and quota trading. Squires et al. [17] used data envelopment analysis to estimate a Malmquist index to examine productivity growth in the Korean tuna purse seine fishery operating in the Pacific Ocean. Felthoven et al. [18] implemented their 2004 framework in a primal, growth accounting application to the Alaskan pollock fishery covering 1994-2003, incorporating environmental conditions, bycatch and stock effects into their model. Paul et al. [19] subsequently expanded this analysis to examine productivity change in both harvesting and processing using a dual, revenue function approach. Walden et al. [20] updated Weninger's aforementioned 2001 analysis to examine productivity change in the surfclam and ocean quahog ITQ fishery using a Malmquist index. Eggert and Tveteras [21] examined productivity change in Icelandic, Norwegian and Swedish fisheries between 1973 and 2003 using a growth accounting framework. Torres and Felthoven [22] conducted a study similar to their earlier work in the Alaskan pollock fishery using a longer panel (1994-2009) and improved econometric techniques to account for the mixed distribution of the production data within a revenue function specification. Most recently, [5] used the Lowe index to measure multi-factor productivity change for all the U.S. catch share fisheries, including the fishery under study, but did not decompose TFP into the efficiency components, did so for a shorter time series, and accounted for fewer inputs.3

The objective of this research is to use a rigorous set of tools grounded in production theory to examine the ways in which the policy change and resulting behavioral adjustments may have affected the economic performance of the fleet. This is accomplished using an approach rooted in [23], and subsequently developed

3 As part of a national effort to estimate productivity for all U.S. fisheries managed bycatch shares, the authors agreed to utilize the same methodology across fisheries, so data limitations in some fisheries necessitated a simpler approach across all fisheries in that study.

B.E. Fissel et al. / Marine Policy I (■■■■) Ill-Ill 3

further in a decomposition methodology proposed by O'Donnell [24,25]. This approach satisfies all economically relevant axioms for index number theory, including monotonicity, homogeneity, identity, commensurability, proportionality, transitivity, and time and space reversal [25]. It also avoids many of the shortcomings and common assumptions in productivity estimation as discussed in the survey paper by Felthoven and Paul [14]. Most notably, this approach also permits the decomposition of estimates of total factor productivity into technical, environmental, and scale-mix components, thereby providing greater insight in the dimensions in which vessel operators are successfully exploiting the greater flexibilities afforded by the way the fishery is now managed.

3. Fishery description

The measures of economic performance described up to this point were constructed for the A80 fleet, a group of catcher-processor (CP) vessels that account for most of the flatfish (e.g., yel-lowfin sole, rock sole, flathead sole) caught in the Bering Sea, as well as Atka mackerel and Pacific ocean perch caught in the Aleutian Islands. The fleet comprises 24 medium to large CP vessels (average length is 159 feet) using bottom trawl gear with limited factory space and processing capability. From 2008-2010, the majority of vessels (16, between 14 and 16 of which were active) were in one cooperative (The Alaska Seafood Cooperative (AKSC)),4 with the remaining 8 (only 6 active) vessels being in the limited-access fishery. Starting in 2011, CPs in the former limited access fishery formed the Alaska Groundfish Cooperative (AKGC), leading to all vessels in the fleet belonging to one of two cooperatives. Since the formation of the AKGC, there have only been 4 active vessels in that cooperative, while the AKSC had 16 and 15 active in 2011 and 2012, respectively. These voluntary harvest cooperatives manage the target allocations and PSC allocations amongst themselves. For a more complete description of the fishery, see Ref. [26]. The statistics in Table 1 summarize the annual production, value and effort variables of the fleet.

These CPs typically use bottom trawl gear to catch and then freeze fish whole or with the head and guts removed. Because of the multi-species nature of the fishery, and each species having a different relative value as well as size and roe content, discarding has historically been a management concern for this fleet. Competing for catch led vessels to discard many of the less valuable fish in order to maximize the value of the fish going through their processing lines and into limited freezer space on trips which often last several weeks between offloads. Binding constraints on the discarded species at times meant that target fisheries were closed and valuable fish were left in the water. In response to these concerns, the Council required full retention of pollock and Pacific cod, and a minimum groundfish retention standard of 85%.5 To assist the fleet to increase retention, in 2008 the Council initiated A80 to the BSAI FMP.6

4 Prior to 2010, it was known as the "Best Use Cooperative".

5 This groundfish retention standard was phased in at 65% for 2008, 75% for 2009, 80% for 2010, and 85% for 2011 and thereafter. However, the groundfish retention standard was subsequently removed in March 2013 after compliance and enforcement costs were higher than anticipated for both NMFS and the Amendment 80 sector and the fleet achieved a retention rate higher than the standard once operating under a cooperative program. See 79 FR 12627, available at: http:// alaskafisheries.noaa.gov/frules/78fr12627.pdf.

6 The North Pacific Fishery Management Council designed the Amendment 80

Cooperatives Program to allow eligibility based upon those persons who: (1) did not meet the qualification criteria of an American Fisheries Act trawl catcher/processor sector as defined in Section 219(a)(7) in the American Fisheries Act; and (2) held a portion of the catch history of Amendment 80 species during the period from 1998 to 2004.

Initial allocations were issued to eligible vessel owners based on catch history. A80 allocates Atka mackerel, Pacific cod, Aleutian Islands Pacific ocean perch and three species of flatfish (yellowfin sole, rock sole and flathead sole), as well as sideboards7 for pollock, Pacific cod, Pacific ocean perch outside of the Aleutian Islands, northern rockfish, pelagic shelf rockfish and a prohibited species catch allocation for halibut and crab. Quota share-holders may, on an annual basis, elect to form a cooperative with other A80 quota share-holders to receive an exclusive harvest privilege for the portion of the catch limit resulting from their aggregated quota share holdings. Annual allocations of quota can be leased annually within and between eligible cooperatives, but quota shares can only be transferred along with the vessel and catch history. Those quota share holders electing not to join a cooperative participate in the Amendment 80 limited access fishery, which receives an overall allocation equal to the sum of what participants would have been allocated had they joined a cooperative. These individuals do not, however, receive an exclusive harvest privilege for any of the A80 species or PSC allocation that was allocated to the limited access fishery on their behalf. Each vessel in the limited access fishery is required to maintain the groundfish retention standard while the cooperative vessels only need to meet the requirement as a cooperative in aggregate.

