I spent January 29 and 30 at the Sheraton Silver Spring participating in this Workshop (where there were a lot of great presentations from NMFS economists). I was given 10 minutes to talk on the "Panel Discussion on Management Needs, Future Directions." I tried to make four points (which is pretty much my research agenda if I can ever focus on NMFS data).
1. Increased use of existing MRIP data is needed
The councils need fairly mundane valuation models, tailored geographically, for management decisions. A basic NRUM with species groups with the most recent MRIP data would be a major advancement. It would provide information on the value of catch that could be used in discussion of recreational quota, bag limits, etc. There is a lot of data, it is generally good (in the most basic sense), and it is underutilized.
2. Choice experiments are not a panacea for contingent valuation
NMFS has become reliant on choice experiment surveys for valuation. The major benefit of choice experiments in the marine recreational fishing context is that they are good for valuing individual characteristics of the fishing trip and getting information on the no trip option. Other than their great expense, the drawbacks of using choice experiments in the recreational fishing context is not well known.
Choice experiments haven’t faced the scrutiny of CVM surveys. In my opinion, they quickly rose to prominence as an alternative to CVM in the 1990s by assertion. I say that because choice experiments potentially face the same sort of problems as CVM (e.g., anchoring, ordering, scope, etc) but the profession has not demanded examination of these issues.
CVM surveys have been shown to suffer from anchoring bias, ordering effects and incentive incompatibility in multiple choice questions. The reaction has been to ask a single question. Similar effects have been found with choice experiments with no apparent reaction. Only a few studies have considered these problems with choice experiments.
CVM surveys have been criticized for passing internal scope tests but not split-sample external scope tests. The reaction has been to disparage CVM applications that don’t pass external scope tests. With the exception of a few studies, choice experiments routinely pass internal scope tests with no criticism that these are weaker tests.
The use of choice experiments by the NMFS is to be commended. But more recognition of their challenges is warranted.
3. Major improvements can be the MRIP with simple add-on surveys
MRIP asks for the number of trips taken within the state of intercept. The linked model can be used to assess changes in catch and site access on the overall number of trips. But this is inconsistent with the RUM because the RUM does not have an opt-out alternative. A negative change in fishing characteristics would cause trips to reallocate to other sites, modes and target species while the overall number of trips stays the same.
Simulating the effect of these negative changes in a linked model is inconsistent and biases the estimate of the change in the number of trips. The bias cannot be signed due to the forcing of trips. If no trip would be preferred and the angler is required through simulation to travel elsewhere then the bias is positive (i.e., too many trips would result from the linked model simulation). If the angler would prefer a trip to another site then the bias is negative (i.e., too few trips would result from the linked model simulation).
The basic RUM would greatly benefit from two additional pieces of data in add-on surveys:
- A CVM question that asks if the most recent trip were $X higher, would you still have taken the trip? This would provide an opt-out option for modeling changes in the number of trips and/or,
- Site specific RP trip data would allow for estimation of a repeated RUM (e.g., this was done in the 1997 SE add-on).
Both of these questions are relatively easy to implement in a short add-on survey. My guess is that this would be much cheaper than the choice experiment surveys.
Another extension is to ask contingent behavior (SP) follow-up questions. Contingent behavior trip questions are also cheap to design and feasible to implement. Questions about changes in bag and size limits, trip cost and expected catch would be valuable.
4. Joint estimation of MRIP and NMFS choice experiment data would strengthen both models
It isn’t clear if the coefficient on the cost variable in choice experiments is similar to the RUM travel cost estimate, which is probably more reliable (if it is measured accurately). Since the catch value is an inverse function of the cost coefficient, a biased cost coefficient could have large effects. It should be straightforward to combine the MRIP data to the NMFS choice experiment data and jointly estimate the model to determine if the cost coefficients are different. This was the point made by Adamowicz et al. (1994) in JEEM that helped launch choice experiments in environmental economics but joint estimation of choice experiment and revealed preference data has largely been ignored since.
Also, simple preference questions can be jointly estimated with MRIP data. Single questions on the 1997 Southeast and 1998 Pacific add-on MRFSS economic surveys could be used to jointly estimate management preferences.