Prepared By: Fisheries Statistics and Economics Division, Silver Spring, Maryland, 1996 Maury F. Osborn, Program Manager David A. Van Voorhees, Gerry Gray, Program Statisticians Ronald Salz, Program Fishery Biologist Elizabeth Pritchard, Program Computer Specialist Mark C. Holliday, Chief, Fisheries Statistics and Economics Division
U.S Department of Commerce - Mickey Kantor, Secretary
National Oceanic and Atmospheric Administration - D. James Baker, Under Secretary
National Marine Fisheries Service - Rolland A. Schmitten, Assistant Administrator
PREFACE
The National Marine Fisheries Service (NMFS) initiated a series of surveys in 1979 to obtain standardized and comparable estimates of participation, effort, and catch by recreational anglers in the marine waters of the United States. Continued efforts to develop and maintain a comprehensive marine recreational fisheries data acquisition and analysis system implemented the first priority of the NMFS Marine Recreational Fisheries Policy established in 1981.
This publication was prepared under the supervision of John F. Witzig. Maury Osborn adapted it for homepage use. Other NMFS personnel involved in report preparation, survey design, survey implementation, and analyses of data were Gerald J. Butler, Ronald J. Essig, Garry Gray, Mark C. Holliday, Maury F. Osborn, Liz Pritchard, Ron Salz, Wade Van Buskirk, and David A. Van Voorhees. The NMFS would appreciate comments on the information presented in this report, as well as suggestions for improvement or changes in the tabular presentation. Please address comments to: U.S. Department of Commerce, NOAA, NMFS, Fisheries Statistics Division - F/RE1, Room 12456, 1335 East-West Highway, Silver Spring, MD, 20910.
The data used for the survey is collected by two independent, but complementary, surveys:
Numerous NMFS methodological studies indicated that the survey should be structured around this data collection approach. These studies showed that a telephone survey could be used to collect reliable data on recreational fishing effort. Data on fishing trips became less reliable beyond a 2-month period due to recall problems. Information on the actual catch such as species identity, number, and both weights and lengths of fish caught could not be reliably collected by telephone. Catch data are obtained from anglers intercepted by trained interviewers stationed at fishing access sites. Data from the two independent surveys are combined to produce estimates of total participation, effort and catch. Using the complemented surveys approach, marine recreational fishing estimates (not including shellfishing) are calculated for six two-month periods (waves) in each year. Table 1. Types of data collected by the complementary survey methods.
| Intercept Survey | Telephone Household Survey| Number, weights
and lengths of fish caught by species | Presence of marine recreational
anglers in the household | State and county of residence
| Number of anglers per household | Avidity level - trips
per year | Fishing trips in 2-month period | Mode of fishing
| Mode of each trip | Primary area of fishing | Primary
location of each trip | |
There are geographical and temporal exceptions that are explained in the Sampling Coverage section of this report. Sampling efforts during January and February since 1980 were generally limited to the Pacificcoast, Gulf coast states and the Atlantic coast of Florida.
Total survey effort during a one-year period usually involves more than 76,000 intercept interviews and over 265,000 telephone interviews. The following sections briefly summarize the methods and procedures employed in the telephone survey, the intercept survey, and the calculation of estimates from the information collected by the two surveys.
The telephone coastal household survey is carried out in 2-week periods of interviewing starting the last week of each 2-month period of fishing activity (waves) and continuing in the first week of the following month. For example, for the January/February wave, coastal households are called during the last week of February and the first week of March. Respondents are asked to recall on a trip-by-trip basis all marine recreational fishing trips made within their state during the 60 days prior to the interview. Telephone sampling effort is directed at households located in coastal counties. Coastal counties are defined as:
Depending on the geographic area, about 70 to 90 percent of the anglers interviewed by the intercept survey live within the telephone survey calling area. The telephone interview sample quota for each wave varies with the amount of fishing activity expected.
The allocation of calls made in each telephone prefix is based on the frequency of households assigned that prefix. The appropriate number of sample points (household telephone numbers) is then randomly generated for each county. All households are eligible for contact each wave, regardless of whether they were contacted in a previous wave. Telephone interviews are conducted between 10:00 am and 9:30 pm (respondent's local time) on weekdays and weekends.
