Centner: Sensitivity of angler benefit estimates to tfie definition of substitute sites considered by tfie angler 163 



The summation of the indirect utilities is across the 

 choice set facing each individual, S,, and not the global 

 choice set, S. 



Since 1979, data have been collected on marine recre- 

 ational angling during the Marine Recreational Fishery 

 Statistics Survey (MRFSS). The MRFSS consists of 

 two independent but complementary surveys: a field 

 survey and a telephone survey, conducted annually in 

 six two-month "waves." The field survey is an intercept 

 survey of anglers conducted at fishing access sites and 

 is designed to obtain a random sample of recreational 

 trips for computing catch per unit of effort. Fish re- 

 tained by interviewed anglers are sampled for length 

 and weight. Fish not retained by the angler are not 

 observed, but count data on this unobserved catch are 

 collected. The data on harvest provide a picture of the 

 size distribution of the kept fish from the stock. If a 

 fishery is regulated by a minimum size limit, a catch- 

 and-keep rate calculated from these data indicates the 

 catchability of fish large enough to keep. As such, it is 

 the observed rate at which anglers can catch and keep 

 fish from a stock. 



The intercept sample is stratified by state, wave, fish- 

 ing mode, fishing area, catch type, and species. Specific 

 data elements collected during the intercept survey 

 include state, county, and zip code of angler's residence, 

 hours fished, primary area fished, target species, gear 

 used, and days fished in the last two and 12 months. 

 During the intercept portion of the survey, data are 

 collected on the length and weight of all fish species 

 retained by the angler and the species and condition of 

 all catch not retained by the angler. Upon completion of 

 the base MRFSS, anglers in the Northeast (NE) (Maine, 

 New Hampshire, Massachusetts, Rhode Island, Con- 

 necticut, New York, New Jersey, Delaware, Maryland, 

 and Virginia) were asked to complete a short add-on 

 questionnaire in 2000. This questionnaire provided 

 information on whether or not the trip was a single-day 

 or longer trip and, if it was a multiple-day trip, whether 

 fishing was the primary purpose of the trip. Data were 

 also collected on the angler's saltwater fishing experi- 

 ence (in number of years), boat ownership (whether 

 owned or not), and whether or not the individual took 

 time off without pay to take the fishing trip. If the in- 

 dividuals responded in the affirmative to the later, they 

 were asked the number of hours in their work week and 

 their personal income. The survey instrument is avail- 

 able at the NMFS web site (NMFS^). 



In order to reduce the complexity of the modeling ef- 

 fort, the angler's choice to fish rather than participate 

 in some other recreational activity, the angler's choice 

 to fish in a private or rental boat mode, and the angler's 

 decision regarding a species target are exogenous to 

 the model. Because the area fished is not documented 



^ NMFS (National Marine Fisheries Service). 2006. Fisher- 

 ies Statistics and Economics Division. Survey Instruments. 

 Website: http://www.st.nmfs.gov/stl/econ/surveys/survey_ 

 timeline.html (accessed on 13 August 2006). 



in the MRFSS, a fishing site is defined as the point of 

 fishing access. As mentioned previously, the treatment 

 of all substitute sites can be quite costly from a data 

 standpoint for a number of reasons. Because thousands 

 of individual sites in the North East (NE) region are 

 recognized in the MRFSS, estimation can be a lengthy 

 process, particularly with nested models. In addition, 

 not all species are sampled in all survey waves at all 

 sites in all modes; therefore the calculation of historic 

 catch rates at the individual site level results in many 

 empty cells. To speed estimation and to fill some of 

 these empty cells, all sites within a coastal county were 

 aggregated into one site that represented that county. 

 Across the NE, there are roughly 63 coastal counties, 

 and therefore 63 sites. In order to examine whether this 

 aggregation strategy induces any bias into the estima- 

 tion of the conditional logit model, a variable im) was 

 created that represents the number of MRFSS sites 

 aggregated into each new site. The rule that a county 

 equals a site was not strictly followed in all cases. Some 

 geographically diverse counties (i.e., those counties with 

 both ocean frontage and bay frontage) were separated 

 into two sites because of the different opportunities 

 provided by these different types of water. 



Both the historic five-year average catch rate (catch 

 rate) and catch-and-keep rate (KRATE) were calculated 

 for the boat mode for each wave and site combination. 

 KRATE measures the catchability of a striped bass 

 large enough to keep, incorporating the five-year aver- 

 age probability of catching a striped bass large enough 

 to keep. The distinction between the catch rate and 

 KRATE is particularly important for striped bass be- 

 cause this species is heavily regulated. Historic KRATE 

 was used in the model because it represents the portion 

 of the catch that an angler would be able to keep, not 

 just the increase in overall catch. It is also the mea- 

 sure of angler quality used in the Whitehead and Haab 

 (1999) study. Even after the site aggregation, some 

 counties did not contain enough data points on striped 

 bass catch from the boat mode over the 5-year period. 



Whitehead and Haab (1999) replaced missing catch 

 rates using the catch rate from the nearest neighboring 

 site in some cases and with zero values, in other cases 

 depending more or less on mode. Hicks et al. (1999) 

 recognized this approach to be ad hoc and estimated 

 his model using both nearest neighbor and zero value 

 assignment, another ad hoc approach, and found that 

 the treatment of missing values did not significantly 

 affect the welfare estimates. He concluded that the zero 

 assignment is perhaps less arbitrary because the empty 

 cells actually convey information. That is, if there are 

 no observations of average catch within a particular 

 wave-Hmode+site-i-species combination, the site is not 

 very productive over that combination. As a result, zero 

 assignment requires less judgment by the researcher; 

 therefore that is the approach used here. 



Estimating any demand equation requires a price 

 variable. Because recreational fishing experiences are 

 not openly traded in markets, travel cost (both the ac- 

 tual cost of travel plus the opportunity cost of time) is 



