foModeling Fisher Behavior under Changing Policies ......• Timely analysis • Test if new models...
Transcript of foModeling Fisher Behavior under Changing Policies ......• Timely analysis • Test if new models...
foModeling Fisher Behavior under Changing Policies, Economics, and Environmental Conditions:
Spatial Models & FishSET
Alan Haynie, Alaska Fisheries Science Center, Seattle, WA
AFSC Socioeconomic Program Review, Wednesday, July 19, 2017
Overview• FishSET – the Spatial Economics Toolbox for Fisheries
• Amendment 80 multispecies fisheries impacts• Unintended consequences - Red King Crab Savings Area• Amendment 80 & halibut bycatch reduction• Steller Seal Lion protected area impacts
• Chinook bycatch: from spatial management to incentives
• Elephants! U.S. Department of Commerce | National Oceanic and Atmospheric Administration | NOAA Fisheries | Page 2
What is FishSET? Key Goals and Elements
1. Better integrate data
2. Develop best modeling practices
3. Develop a Toolbox of many models• Timely analysis• Test if new models are better
Page 3
Data to explain the factors that impact fishing
• Spatial fishing information • Vessel characteristics• Price Info
• From markets• From vessel surveys
• Biological survey info
• Environmental data• Satellite observations• Weather station data• Buoy data• Bathymetry; Ice data;
ROMS; Habitat• Other
Photo:CSMphotos
Page 6
A Big Data Focus: Identifying Fishing from Vessel Monitoring System (VMS) Data
Models allow us to use VMS data to estimate where unobserved fishing is occurring.
Source: Watson & Haynie (unpublished)
Size-targeting and production strategies in the Bering Sea pollock catcher processor fishery
Chen & Haynie(in prep)
Page 8
Pilot Projects across the USABureau of Ocean Energy
Management Project Wind Energy Site Evaluation
Turtle Protection Measures
West Coast Rockfish Conservation Areas
Steller Sea Lions
Hawaii Longline Fishery
Page 9
• Expected catch/revenue• Travel costs• Vessels size & character• Biological & environmental
characteristics of areas• Regulations, quotas, and bycatch
How do fishers choose where to fish?
Example types of fisher location choice models • Area-specific constant logit (ASC logit)
• Nested logit model• Participation model• Models with bioeconomic linkages
• Mixed logit model(s)• Plus: a variety of specialized fisher location choice
models (E.g., EPM, DRUM, etc.)
Page 11
FishSET Tools
Page 12
Data management and integration tool
Page 13
Mapping Tools
Equation Terms Settings
Prediction Catch (C)
Prediction Variable Catch Var
Individual (i) Vessel
Group (g) Company
Fleet (f) auto set All Vessels
Time Variable Haul Date
Recent (r) 1 day
Short Term (s) 14 days
Long Term (l) all time
Individual’s Zone (z) auto set
Area (a) area variable
All zones on map (m) auto set
C = Ci, r, z+ Ci, s, z + Ci, l, z+ Cg, s,a+ Cf,l,m + E
Menu…
C
Ci, r, z Ci, s, z Ci, l, z Cg, s,a Cf,l,m
Linear Fit Results
Rsqr 0.9999
β0 0
βi, r, z
βi, s, z
β, l, z
βg, s,a
βf,l,m
Flexible tools to form expectations
Page 15
Modeling Tools
Model Comparison & Reporting Tool
• Pseudo-R-squared
• AIC, AICc, BIC , etc.
• In-sample and out-of-sample predictive skill
Page 16
Reporting tool
Page 17
Page 18
Allows analysts and policy-makers to compare the impacts of potential policies
Policy Simulation Tool
Help & teaching tools …
Page 19
Next Steps
• Complete additional models and distribute Matlab Version of FishSET
• Develop R Verison• Link with other related projects like POSEIDON• Develop standing models of key fisheries.
Page 20
Next: 2 stories are about the BSAI multispecies catcher-processor trawl fishery
Fish Story # 1
“What Are We Protecting? The Challenges of Marine Protected Areas for Multispecies Fisheries”
Josh Abbott and Alan Haynie, 2012.
Ecological Applications, 22(3): 762–777.
Photo: M. ReimerPhoto: M. Reimer
Case Study: The Eastern Bering Sea Rock Sole/ Pacific Cod Fishery
• Group (<20) of trawl vessels • January to March
• Until 2008 faced common-pool TACs on target and prohibited species catch (PSC)• Red king crab• Pacific halibut
• Before 1995 when Red King Crab Savings Area was implemented, red king crab bycatch typically closed fishery
Photo: M. Reimer
Q1: How did the distribution of fishing effort change in the wake of the closures?