The intercept survey consists of interviews which gather catch and demographic data from marine recreational anglers who have just completed fishing in 1 of 3 fishing modes:
The intercept survey continuously samples angler catches during the 6 two-month sampling periods from January through December. Intercept sampling is stratified by state, mode, and two-month wave with a minimum of 30 intercepts in each stratum. Beyond this minimum, samples are allocated in proportion to average estimates of fishing pressure from the three previous survey years. Complete coastwide lists of access sites for marine recreational fishing were created in 1979 and are continuously updated. (Site lists will be posted on the home page in the future.). Sites are chosen for interviewing assignments by randomly selecting from among the listed access sites weighted by estimates of expected fishing activity. Sampling is distributed among weekdays, weekends and holidays in such a manner as to assure that about 60 percent of the interviews are collected on weekends and holidays on the Atlantic and Gulf coasts and 75 percent on the Pacific coast. Anglers are intercepted, screened, and interviewed at assigned access sites upon completion of their fishing trips. A small number of interviews (less than 5 percent) are conducted with beach/bank shore mode anglers who have not completed their trip. At heavy use sites, every nth angler is intercepted and interviewed. For example, every second or third angler might be interviewed if the site is too busy to interview all anglers. Each interview consists of:
Interview procedures vary slightly among fishing modes:
Interviewing procedures have been developed to allow separate recording of information on the following:
The estimates derived from the telephone and intercept surveys fall into three categories:
Fishing effort is the estimated number of fishing trips taken by individual anglers. The number of individual fishing trips is estimated for each state, coastal county, mode, and bimonthly wave. Data from the telephone survey of households are used to calculate mean numbers of trips per household in each fishing mode during each wave. This number is multiplied by the number of permanent, full-time occupied households in the coastal county (Bill Communications, Inc. 1995) to estimate total number of fishing trips in each mode by coastal county residents. Data on the number of households in the coastal zones are updated annually. The telephone survey can not provide information on the number of trips taken by people who reside in households beyond the 25- or 50-mile coastal zone from which the telephone numbers are drawn. Ratios obtained from the intercept survey are used to estimate the numbers of trips taken by out-of-state residents and state residents of non-coastal counties. For example:
Similar procedures are used to estimate fishing trips taken in the state by anglers residing in other states. In certain circumstances a ratio obtained from the intercept survey is used to estimate trips taken by anglers who reside in coastal counties but do not have telephones. Ratios are also used to adjust effort estimates if the proportion of coastal county residents living in full-time occupied households with telephones differs significantly between the intercept survey sample of anglers and the most recent census.
The catch of each finfish species is estimated for each subregion, state, fishing mode, primary fishing area, and wave. The total number of fish caught in a particular fishing mode and area of fishing is estimated from:
Telephone Data Intercept Data Results
Number of finfishing trip Mean catch per trip Number of each by mode and area X by species caught by = species caught by mode mode and area and are Catch estimates are added among strata to obtain estimates of catch of each species at the subregion, state, mode, primary area, or wave levels. All fish that are caught by intercepted anglers are not available for the interviewer's inspection. The intercept interview and the estimation procedures distinguish between those fish brought ashore in whole form, and those not brought ashore in whole form:
Expanded catch estimates are made for these three types of catch. The purpose of estimating three catch types is to distinguish between those species identified and measured by trained interviewers, and those species reported to the interviewers by anglers. Previously cited methodological studies indicated species are often misidentified by anglers and their reported measurements subject to several types of bias. Lengths and weights are obtained by sampling the fish caught and brought ashore in whole form by intercepted anglers.
Most of the trips sampled in the intercept survey are completed trips, with anglers being interviewed only at the end of the fishing trip. Some incomplete trips are sampled in the shore mode, and they are converted into complete trips by multiplying the recorded catch per hour by the anticipated total trip length.
The estimated number of participants, derived from telephone and intercept data, has to account for varying levels of reported fishing avidity.
This section describes the NMFS procedures chosen to identify and adjust extreme or "outlying" observations and to adjust for other sampling practicalities.
Population estimates such as total fishing effort are subject to wide variability when based on a relatively small number of interviews. The protocol used in the MRFSS to produce estimates of total catch and effort is very sensitive to the inclusion of a few extreme observations in reported trips by individual households and in intercept survey ratios of coastal to non-coastal and out-of-state anglers. Telephone survey households that report an extreme number of fishing trips for the sample period tend to have a disproportionate effect on the estimate of average fishing effort, producing unrealistically high estimates of total fishing effort. These extreme estimates are adjusted in the folowing manner:
Estimation of fishing effort for the party/charter boat and charter boat sectors of the recreational fishery is difficult due to the relatively low incidence of reported fishing activity in these modes by households contacted in the telephone survey.