A80 also changed the way that PSC is allocated across species. PSC used to be allocated based on anticipated PSC usage for each target species in the fishery but after A80, an overall PSC allocation is provided to each cooperative (and the limited access fishery) that can be utilized in any target fishery. As a result, the limited access fishery can still lead to a race among vessels for both the limited access share of the TAC and the PSC allocation, while the cooperatives have a specific allocation of the TAC and PSC that can be fished by any cooperative member vessel throughout the year. The effects of the derby style limited-access fishery can exacerbate problems in years where TACs or PSC limits are binding. This leads to the hypothesis that the limited access fleet would not experience the same gains from rationalization as the cooperative sector that has increased flexibility in dealing with PSC and the heightened groundfish retention standards.

4. Methods

Measures of firm performance (e.g., profitability, productivity, efficiency) can be used to inform economic decision-making. However, to better understand observed performance one should compare it with some benchmark, such as the maximum level of performance that is technically feasible, and strive to understand the reasons why entities might be performing at different levels. This research focuses on the total factor productivity of A80 vessels to assess whether changes in productivity and efficiency have occurred since implementation of the catch share policy, and whether such changes are different for those participating in cooperatives. The analytical framework underpinning the paper is an aggregate quantity-price framework developed by O'Donnell [24,25,27-30].

4.1. Production technologies

To obtain estimates of the measures that characterize different dimensions of performance for this fleet the production technology must be modeled. In O'Donnell [30], production technology is

7 Sideboards essentially limit catch on species not included in the quota-share based cooperative allocation. They are intended to limit the ability of vessels in rationalized fisheries from exceeding historic levels of participation in other fisheries.

4 B.E. Fissel et al. / Marine Policy I (l

Table 1

Summary of A80 fleet production and effort.

Year Vessel count Total days Product value (2010 USD) Product weight (mt) Total fuel use (gal) Crew days Average length (ft)

2005 22 4669 $ 239,511,186 141,637 21,451,945 171,302 172

2006 22 4543 $ 248,603,206 138,970 16,684,075 166,032 172

2007 22 4452 $ 252,796,723 133,539 14,243,051 162,061 172

2008 22 4377 $ 278,662,808 170,508 12,551,890 156,482 172

2009 21 3719 $ 229,714,467 161,538 10,663,954 137,739 169.3

2010 20 3924 $ 273,465,897 180,469 11,135,063 149,216 172.5

2011 20 3728 $ 346,776,392 183,678 11,342,043 140,107 172.9

2012 19 3723 $ 353,671,935 187,391 10,627,799 141,041 176.3

defined as a technique, method or process for transforming inputs into outputs. Furthermore, the set of all technologies available in period t is referred to as the period-t metatechnology. It is convenient to think of a technology as a book of instructions and the period-t metatechnology as a library. Let q = (q1, ..., qN*)' and x = (x1, ..., xM*)' denote vectors of outputs and inputs, respectively. The set of output-input combinations that are possible using the period-t metatechnology in a production environment characterized by z = (z1, ..., z**)' is formally defined as:

T'(z) = {(x, q): x can produce q in period t in environment z} (1)

The same regularity assumptions as [31, p. 26, 27] are maintained, namely that inactivity is possible, output sets are closed and bounded, input sets are closed, there is no free lunch, and outputs and inputs are weakly disposable. Under these assumptions, an equivalent representation of Tt (z) is the output distance function (ODF), formally defined as

DO(x, q, z) = inf{5 > 0: (x, q/S) e Tt(z)}.

The ODF is nonnegative (NN) and homogeneous of degree one (HD1) in outputs. O'Donnell [30, Proposition 8] shows that if technical change is implicit Hicks output neutral (IHON) and the metatechnology is output homothetic (OH), then the ODF takes the form DO(x, q, z) = Q(q)/Ft (x, z) where Q(q) = Dq(x, q, z) is an aggregate output and Ft (x, z) is a production function. In practice, it is common to assume that technical change is IHON, that the metatechnology is OH, and that the production function is of the Cobb-Douglas form:

ln Ft (x, z) = Y (t) + £ pj ln zj + £ pm ln xm

j=i n=i (2)

where £m pm = r and y (t) is a technology shifter than can be specified in various ways to accommodate technical change. The metatechnology exhibits decreasing returns to scale, constant returns to scale or increasing returns to scale as r is less than, equal to, or greater than one. If inputs are strongly disposable, then pm > 0 for all m. This paper estimates the specific form of the production function represented in Eq. (2), including fixed effects and a specific form of y (t), to derive components of interest such as efficiency measures and productivity indices, described in turn below.