Q3: How did the closures impact red king crab bycatch?
Q3: How did halibut bycatch change as a result of displacement from the closure?
Fish Story # 2
Changes with Amendment 80 to the BSAI Fishery Management Plan
“Hidden Flexibility: Institutions, Incentives and the Margins of Selectivity in Fishing.”
Abbott, Haynie, and Reimer. Land Economics, February 2015.
Amendment 80 (A80) in 2008Goals:• Catch, profits, flexibility• Bycatch & discards
Vessels could:1) join a cooperative (16 vessels) OR2) stay in limited-access (6 vessels)
Also in 2008, Pacific cod allocation decreased.
Photo: Groundfish Forum
How did vessels reduce their bycatch?“Multiple margins”
1. Large-scale choice of fishing ground 2. “Reactive” spatial avoidance3. Reductions in night fishing
• a decrease of 15 to 18% relative to 2007; pronounced seasonality in the reduction.
Excluders, too, but we don’t have data on these.
Large-scale spatial avoidance: Sep - Dec
• No discernable-large scale pattern of avoidance
• Consistent with a late-season relaxation of avoidance efforts after uncertainty over multi-species quota scarcity is resolved
Industry response to late season halibut bycatch: Amendment 80 Halibut Bycatch Avoidance Plan
• Developed by industry in response to Council request, implemented by industry; reported to the Council in December 2016.
• Would impose large fines on high-bycatch outlier vessels based on seasonal performance
• Implemented through contractual agreement– no fines were imposed because all vessels were close to average rates and rates declined.
The Economic Impacts of Spatial Closures: Evidence from the 2011-2014 Steller Sea Lion Protective Measures in the North Pacific
Matthew N. Reimer1 and Alan C. Haynie21Institute of Social and Economic Research, University of Alaska Anchorage 2Alaska Fisheries Science Center, NOAA
Motivation: Spatial closures are a prominent tool forecosystem-based management in commercial fisheries, butempirical estimates of the short-run costs incurred by thecommercial fishing industry, such as increased costs and/orforegone revenue, are relatively scarce.
Project goal: We conduct an ex post evaluation of the short-run economic impacts associated with the 2011-2014 spatialclosures to the North Pacific groundfish fleet for the protection ofthe endangered western stock of Steller sea lions (SSL) in theAleutian Islands (AI).
Background
Introduction
In 2011, NMFS prohibitedthe retention of Atkamackerel and Pacific cod inmanagement area 543 byboth trawl and non-trawlgear, as well as in SSLcritical habitat areas by trawlgear in management area542 (Figure 1), resulting in asharp decline in the totalallowable catch (TAC) forAtka mackerel (Figure 2).
Treated vs. Control Groups: Of the 18 vessels in oursample, only 7 actively targeted Atka mackerel and Pacific cod inthe closed areas prior to the closures. These 7 vessels serve as thetreated group. The other 11 vessels serve as the control group,which we use to estimate what net revenue would have been in theabsence of the SSL protective measures.
Figure 1: National Marine Fisheries Service (NMFS) management areas for theBering Sea and Aleutian Islands and the critical habitat areas (shaded circles).Courtesy of Hicks and Schnier (2010).
Figure 2: Allowable biological catch(ABC) and total allowable catch (TAC) forAtka mackerel, by NMFS managementarea (541, 542, 543) and the entireAleutian Islands management area(Total).
DataOur main variable is annual net revenue as reported in theEconomic Data Reports for each vessel in the Amendment 80 (A80)fleet. We compute net revenue as annual gross revenue (e.g.,product sales and income from other sources) minus annualvariable cost (e.g., labor, fuel, maintenance and repair, food,packaging, and fish taxes).
The A80 fleet consists of catcher processors that use non-pelagictrawl gear to target non-pollock groundfish species. A80 vesselsreceive a share of the TAC for six primary species and vest them ina cooperative. Our sample consists of 18 vessels that fished in allyears during the period 2008-2014.
MethodsWe conduct a comparative case study, which estimates theevolution of an outcome variable for units affected by a particularintervention and compare it to the evolution of the same outcomeestimated for a control group of unaffected units.
We Estimate Three Model Types & Compare:Difference-in-Differences (DnD): Uses the entire control group as acomparison group. Allows treated and control vessels to be differentin ways that do not change over time. It requires trends in theoutcome variable to be the same for the treated and control groupsin the absence of the treatment (Figure 3).
