A Comparison of Revealed, Stated, and Actual Behavior in Response to a Change in Fishing Quality

Abstract

Due to water quality problems at its fish culture stations, Pennsylvania stocked 28% fewer catchable trout in 2002 than it had in prior years. The impact of this stocking decrease on license and trout stamp sales was projected based both on an econometric analysis of historical license sales (revealed behavior) and on a telephone survey of trout anglers (stated behavior). The econometric analysis showed no relationship between stocking levels and angler participation. However, in the telephone survey, 11.5% of current trout anglers stated that the cut in stocking would lead them to stop fishing for trout, including 3.2% who would stop fishing altogether. The projections based on stated behavior did a better job predicting resident license sales than did the projection based on revealed behavior, but anglers overstated their tendency to stop buying trout stamps.

This research was financially supported by the Pennsylvania Fish and Boat Commission. The authors are grateful to Yanguo Wang for econometric assistance. The authors alone are responsible for all statements and errors.

Keywords:

• sport fishing
• license sales
• stated preference
• revealed preference

Notes

This research was financially supported by the Pennsylvania Fish and Boat Commission. The authors are grateful to Yanguo Wang for econometric assistance. The authors alone are responsible for all statements and errors.

USFWS national estimates of fishing license sales also showed a pattern of increasing sales during the 1970s and 1980s, and a decline during the 1990s, although the percent changes were smaller than in Pennsylvania.

The specific wording of the question was “Suppose that X% fewer trout are going to be stocked next year than this year. Would you still buy both a fishing license and a trout stamp for 2002?” with X = 10, 30, or 50. A follow-up question identified respondents who would buy a license but not a stamp.

Based on linear interpolation between data from 2000 census and PA State Data Center’s July 1, 2002 population estimate.

Item non-response rates were 0.2% for the question regarding plans for 2002, and 1.3% for the question regarding license purchase in 2000.

Item nonresponse for this question was 0.3%.

Confidence intervals for these predictions are generated using Monte Carlo techniques (Krinsky & Robb, 1986).

In practice, the analyst would not have advanced knowledge of the unemployment or inflation rate, and would have to rely on projections of those underlying economic variables.

The license sales projection from the econometric analysis and its confidence interval were calculated using the SAS ARIMA procedure.

Confidence intervals for the license and stamp sales projections from the survey responses include two sources of uncertainty, the sampling variances of the baseline rates of entry and exit, and the sampling variance of parameters of the logistic regressions from Table 2. The combined effect of these two sources of uncertainty was simulated using Monte Carlo techniques.