4.2. Measures of efficiency

Vessel and time subscripts are introduced into the notation so that, for example, qit = (qit1, ..., qitN*)' now denotes the outputs of vessel i in period t. O'Donnell [25] defines the TFP of this vessel as TFPit = Q (qit)/X (xit) where Q (.) and X (.) are NN, ND and HD1 aggregator functions. An associated measure of overall productive performance is the measure of TFP efficiency (TFPE) defined by

O'Donnell [25, p. 880]:

TFPEit = TFPit/TFP*, (3)

where TFPt* denotes the maximum TFP that is possible in period t. TFPE can be written as the product of a measure of vessel efficiency (VE)8 and a measure of environmental efficiency (EE). As the names suggest, VE is the component associated with the vessel, and EE is the component associated with characteristics of the production environment. This separation is useful in decomposing changes in performance associated with improvements in the vessel or equipment or the operator's skill from, say, improvements associated with changes in ocean conditions, species growth rates or other factors that are outside the operator's control (i.e., environmental variables). The VE and EE of vessel i in period t are formally defined as [30]:

VEit = TFPit/TFP' (zit) (4)

EEit = TFP1 (zit )/TFP* (5)

where TFPt (zit) is the maximum TFP that is possible using the period-t metatechnology in an environment characterized by zit. VE can be further decomposed into a measure of output-oriented technical efficiency (OTE) and output-oriented scale-mix efficiency (OSME). OTE measures reflect a vessel's ability to maximize catch, while OSME reflects its ability to exploit scale and scope economies, and change input and catch composition. Changes in OSME occur as operators take advantage of the relative contribution of inputs or the profitability of different species, which one would expect to be augmented under catch shares. The OTE and OSME of vessel i in period t are formally defined as [30]:

OTEit = DO (xit, qit, zit) (6)

OSMEit = VEit/OTEit. (7)

OTE is a measure of efficiency that is usually attributed to Refs. [32] and [33]. OSME is a measure of economies of scale and scope defined by [27]. Eq. (7) says that OSME is the component of VE that remains after the OTE component has been removed.

4.3. Productivity indices

Because the interest lies in comparing productivity across vessels over time, the index that compares the TFP of vessel i in period t with the TFP of vessel k in period s is:

TFPIksit = TFPit/TFPks = Qlksit/XIksit (8)

where Qkit = Q (qit)/Q (qks) and Xlksit = X (xit )/X (xks) are output and input quantity indexes. Any NN, ND and HD1 aggregator functions

8 O'Donnell [30] couches his discussion in the context of firms and thus uses the notation "FE", whereas the vessel-based discussion in this analysis uses of the term "VE."

B.E. Fissel et al. / Marine Policy I (l

can be used to construct measures of TFP and TFP change [30]. If the aggregator functions are linear, and if average prices are used as weights (i.e., measures of relative value), then Q (qit) <x p'qit and X (xit) « w'xit where ¡5 and w are vectors of average output and input prices respectively. The associated output and input quantity indexes are indexes that are usually attributed to Ref. [23]:

QJksit = (p ' Qit )l(p ' qks )

Xlksit = (w%t llwXks )

These indexes are proper indexes in the sense that they satisfy the weak monotonicity, linear homogeneity, identity, homogeneity of degree zero, proportionality, time-space reversal, transitivity and circularity axioms listed in Ref. [30]. The associated TFP index is the Lowe TFP index proposed by Ref. [25]:

p'qit) w'xit

p 'Qks , { WW'Xks

exp( l-tf)" r J* n J=1 zjit Pf ' exp(-Uit)"

exp( i-YS) , zjks, _ exp(-Uks)_

exp(Vit)

exp(Vks)

The first term in square brackets on the right-hand side is a measure of technical change (TI), the second term is a measure of environmental efficiency change (ZI), the third term is a measure of technical efficiency change (OTEI), and the fourth term is a measure of scale-mix efficiency change (OSMEI), and the last term is a statistical noise index (NOISEI). If there is no statistical noise, then (14) collapses to the decomposition (12). All of this information is easily obtained from the data used in this analysis and resulting parameter estimates from estimating Eq. (13) as a stochastic production frontier discussed later in the paper.

5. Data

Any proper TFP index can be decomposed into measures of technical change, environmental change, and efficiency change. For example, the efficiency measures discussed in Eqs. (3)-(7) im-plythat for vessel i in period t, TFPit = TFP* x EEit x OTEit x OSMEit. Thus, the index defined by Eq. (8) can be decomposed to compare this vessel-year performance with that of vessel k in period s as:

TFPIksit = TIst x Zlksit x OTEIksit x OSMEIksit (12)

where TIst = TFP*/TFP* is the measure of technical change defined by [24 pp. 537], ZIksit = EEit/EEks is the measure of environmental (efficiency) change defined by Ref. [30], OTEIksit = OTEit/OTEks is an index of output-oriented technical efficiency change, and OSMEIksit = OSMEit/OSMEks is an index of output-oriented scale-mix efficiency change.

4.4. Model specification

The Lowe TFP index (Eq. (11)) can be computed simply with available data on input and output prices and quantities. By estimating the production function in Eq. (2) the decomposition given by Eq. (12) can be used to identify changes in the various components of the TFP index.

The econometric specification uses the assumption that the technical change is IHON and linear in t, the metatechnology is OH, and the production function is the Cobb-Douglas form, leading to a production function of the form:

ln Qit = Yo + Yit + £ Pj ln zj + £ m ln Xm + Vit - Uit

j=1 m=1 (13)

where Qit = ¡j'qit is a Lowe aggregate output, t is a time trend, zjit is an environmental variable, xm is an input, vit represents statistical noise and uit is a technical inefficiency effect. Furthermore, it is assumed that vit~iidN(0, a2), uit~iidN+(0, ) and vit and uit are distributed independently of one another and the other regressors.9 If these conditions hold and there is no endogeneity, consistent estimates of all parameters are obtained with maximum likelihood estimation. In addition, with this particular specification, the Lowe TFP index in Eq. (11) can be decomposed as in Eq. (12):

