Multiple Margins of Fishing Behavior: Implications for Predicting the Effects of a Policy Change
Matthew Reimer University of Alaska Anchorage Institute of Social and Economic Research
Multiple Margins of Fishing Behavior • Fisheries management history: reactive policymaking - fishermen have far more flexibility than originally thought - new policies react to unanticipated consequences of previous policies • Example: BC Salmon Limited Entry Program 1969 •
•
Rapid erosion of effort controls as fishermen expanded into “free” effort dimensions.
Initial limits on number of vessels ➡ Limit on tonnage of vessels ➡ Limit on length of vessel ➡ Limit on gear types ➡ Limit on combining licenses
Ignoring behavioral margins = regulatory surprises • Traditional models of the fishing production process ignore the primary behavioral margins of fishermen !
- Early models: aggregate production functions relating industry catch to industry fishing effort !
- Powerful insights: e.g. open access = biological overexploitation + dissipation of rents (Gordon, 1954) !
- Interpretation: “too many boats chasing too few fish.” Fix incentives along this margin, and fisheries problem solved! !
- Accumulation of experience shows that other margins are likely to matter. !
“The New Fisheries Economics: Incentives Across Many Margins” M.D. Smith (2012)
• Multiple margins (extensive and intensive) across which fishermen act - amount of gear - fishing grounds - type of gear - product types - number of trips - fish size - trip length - entry or exit - target species - different fisheries !
• To what extent should managers control these margins? - How can managers control these margins? - What are the consequences of ignoring them? - How do we know ex ante what margins are important?
Fishery Production Models and Policy Invariance CONCEPTS
possibility curve could be drawn for each input level. Furthermore, w the combination of outputs that maximise profit, given an input level to that which maximises revenue.^ The revenue equivalent to the iso isorevenue line, which has slope equal to {-pxlpi), the negative rati prices. The optimal (revenue-maximising) point is determined b tangency between this line and the production possibility curve, Figure 3.2.
• Conventional aggregate fishery production models that ignore primary behavioral margins do not identify policy invariant parameters y2
92
• Catchability parameters and selectivity curves convey biological, technical, and behavioral relationships
Policy Invariant?
Pppc (x)ixrxio) ~ ~ " ~ ~ " ^ ^ / ^
0
y1 qi
The Importance of Policy Invariance: Lucas Critique (1976) • It is naïve to try to predict the outcome of a policy intervention entirely on the basis of a relationship that systematically alters with a change in policy.
• “Policy invariance facilitates the job of forecasting the impacts of interventions. If some parameters are invariant to policy changes, they can be safely transported to different policy environments.” (Heckman, 2010)
What are the implications of ignoring key behavioral margins when predicting the effects of a policy? Outline • Traditional approaches to modeling fishery production - How have these been used to predict policy interventions?
• Simulation exercise - How does a policy intervention change production relationships when accounting for behavioral margins?
• Empirical investigation of the Bering Sea groundfish fleet - Are empirical aggregate production relationships invariant to a policy change that changed fishermen incentives?
• Current and future directions - Implications for policy evaluation: where do we go from here?
A Pressing Question: Are Catch Shares Appropriate for Multispecies Fisheries? • Catch shares—a secure privilege to harvest a proportion of a fishery’s total allowable catch. - can be allocated for multiple species - can be allocated to individuals, groups, or communities - often seen as “the way” to end the “race-for-fish” under open access institutions
A Pressing Question: Are Catch Shares Appropriate for Multispecies Fisheries? • Can fishermen match their catch composition with a portfolio of quota allocations? - fishing gear is not perfectly selective (Squires et al. 1987) - could encourage illegal discarding and data fouling (Copes, 1986) - “choke” species and unharvested quota - selectivity depends on targeting ability (Pascoe et al. 2007)
• Targeting ability—or output substitution capabilities—can be represented via a multi-output production function - curvature of the production frontier indicates a fisherman’s “ability” to substitute between species
Output Substitutability with “Bad” Outputsb = und 0 ARTICLE IN PRESS
2. Directionalpossibility output distance function with desirable and u Figure 4.1:Fig. Production set example - Example R. Fa¨re et al. / Journaloutputs. of Econometrics 126 (2005) 469–492 Adapted from F¨are et al. (2005). y = desirable output
g = (gy, -gb)
This idea is made for output the general case of byproducts in Murty b = clearly undesirable null jointness and weak disposability assumptions are a reduced form wa This function seeks through theoutputs. simultaneous maximum reducti Directional output distance function with and desirable undesirable between good badand outputs abatement e↵orts that are not exp 0
Fig. 2.