Propensity-score-weighted (PSW) DnD: Matches treated vessels tosimilar control vessels based on pre-intervention baselinecharacteristics. Creates a comparison group that is more similar tothe treated group than in the DnD.
Synthetic Control Method (SCM): Constructs a “synthetic”comparison vessel for each treated vessel using a weightedaverage of vessels in the control group based on pre-interventioncharacteristics. Creates a comparison vessel that is more similar toeach treated vessel compared to the DnD and PSW.
Mechanisms: The estimators above are not able to explainHOW an effect on net revenue may arise. We estimate a StructuralEquation Model (SEM) in order to decompose the total effect on netrevenue into its constituent mechanisms (Figure 4). We focus onthe following mechanisms:• Quota mechanism: affects net revenue by reducing the harvest
quotas in the closed areas (Figure 1).• Closure mechanism: affects net revenue by displacing vessels
from historically productive fishing grounds.• Value mechanism: affects net revenue as vessels adjust
productive activities to increase the value of their production.
Figure 4: Causal diagram of themechanisms through which SSL measurecauses net revenue. The directed edge xysignifies that x causes y.
Figure 3: Average net revenue forthe treated and control vessels.
Results
-4
-2
0
2
4
6
2011
2012
2013
2014
2011
2012
2013
2014
2011
2012
2013
2014
2011
2012
2013
2014
2011
2012
2013
2014
Quota(mt thousand)
Harvest(mt thousand)
Price($/100 mt)
Revenue($ million)
Cost($ million)
p > 0.1 0.1 ≥ p > 0.05 0.05 ≥ p > 0.01 0.01 ≥ p
Overall Effect: Allestimators find a sizeablenegative effect on net revenue inthe first two years after theintervention (~30% reduction),approach zero by the third year,and become slightly positive by2014 (Figure 5). However,average effects are notsignificantly different from zeroat the 10% level for any year.
Figure 5: Point estimates of the annualaverage effect of SSL measure on netrevenue for the DnD, PSW, and SCMmodels. Intermediary Effects:
The SSL protective measures hada direct adverse (and significant)effect on variable cost, as well asthe quota received by vessels(Figure 6). However, these wereoffset by direct increases inrevenue (conditional on price andharvest) and wholesale prices(conditional on harvest).
Figure 6: Point estimates of the annualaverage effect of the SSL protectivemeasures on quota received, harvest,wholesale price, revenue, and variablecost.
Conclusion
Mechanistic Effects: TheSSL protective measuresnegatively impacted net revenuethrough a reduction in quota(Quota Mechanism) and bydisplacing vessels fromproductive fishing grounds(Closure Mechanism). However,these were partially offset bysubstituting towards harvestingother species with higher pricesand/or recovery rates (ValueMechanism) (Figure 7).
Figure 7: Point estimates of theclosure, quota, and value mechanisms,along with the total effect of SSLmeasure on net revenues, from the SEMmodel.
There is little evidence of a negative effect of the SSL closures onthe net revenue of affected vessels. However, our decompositionidentifies multiple counteracting mechanisms. Vessels werenegatively impacted through the reduction in Atka mackerel quotaand from being displaced from historical fishing grounds, but wereable to partially offset this by substituting to other species outsideof the closure. These results confirm the common verdict that theshort-run cost of a spatial closure will depend on the opportunitiesoutside of the closure. Our results demonstrate the importance ofunderstanding the mechanisms underlying the economic impacts ofspatial closures, for both generalizing results beyond the setting athand and for aiding policy makers in designing policies that fostermechanisms that achieve the intended policy objectives.
Chinook bycatch in the Bering Sea pollock fishery• No total cap on Chinook
bycatch prior to 2011; • Large area and
rolling hotspot closures used
• Hard cap and other incentive measures implemented in 2011 with Amendment 91
U.S. Department of Commerce | National Oceanic and Atmospheric Administration | NOAA Fisheries | Page 35
Opportunities
U.S. Department of Commerce | National Oceanic and Atmospheric Administration | NOAA Fisheries | Page 37
From SSC Minutes: “The incentives section (7) provided a very helpful discussion that should not get lost in the detailed evaluation of future alternatives. Incentives provide a supplement (and potentially even an alternative) to index-based approaches.”
Public goods provision of ideas• Council settings (presentations,
analyses, plan teams, etc.)• IPHC• IEA, ICES, PICES, MSEAS,
SICCME, ESSAS, etc.• Inter/multi/trans-disciplinary work• Diverse scientific settings
U.S. Department of Commerce | National Oceanic and Atmospheric Administration | NOAA Fisheries | Page 39
Page 40
Thanks!
Questions?
Photo CSMphotos