9 The notation N+ refers to a half-normal distribution.

The data relied upon to estimate the production technology and various indices discussed thus far include information on output quantities and prices derived from weekly or daily production records of total product weight (in metric tons) and product value, respectively, from all species caught on days where some amount of any A80 species was also produced.10 Input quantity and price data were included for labor, capital, and fuel using records from federal vessel registration files, daily production reports, and mandatory annual economic data reports (EDR) which track input use and cost. The labor input is defined as the number of days the vessel was in the A80 fishery multiplied by the crew size for those days. Total annual labor cost data from the EDR is divided by the number of production days in the A80 fishery to estimate a daily wage for crew members. The annual capital quantity used is the product of vessel length, horsepower and annual days at sea.11 The price of the capital quantity is taken from the rate for BAA grade bonds following [34]. The total fuel used and total fuel expenditures (used to calculate average fuel prices paid) by the fleet comes from the EDR data. As the EDR data are annual, all EDR data are prorated using the share of a vessel's total processing days where they processed any A80 species. One of the primary differences between the TFP estimates in this paper and those in [5] is the imputation of fuel use in the pre-A80 period, whereas [5] estimated TFP using capital, labor, and fuel use only for the post-A80 period. While it is possible to estimate TFP metrics using only capital and labor data over the entire time period, this application requires additional steps to impute fuel use in the pre-A80 period as described in the following section. This paper estimates TFP for both the pre-A80 and post-A80 period with greater data resolution because it is important to capture the changes in productivity resulting from the implementation of A80, and including the change in fuel use is an important factor to

10 Prior to 2008, CPs were required to submit weekly production reports and therefore production data is summarized at the week level prior to 2008 and daily for 2008 to present. In either case, the data used in the study are annual aggregates.

11 Attempts were made to utilize an economic measure of capital, the estimated value of the vessel, which is self-reported annual in the EDR reports. Inspection of these data revealed considerable variation in values among seemingly similar vessels in terms of production composition and value, as well as vessel size. This could potentially be due to respondents interpreting the determinants of vessel value in different ways, despite instructions clarifying what to include and not include. Regardless, the values for some may include items such as the perceived value of vessel permits in addition to the replacement value of the vessel. Use of this capital variable in the models resulted in insignificant parameter estimates for capital, and given these findings and data quality concerns discussed thus far, a more traditional measure of physical vessel capital was used.

B.E. Fissel et al. / Marine Policy I (l

consider.

5.1. Input imputation for pre-A80 years

One of the goals of this research is to examine the change in economic performance arising from the transition to catch shares and cooperatives. This requires one to examine performance before and after the policy change. Unfortunately, the collection of cost data in the EDR was coincident with the introduction of Amendment 80 in 2008. In order to conduct the pre- and post-policy comparisons some of the input data on labor and fuel costs had to be imputed for the pre-A80 years (2005-2007) included in this study. Labor constituted the largest share of operating costs with an average 52.5% of the combined labor, food and fuel costs for 2008-2012. While estimates of crew days were available for the entire sample period, crew wages were not collected until 2008 and needed to be imputed for earlier years. Construction is known to be a common off-season job for fishermen and construction wages may be considered to be a reasonable reservation wage for crew. Visual inspection of the variation in construction wages was comparable to that of the A80 fleets' cost-per-day of labor for the years 2008-2012. Therefore, wages are imputed for 2005-2007 by estimating a regression of daily wages using construction wage data from the Bureau of Labor Statistics Occupational Employment Statistics database (http://www.bls.gov/oes/) and vessel-specific dummy variables. The fitted model had an R-squared of 0.78 indicating a reasonably good fit.

Fuel is the second most significant variable cost of fishing. Between 2008-2012 fuel costs averaged 32% of the combined labor, food and fuel costs. Fuel is a potentially more volatile aspect of vessel operating costs because fuel prices tend to be more variable than other input prices and fuel consumption is dependent upon the number of days fished, targeting strategies and location choices associated with finding fish and avoiding incidental catch of prohibited species. Fuel use is imputed by specifying a regression in which fuel use is dependent upon vessel length, fishing effort and fuel prices. Effort was measured through processing days, crew days and production weight. Fuel prices were calculated from EDR fuel costs divided by the fuel use for the years after 2008, and prior to 2008 fuel prices are derived from Dutch Harbor Alaska (a large fuel supplier and port for many A80 ships). A time trend is also included to help account for unspecified trends (e.g., in demand) in prices.12

5.2. Input and output indices

Lowe input and output indices were constructed to characterize total factor productivity. The Lowe output (input) indices are calculated using average prices calculated as the ratio of revenue (cost) over output production (input use) totaled across time and vessels. Comparison of vessel level input and output indices (X and Q, respectively, in the graph) shows an increase in productivity after rationalization of the A80 fleet (Fig. 1).

Examination of output indices averaged over vessels shows that aggregate output has been steadily increasing (Fig. 2). Output indices of rock sole and non-A80 species show discernible increasing trends in the post-A80 period13; these species/groups comprised 13.7% and 17.3% of the average revenue shares, respectively, and hence were influential drivers of the upward trend in the aggregate output index. Most other species do not show a

12 Data imputation regression results are available from the authors. upon request.

13 The non-A80 species that lead to the large increase in the output index after

the implementation of A80 consists primarily of increases in catch of arrowtooth

flounder, Greenland turbot, and Alaska plaice.

Inputs (X)

Fig. 1. Vessel outputs and inputs pre- and post-rationalization.

distinct post-A80 trend. The exception is flathead sole (4.5% revenue share) which is the only species whose output index decreased. The aggregate input index is decreasing in the pre-A80 period, driven primarily by decreases in fuel use, which constituted a third (28.7%) of the input cost share on average. The labor (52.5% cost share) index and capital index (18.8% cost share) show similar variation over time and in the post-A80 period were the most influential factors determining aggregate input variation. While post-A80 aggregate input use has shown some variation, there is no apparent upward trend as with output.