(q/p) researchers have modeled environmental out well. Some outputs areofeffectively treated as isinputs with to similar strong Satisfaction conditions (i)-(vi) sufficient express the which would yield an unbounded output set; for examples the use of aand transformation function T (McFadden, 1978): outp Hailu Veeman (2001). However, an unbounded Output Set : possible if traditional inputs are given. (y + Although β*gy, b - β*gb) our set representations P (x) = {(y,ofb)the : Ttechnology (x, y, b) are 0} ,c are not very helpful from a computational viewpoint. He representations of technology, which allow us to mainta where T (x, y, b) = 0 represents the production possibilities fron accommodate thePWproduction of byproducts, and are consis (x) Our preferred modeloutput is the that directional output set ofapproach. all combinations of “efficient” can be obtained g¼ 3 ðgy ; gb Þ be a directional vector (illustrated below), puts directional x. The production transformation function T (·) provides a F = (b, y) distance function is defined as relation that represents the multidimensional set of feasible o ~o ðx; y; b; gy ; $gb Þ ¼ maxfb : ðy þ bgy ; b $ bgb Þ 2 PðxÞ D 2
3
expansion in good outputs. Fig. 2 illustrates the directional o
A Pressing Question: Are Catch Shares Appropriate for Multispecies Fisheries? • Data from non-catch shares fisheries suggest catch shares face serious challenges due to weak substitution possibilities between species - e.g. Squires (1987); Squires and Kirkley (1991,1995,1996); Pascoe et al. (2007,2010)
• Evidence from multispecies catch shares shows greater flexibility than previously thought - e.g. Branch and Hilborn (2008); Sanchirico et al. (2006)
• Perhaps output substitution revealed through ex ante empirical investigations reveal more about behavior and incentives than actual technical relationships.
Neglected Margins of Production
Fishing production depends on temporal and spatial choices.....
Multispecies production function: microfoundations local production y = y1 y2 function:
x y2
global production function
10000
y
x2 x3
4000
x4
6000
v2
x1
✓1jt ✓2jt
8000
X
=
y1 0
2000
4000
6000
8000
10000
Simulation Model T wo Species : bycatch(1) and target(2)
(a) bycatch(1)
(a)
(b)
25
25
25 10
25
25
3000
3000
3000
10
30
1600
9
9
2700
2700
2700
27 1400
8
8
20
2400
(b)
7 6
2100
5
3000 1800
15 4 3
20
20
target(2)
(c)
bycatch(1) 20
2400
1200 1 1
2
3
4
5
6
7
8
9
2400 900
10
600
5
2100
15
15
15
1800
1800
6
21
160015
10
1200
10
10
15
20
25
0 1800
10
1200
140010
20 900
5
600
5
300
15 5
10
15 5
2010
2515
20
25
5
0
1
25
0
5 200
5
10
15
5
20
25
6
7
8
9
10
90
400
60
1000 10
600
300
20
4
200
400
15
3
5
900 600
2
1200
800 10
12
600
300
0
15
2
900
600
5
18
800
1
1500 1200
1000 4 3
1500
1500
300
5
1200
5
1500
2
24
7
2100
2100
25
20
2400
2700
10
10
(b) target(2)
(c)
0
15
5
20
10
25
0 15
30
20
25
0
Simulation Model T wo Species : bycatch(1) and target(2)
(a) bycatch(1)
(a)
(b)
25
25
25 10
(b) target(2)
(c) 25
25
3000
3000
3000
10
30
1600
9
9
2700
2700
2700
27 1400
8
8
20
6
20
20
2400
7
20
2400
2100
20
2400
1200
6
2100
2100
24
7
21 5
15
1800
4 3
30
5
15
15
15
1800
1500
18
4 3
1500
1500
1000
15
1800
15
800
2
2 1 1
2
3
4
5
6
7
8
9
10
10
25
1200
10
1200
10
10
1200
10
900
900
1
900
12
600
1 2
3
4
5
6
7
8
9
10
90
400 600
5
Target Species 15 20
5
600
5
300
15
20
25
0
5
10
15 5
2010
2515
200
300
0
20
25
5
0
10
10
10
300
0
5
10 Bycatch Species
60
5
5
5
600
5
15
20
15
5
20
10
25
0 15
30
20
25
0
Simulation Model At each t :
max {E[y2tj ]
j2{1,...J}
⇢E[y1tj ]
rDt (j)}
Dt (j) = distance to j f rom current location
⇢ and r are institutional parameters ⇢ = shadow cost of bycatch
low ⇢ means fishermen do not internalize the external cost of bycatch.