6. Results

The production function described in Eq. (13) was estimated as a stochastic frontier using annual data for each vessel for the years 2005-2012. The dependent variable, Qit, is the vessel-specific annual sum of the product weight from all species, weighted by the average price for each species. Three classes of explanatory variables were used in the final specification of the stochastic frontier model: technology, input use, and environmental variables. The technology variables included a time trend (Trend), a post-rationalization dummy variable (Post-Rat), a post-rationalization trend (Trend*Post-Rat), and dummy variables for post-rationalization participation in the limited access fishery (Post-Rat LA), membership in the Alaska Seafood Cooperative (AK Seafood Co-op) or the Alaska Groundfish Cooperative (AK Groundfish Co-op) and trend interactions with the Co-op status variables. The input factors, x, used were fuel (Fuel), labor (Labor) and capital (Capital). Vessel fixed effects were included to account for vessel specific differences in production. The environmental variable, z, was the aggregate A80 quota allocation (Quota Allocation).

Other variables were also included in the model but preliminary analysis indicated that some variables were not critical determinants of output and were excluded from the model. Specifically, initial inclusion of several A80 species-specific biomass estimates and found many to exhibit negative coefficient estimates that often produced counter-intuitive results (suggesting increasing biomass actually lowers catch, ceteris paribus). And, the species biomass variables with significant positive coefficient estimates were minor species that are not primary components of A80 vessel production. A variable constructed as the sum of the A80 quota allocations, was found it to be more consistent with the expected environmental (stock) effect than the species-specific biomass variables, and also a stronger determinant of changes in annual catch levels than an alternative biomass specification (a revenue weighted average of A80 species biomass estimates

B.E. Fissel et al. / Marine Policy I (l

Fig. 2. Average production output and input Lowe indices. (note this figure is two panels).

-B— A80 quota allocation A A80 catch 0 A80 blomasslndex Fig. 3. Amendment 80 catch, biomass, and quota allocation.

(Fig. 3)). The evidence from this analysis indicates that the A80 allocation performed better because the amount of the species actually available to the fleet is typically not determined by biomass, but rather by the size of the overall total allowable catch (TAC), which is often set well below the sum of the maximum allowable biological catch (ABC) of all species due to the 2 million ton cap on the total groundfish that can be taken annually in the Bering Sea.14 Catch of the key A80 species is also typically well below the ABC, which reduces the likelihood that resource scarcity as a result of abundance will be a major determinant of productivity. In aggregate, catch comports more with the quota allocation than aggregate changes in biomass.

Three models are presented in Table 2, which reports the results of estimating the stochastic frontier in Eq. (13).15 Model A was estimated with all variables included. Models B and C include different specifications of technical change to highlight different

14 The majority of the 2 million ton cap goes to the pollock fishery whose TAC has gone from 1.5 million tons in 2005/2006 to 0.81 million tons in 2009/2010 to 1.2 million tons in 2011/2012. The remainder of the 2 million tons gets split among all other groundfish and in some years there is much more room under the cap than in others. As the majority of A80 species TACs are well below their ABC, there is a lot of room to sustainably increase harvest of these species from a biological standpoint, but there are physical and technological constraints as well as bycatch constraints limiting the fleet's ability to harvest the entire quota they are allocated. Therefore, environmental conditions were modeled using the A80 species quota allocation as it is more closely tied to catch without necessarily moving in direct proportion with it.

15 The model was estimated using the 'frontier' package in R [42].

features of how it has changed over time. Model B removes the coop trend interaction terms to characterize the aggregate TFP trend. Model C highlights the post-rationalization trends attributable to co-op participation by removing the aggregate trend and co-op dummies and includes a pre-rationalization trend. Results indicate that labor, fuel and capital are all highly influential inputs determining catch (Table 2). Testing confirmed that the coefficients on the input factors are jointly statistically significant at the 1% level. A test for constant returns to scale rejected the null hypothesis that the sum of the input factor coefficients was equal to one at the 1% significance level. For all models the sum of the input factor coefficients was approximately 1.14, indicating that vessels are currently operating in a range of increasing returns to scale.16 Given that production functions can be susceptible to econometric endogeneity [35,36] Durbin-Wu-Hausman endogeneity test was conducted with output prices as instruments for input factors [37]. While output prices may be considered weak instruments, they were the best instruments available. The test failed to reject the null hypothesis that explanatory variables and errors are correlated (i.e., the presence of endogeneity was rejected), and hence, the use of the input factors is preferred.17

The time trends in the technical change specification were a prominent feature of the model. The relatively short three-year pre-rationalization sample likely confounded identification of distinct post-rationalization trends. Inclusion of the time trend shows that aggregate productivity has been increasing at a rate of roughly 10% per year (Models A and B). Tests of a break in the general trend post-rationalization were rejected.18 Relative to the general (increasing) trend in productivity, the Alaska Seafood Coop vessels were more productive, those in the post-rationalization limited access (Post-Rat LA) fishery were less productive, and

16 The time series length is too short to conduct a meaningful test of autocorrelation in the residuals. A test was indicative of the presence of hetero-skedasticity across vessels in the residuals, but not across input factors. There was no clear way to account for this given the software. The unaccounted for hetero-skedasticity could impact the efficiency of our estimators, but they are still unbiased and consistent, hence coefficient estimates (and indices) are unlikely to be substantially affected.

17 Other variables that could be argued to be exogenous instruments were also tried with the same result. These include the biomass index, quota allocation, and fuel prices. Because of the short time series the use of lagged output prices as instruments significantly reduced the number of observations; hence our final test results used only contemporaneous output prices.

18 Because all vessels were either limited access or in a Co-op, the post-rationalization dummy is equal to the sum of the limited access and Co-op dummies. The same relationship exists between the post-rat trend and the trend-Co-op interaction variables. Hence, in Model A not all could be included.

B.E. Fissel et al. /Marine Policy l (

Table 2

A80 productivity stochastic frontier model results.

Coefficient Model A Model B Model C

Intercept -0.015 - 0.851 - 5.749***

(0.998) (0.994) (0.979)

Trend 0.117 0.082***

(0.074) (0.014)

Pre-Rat Trend 0.110***

(0.021)

Post-Rat Lim. Acc. 0.215** 0.132

(0.090) (0.107)

AK GF Co-op 0.382 - 0.103

(0.862) (0.107)

AK SF Co-op 0.200** 0.188**

(0.102) (0.074)

Trend Post-Rat LA* - 0.080 0.081 ***

(0.106) (0.020)

Trend AK GF Co-op* -0.153 0.031 ***

(0.204) (0.010)

Trend AK SF Co-op* - 0.040 0.103***

(0.096) (0.006)

log(Quota Allocation) 0.156* 0.214*** 0.590***

(0.088) (0.081) (0.070)

log(Labor) 0.422*** 0.422*** 0.339***

(0.089) (0.114) (0.074)

log(Fuel) 0.446*** 0.406*** 0.492***

(0.104) (0.089) (0.072)

log(Capital) 0.277*** 0.315** 0.305***

(0.106) (0.124) (0.105)

a2 0.024*** 0.026*** 0.034***

(0.003) (0.004) (0.003)

Y 0.980*** 0.988*** 1.000***

(0.082) (0.141) (0.001)

Log likelihood value 157.85 153.41 155.167

Mean efficiency 0.889 0.885 0.874

BIC -131.23 -137.73 - 141.24

Total number of observations — 168

Standard Errors are in parentheses below the parameter estimates. The parameters a1 and y are related to the variance of the normal, vit, and half-normal, uit, error terms (a = aV + aU , 1 = au/av , aV = Var(vit) and aV = Var(vit)). The model was estimated with vessel fixed effects which have been excluded from this table.

* Significance codes=0.1.

** Significance codes=0.05.

*** Significance codes=0.01.

those that formed the Alaska Groundfish Co-op saw little if any change in productivity relative to the Post-Rat LA period. The decrease in relative productivity of the Alaska Groundfish Co-op in Model B, indicated by the negative coefficient, may at first appear

slightly puzzling. However, this is a result of the trend specification and the fact that the AK Groundfish Co-op was formed in 2011 largely from the vessels that had chosen to remain in the limited access fishery between 2008-2010. Model C shows more explicitly that the productivity increase in the Alaska Seafood Co-op has been more pronounced.

Productivity indices were constructed using Model A, which was selected because it was the most comprehensive. By examining the individual components of the Total Factor Productivity index (TFPI) in Fig. 4 one can see that the primary driver of productivity growth is the technical change component (TI). Environmental efficiency increased somewhat through most of the observation period due to the increases in quota allocations for target species between 2006 and 2010. Technical efficiency and scale-mix efficiency show minimal variation over the sample period. The differences between the Alaska Seafood Co-op vessels and those vessels in the Post-Rat LA fishery, which then formed the Alaska Groundfish Co-op (referred to as "PRLA+AKGC"), can be seen in the right panel of Fig. 4 (note that for the year 20052007 the indices in the Alaska Seafood Co-op plot are for those vessels that joined the co-op in 2008). The Alaska Seafood Co-op vessels experience a much larger increase in TFPI throughout the study period compared with the PRLA+AKGC vessels, which still experience a significant increase in TFPI after rationalization even while participating in the Post-Rat. LA fishery. These vessels experience a slight increase in TFPI in the first year of operating as a cooperative (2011), but these gains are erased by a decline in TFPI in 2012.

Table 3 provides the specific index values. Because of the small impact from the technical efficiency (OTEI) and scale-mix efficiency (OSMEI) and noise indices (NOISEI) these have been aggregated into a variable labeled OTHER.

7. Discussion and conclusion

There are several mechanisms in the A80 Program that may help provide context for the observed increases in total factor productivity growth, and in particular, technical change that shifted out the production frontier. First, allocating catch shares to vessels has eliminated the race for fish, and could have subsequently reduced fishing costs, particularly fuel use which dropped 15% from the first year after implementation, 2008, through 2012. Second, the cooperative structure of the Amendment 80 Program could be improving communication among cooperative members and reducing search costs. Third, A80 changed the way in which halibut PSC allocations were made which has likely had a significant impact on each member vessel's ability to increase their catch due to the changes in the way PSC are allocated and utilized within the cooperative [2]. Prior to the A80 program, halibut PSC limits would close the entire target fishery for rock sole and yel-lowfin sole prior to full exploitation of the TAC. Because target fisheries are no longer closing because of halibut PSC limits, vessels have been able to increase the number of active days in the fishery from an average of 258 days during the three years prior to A80 (2005-2007) to an average of 322 days over the period 20082012. This added flexibility, in addition to increases in quota allocated to the sector, has allowed the sector to increase their average catch from 200,346 t from 2005-2007 to 241,087 t from 2008-2012 [38, Section 7.5]. While this obviously requires greater use of fuel, labor, and capital (which are accounted for in the indices), the gains in the environmental efficiency index discussed in the earlier section highlight the contribution of the quantity and nature of quota allocations on productive performance.

Indices did not display sizable changes in technical efficiency within this fleet after passage of Amendment 80. This suggests

B.E. Eissel et al. / Marine Policy I (l

Table 3

Productivity indices (Model A).