starting location
Simulation Model At each t :
max {E[y2tj ]
j2{1,...J}
⇢E[y1tj ]
rDt (j)}
Dt (j) = distance to j f rom current location
⇢ and r are institutional parameters r = shadow cost of travel
high r means fishermen perceive an opportunity cost of not fishing.
starting location
Simulation Model At each t :
max {E[y2tj ]
j2{1,...J}
⇢E[y1tj ]
rDt (j)}
Example high ⇢ and low r represents a fishery with catch shares. low ⇢ and high r represents a fishery without catch shares.
starting location
High ρ Low
r
25
0.009 0.008
20
0.007 0.006
15
r
Low ρ Low
Revealed Production Sets
0.01
0.01
25
0.009
0.008
20
0.007
0.006
15
0.005
0.005 0.004
10
0.004
10
0.003
0.003 0.002
10
15
20
25
0.002
5
0.001
0.001 0
5
10
15
20
25
0
10
Target Species 15 20
25
5
30
5
Low ρ High r
5
High ρ High r
0.02
25
0.018 0.016
20
0
5
10 Bycatch Species
15
20
0.02
25
0.018
0.016
20
0.014
0.014 0.012
15
0.012
15
0.01
0.01 0.008
10
0.008
10
0.006
0.006 0.004
5
0.004
5
0.002
0.002
5
10
15
20
25
0
5
10
15
20
25
0
High ρ Low
r
25
0.009 0.008
20
0.007 0.006
15
r
Low ρ Low
Revealed Production Frontiers
0.01
0.01
25
0.009
0.008
20
0.007
0.006
15
0.005
0.005 0.004
10
0.004
10
0.003
0.003 0.002
10
15
20
25
0.002
5
0.001
0.001 0
5
10
15
20
25
0
10
Target Species 15 20
25
5
30
5
Low ρ High r
5
High ρ High r
0.02
25
0.018 0.016
20
0
5
10 Bycatch Species
15
20
0.02
25
0.018
0.016
20
0.014
0.014 0.012
15
0.012
15
0.01
0.01 0.008
10
0.008
10
0.006
0.006 0.004
5
0.004
5
0.002
0.002
5
10
15
20
25
0
5
10
15
20
25
0
Application: The Bering Sea Groundfish Fishery Did rights-based management induce bycatch avoidance?
brokers or wholesalers for direct sale or further processing. Since the early 2000s, twenty-
three vessels have actively participated in the fishery, ranging in size from 91 to 295 feet
Bering Sea and Aleutian Islands (median=154) with horsepower ranging from 850 to 7000 (median=2250).
²
²
ALASKA
ALASKA Red King Crab Savings Area (RKCSA) Pribilof Islands Habitat Conservation Area
Bering Sea
Aleutian Islands
Red King Crab Savings Subarea (RKCSS)
0
35 70
140
210
280
Nautical Miles 350
0 12.5 25
50
75
100
Nautical Miles 125
Figure 5.1: Map of Bering Sea and Aleutian Islands -
Generally speaking, the BSAI groundfish fishery can be divided into two fisheries: the 1
“Groundfish” refers to any fish species that live on or near the bottom of the seafloor.
The Bering Sea Groundfish Fishery • Pre-Amendment 80 (prior to 2008): - Target species TACs allocated as common property over multiple “sub-seasons” - TAC for prohibited species allocated to target species fisheries - Target fisheries typically closed due to binding bycatch TAC - particularly true for halibut - Fishermen “unable” to avoid halibut
• Post-Amendment 80 (2008 and after): - Target species and bycatch allocations vested directly into cooperatives or limited access fishery - Initially one cooperative formed: 16 vessels, 7 companies
Changes in Fishing Practices
• Large-scale shift in effort away from halibut-rich areas !
- Dark = increased effort - Light = decreased effort
January to April
-.2
Chg. in
Changes in Fishing Practices 2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
.4 .2 0
!
-.2
Chg. in Probability (2007=0)
Halibut>10%
• Fine-scale shift in effort after hauls with a large proportion of halibut
2002 2003 2004 2005 2006 2007 2008 2009 2010
Year
Move if dist>=5
- Probability of moving fishing locations after a large halibut encounter (relative to 2007)
Move if dist>=7
movements of a given minimum distance relative to
effects. Z-statistics (in parentheses) are based on cluster-robust standard errors clustered on combinations
Changes in Fishing Practices of vessel, year and week (* p