Year Group TFPI TI ZI OTHER

All Split All Split All Split All Split

2005 AKSC LA (PRLA+AKGC) LA 0.88 0.91 0.81 0.89 0.89 0.89 0.99 0.99 0.99 1.00 1.01 0.97

2006 AKSC LA (PRLA+AKGC) LA 0.98 1.01 0.91 1.00 1.00 1.00 0.99 0.99 0.99 1.00 1.01 0.97

2007 AKSC LA (PRLA+AKGC) LA 1.08 1.11 1.01 1.12 1.12 1.12 1.02 1.02 1.02 0.98 0.99 0.96

2008 AKSC (PRLA+AKGC) 1.42 1.46 1.33 1.47 1.48 1.49 1.05 1.05 1.05 0.97 0.98 0.96

2009 AKSC (PRLA+AKGC) 1.53 1.61 1.35 1.56 1.59 1.49 1.06 1.06 1.06 0.97 0.98 0.95

2010 AKSC (PRLA+AKGC) 1.66 1.56 1.35 1.67 1.72 1.55 1.07 1.07 1.07 0.98 1.00 0.94

2011 AKSC (PRLA+AKGC) 1.74 1.60 1.43 1.76 1.86 1.42 1.04 1.04 1.04 0.98 0.98 0.99

2012 AKSC (PRLA+AKGC) 1.90 1.82 1.38 1.85 2.01 1.37 1.03 1.03 1.03 1.00 1.00 0.99

OTHER is the geometric mean of the OTEI, OSMEI, and NOISEI indices.

that the relative abilities of vessels to operate on the frontier did not change over time; the frontier did shift out in each period after the introduction of catch shares, as evidenced by the positive technical change, but on average the fleet's relative ability to operate on that frontier did not. This finding seems consistent with a fleet whose composition did not change much during the period under examination, and so the heterogeneity and relative performance of the vessels in the fleet remained stable after accounting for the environmental and input use changes. In addition, indices did not display notable changes in scale-mix efficiency, which one can see relates to changes in the relative proportion of factor inputs in Eq. (14). Given the technology used to catch and process fish with trawl gear, this result is not surprising. There is not a great degree of flexibility in substituting labor for fuel or capital when trawling for flatfish. One could argue that a given vessel has a relatively rigid production technology that requires a specified number of bodies to haul gear, cut fish, and box and freeze finished product. Several authors have utilized a uni-dimensional measure of "effort" in characterizing fishing technologies, arguing that they are essentially fixed-proportions technologies [34,39-41]. Discussions with industry personnel indicate that there has been no large-scale change in the way these inputs are combined to catch and process fish. The behavioral changes pertain mostly to the timing and spatial location of fishing, largely in the pursuit of diminished PSC catch.

In addition to observing an aggregate, fleet-wide increase in total factor productivity after A80 was implemented, differential productivity changes were observed according to whether vessels participated in a cooperative, and according to the cooperative in which they chose to participate. Specifically, vessels participating

B.E. Fissel et al. /Marine Policy l (l

in the Alaska Seafood Co-op exhibited a higher productivity increase relative to the Post-rationalization LA fishery, and that productivity gains did not change when vessels moved from the Post-rationalization LA fishery to the Alaska Groundfish Co-op. This latter result may be attributable to this change essentially being a title change for a group of vessels rather changes in the composition and prosecution of the fishery.

In sum, these findings lend support to the notion that the Amendment 80 program had positive economic effects on the productivity of the fleet. Increases in profitability have occurred over time even after accounting for changes in output prices, input use and costs, and quota allocations. These results indicate that increased flexibility in terms of the timing of fishing and utilization of PSC, as well as improved coordination and communication through the development of cooperatives, has helped the fleet operate more productively relative to the period prior to Amendment 80.

Acknowledgments

We are grateful to the U.S. National Marine Fisheries Service Office of Science and Technology and Rita Curtis for providing financial support for this project. We are also grateful to the participants at the 2104 International Institute of Fisheries Economics and Trade Conference, the 2014 Asia-Pacific Productivity Conference, Dale Squires, Eric Thunberg and referees for insightful comments and advice. We remain responsible for any errors. The results are not necessarily those of the U.S. National Marine Fisheries Service.

References

[1] E. Grifell-Tatjé, C.A. KnoxLovell, Productivity Accounting: The Economics of Business Performance, Oxford University Press, New York, 2015.

[2] J.K. Abbott, A.C. Haynie, M.N. Reimer, Hidden flexibility: institutions, incentives, and the margins of selectivity in fishing, Land Econ. 91 (2015) 169-195.

[3] North Pacific Fishery Management Council and National Marine Fisheries Service, Final Environmental Assessment/Regulatory Impact Review/Final Regulatory Flexibility Analysis for Amendment 80 to the Fishery Management Plan for Groundfish of the Bering Sea and Aleutian Islands Management Area. Available at: <http://alaskafisheries.noaa.gov/sustainablefisheries/amds/80/ earirfrfa0907.pdf), 2007.

[4] Northern Economics, Inc., Five-Year Review of the Effects of Amendment 80. Prepared for North Pacific Fishery Management Council, October, 2014.

[5] J. Walden, J. Agar, R. Felthoven, A. Harley, S. Kasperski,J. Lee, T. Lee, A. Mamula, J. Stephen, A. Strelcheck, E. Thunberg, Productivity Change in U.S. Catch Shares Fisheries. U.S. Dept. of Commer, NOAA Technical Memorandum NMFS-F/SPO 146 (2014) 137p.

[6] S. Comitini, D.S. Huang, A study of production and factor shares in the Halibut fishing industry, J. Polit. Econ. 75 (4) (1967) 366-372.

[7] T.M. Jones, J.W. Hubbard, K.J. Roberts, Productivity and Profitability of South Carolina Shrimp Vessels, 1971-75, Mar. Fish. Rev. 41 (1979) 8-14.

[8] V.J. Norton, M.M. Miller, E. Kenney, Indexing the Economic Health of the U.S. Fishing Industry's Harvesting Sector. NOAA Technical Memorandum NMFS-F/ NEC-40. U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Northeast Fisheries Science Center. Woods Hole, Massachusetts, 1985.

[9] D. Squires, Index Numbers and Productivity Measurement in Multispecies Fisheries: An application to the Pacific Coast Trawl Fleet. NOAA Technical Report NMFS 67. U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Southwest Fisheries Science Center, La Jolla, CA, 1988.

[10] D. Squires, Productivity measurement in common property resource industries: an application to the pacific coast trawl fishery, RAND J. Econ. 23 (2) (1992) 221-236.

[11] D. Squires, Sources of growth in marine fishing industries, Mar. Policy 18 (1) (1994) 5-18.

[12] Q. Weninger, An analysis of the efficient production frontier in the fishery: implications for enhanced fisheries management, Appl. Econ. 33 (2001) 71-79.

[13] Jin Di, E. Thunberg, H. Kite-Powell, K. Blake, Total factor productivity change in the new england groundfish fishery: 1964-1993, J. Environ. Econ. Manag. 44 (3) (2002) 540-556.

[14] R.G. Felthoven, C.J.M. Paul, Directions for productivity measurement in fisheries, Mar. Policy 28 (2) (2004) 161-169.

[15] K.J. Fox, R.Q. Grafton, T. Kompas, T.N. Che, Capacity reduction, quota trading and productivity: the case of a fishery, Aust. J. Agric. Resour. Econ. 50 (2)

(2006) 189-206.

[16] R. Hannesson, Growth accounting in a fishery, J. Environ. Econ. Manag. 53 (3)

(2007) 364-376.

[17] D. Squires, C. Reid, Y. Jeon, Productivity growth in natural resource industries and the environment: an application to the Korean Tuna Purse-Seine fleet in the Pacific Ocean, Int. Econ. J. 22 (1) (2008) 81-93.

[18] R.G. Felthoven, C.J.M. Paul, M. Torres, Measuring productivity and its components for fisheries: the case of the Alaskan Pollock Fishery, 1994-2003, Nat. Resour. Model. 22 (1) (2009) 105-136.

[19] C.J.M. Paul, M. Torres, R. Felthoven, "Fishing revenue, productivity and product choice in the Alaskan Pollock fishery, Environ. Resour. Econ. 44 (2009) 457-474.

[20] J.B. Walden, J.E. Kirkley, R. Fare, P. Logan, Productivity change under an individual transferable quota management system, Am. J. Agric. Econ. 94 (4) (2012) 913-928.

[21] H. Eggert, R. Tveteras, Productivity development in Icelandic, Norwegian and Swedish fisheries, Appl. Econ. 45 (6) (2013) 709-720.

[22] M. Torres, R. Felthoven, Productivity growth and product choice in catch share fisheries: the case of the Alaska Pollock, Mar. Policy 50 (2014) 280-289.

[23] J. Lowe, The Present State of England in Regard to Agriculture, Trade and Finance, 2nd ed., Longman, Hurst, Rees, Orme and Brown, London, 1823.

[24] C. O'Donnell, Measuring and decomposing agricultural productivity and profitability change, Aust. J. Agric. Resour. Econ. 54 (2010) 527-560.

[25] C. O'Donnell, Nonparametric estimates of the components of productivity and profitability change in U.S. agriculture, Am. J. Agric. Econ. 94 (2012) 873-890.

[26] D. Witherell, M. Fey, M. Fina, Fishing Fleet Profiles. North Pacific Fishery Management Council. Anchorage, AK. Available at: <http://www.npfmc.org/ wp-content/PDFdocuments/resources/FleetProfiles412.pdf>, 2012.

[27] C. O'Donnell, An aggregate quantity framework for measuring and decomposing productivity change, J. Prod. Anal. 38 (2012) 255-272.

[28] C. O'Donnell, Alternative Indexes for Multiple Comparisons of Quantities and Prices. Centre for Efficiency and Productivity Analysis Working Papers No. WP05/2012 (Version 21 May 2013), University of Queensland, 2012.

[29] C. O'Donnell, Econometric estimation of distance functions and associated measures of productivity and efficiency change, J. Prod. Anal. 41 (2014) 187-200.

[30] C. O'Donnell, Using Information About Technologies, Markets and Firm Behaviour to Decompose a Proper Productivity Index, Journal of Econometrics, forthcoming.

[31] R. Fare, D. Primont, Multi-output Production and Duality: Theory and Applications, Kluwer Academic Publishers, Boston, 1995.

[32] G. Debreu, The Coefficient of Resource Utilization, Econometrica 19 (3) (1951) 273-292.

[33] M. Farrell, The measurement of productive efficiency, J. R. Stat. Soc., Ser. A (Gen.) 120 (1957) 253-290.

[34] D. Squires, J. Kirkley, Production quota in multiproduct pacific fisheries, J. Environ. Econ. Manag. 21 (1991) 109-126.

[35] S. Olley, A. Pakes, The dynamics of productivity in the telecommunications equipment industry, Econometrica 64 (6) (1996) 143-205.

[36] J. Levinsohn, A. Petrin, Estimating production functions using inputs to control for unobservables, Rev. Econ. Stud. 70 (2) (2003) 317-341.

[37] J.A. Hausman, Specification tests in econometrics, Econometrica 46 (1978) 1251-1271.

[38] B. Fissel, M. Dalton, R. Felthoven, B. Garber-Yonts, A. Haynie, A. Himes-Cornell, S. Kasperski, J. Lee, D. Lew, C. Seung, Stock Assessment and Fishery Evaluation Report for the Groundfish Fisheries of the Gulf of Alaska and Bering Sea/ Aleutian Islands Area: Economic Status of the Groundfish Fisheries off Alaska, 2013. Available at: <http://www.afsc.noaa.gov/REFM/Docs/2014/economic. pdf>, 2014.

[39] J. Kirkley, The Relationship between Management and the Technology in a Multi-species Fishery: The New England, Georges Bank, Multi-species Fishery (Ph.D. dissertation), University of Maryland, 1986.

[40] J. Kirkley, I. Strand, The technology and management of multispecies fisheries, Appl. Econ. 20 (1988) 1279-1302.

[41] K.G. Salvanes, F. Steen, Testing for relative performance between seasons in a fishery, Land Econ. 70 (4) (1994) 431-447.

[42] T. Coelli, A. Henningsen, "frontier: Stochastic Frontier Analysis." R package version 1.1-0. <http://CRAN.R-Project.org/package=frontier>, 2013.