To reduce the effect of small sample sizes on the effort estimates for the charter boat fishery:
Total fishing effort estimates at the subregion/state/mode/area/wave (cell) level are comprised of 3 component estimates:
The last two components are estimated using ratios of non-coastal county resident trips and out-of-state resident trips to coastal county resident trips obtained from the intercept survey. These ratios are applied to the base estimate of coastal county resident fishing effort derived from the telephone survey. Unusually high ratios lead to unrealistically high estimates of fishing effort attributable to non-coastal or out-of-state anglers. This is common in the charter boat fishery in the South Atlantic and Gulf of Mexico subregions and is adjusted as follows:
In some cases there are missing data for fishing households contacted in the telephone survey, where some or all of the trip information is not collected (inability to contact an identified angler or respondent fatigue). Although proxy data are collected whenever possible from other qualified household members, there are still circumstances where a household is initially identified as a fishing household, but household fishing data is either incomplete or unobtainable. We use a statistical procedure called "hot deck" resampling to impute values for missing data. This procedure substitutes a randomly selected complete observation obtained from a similar household or angler for each missing observation (eg. number of trips per angler, mode of each trip). For example:
In some cases, no adjustment can be made for missing data. The estimation procedure combines information from the telephone household and intercept surveys. The completeness of the resulting data matrix is occasionally affected by the presence of "missing cells" in which no information is obtained from one or both surveys. The presence of missing cells results in an underestimate of the total number of fish, or an estimate of number of fish but no corresponding estimate of the weight of these fish.
In some cases there is an estimate of landings or harvest, but no fish were measured in that cell and there is no estimated weight. Missing weights are estimated by length-weight equations or using a protocol to impute an average weight for the species from the closest adjacent cell, such as the adjacent mode in the same area and state.
A simple random sample model stratified at the county level is used for conduct of the telephone survey of coastal county households. However, the variance associated with the average number of fishing trips is calculated using a simple random sample model stratified at the state level. Estimation of the variances associated with the average catch and weight of catch estimates obtained from the intercept survey is based on the assumptions that the primary sampling unit was a fishing trip by an individual angler and that there is no clustering effect due to the collection of groups of interviews at each visited site. These assumptions were empirically verified in pilot surveys. Therefore, the variance is estimated using the standard variance equation for a stratified random sample: where: si2 = estimated variance of stratum i, and ni = number of interviews in stratum i. Estimation of the variance of the combined estimates for the two surveys requires special attention.
The sampling variance of the estimated total catch (for individual species and for species groups) is calculated in terms of the expected values and sampling variance of xi (average catch) and yi (total number of trips) for each stratum. Total catch is not normally distributed and therefore direct examination of the precision of the estimates is difficult. However, simulation experiments indicated that a normal approximation is satisfactory for construction of 95 percent confidence intervals around the estimated total catch.
Precision refers to the dispersion of the sample measurements used to calculate an estimate and the resultant variability in the estimate.
The standard error is necessary for calculating confidence intervals around an estimate.
The standard error is also used to calculate the proportional standard error (PSE).
Effort estimates are calculated at the subregion, state, mode, and wave level. Catch estimates are calculated at the subregion, state, mode, area, wave, and species level. All estimates and variances are additive across strata because they are estimated independently. For example, the estimated number of fishing trips in a subregion on an annual basis is the sum of trip estimates from all states in the subregion, all modes, and all waves. The data used to produce summary tables are maintained in their unaggregated form in the MRFSS data base. Catch estimates for some species are aggregated into species groups for snapshot summary tables.
Special care is advised when comparing catch estimates for the 13 years (1979-91) of MRFSS data because of differences in sampling coverage. In the South Atlantic and Gulf subregions the MRFSS has not collected catch data from partyboats since 1985, so estimates for these subregions now only include charter boats. Marine recreational fishing in Texas is monitored by the Texas Department of Parks and Wildlife and has not been surveyed by the MRFSS since 1985. On the Pacific coast, ocean boat trips during certain waves and salmon trips are not sampled because they are surveyed through the state natural resource agencies.
Large increases or decreases in the estimated total catch of individual species groups may be due to the inclusion in the sample of unusually large catches of a species by one or a few anglers in a particular wave and mode rather than an actual biological or economic phenomenon. Calculation of an estimated catch for the species for an entire state then results in a high estimate. Pulse fisheries can also result in unusually high estimates. In addition, examination of the standard errors or coefficients of variation associated with the estimates will indicate if significant differences exist between the two estimates. Trends in the catch estimates from 1979 should also be evaluated when examining the total catch estimates.
In 1995, we conducted a major effort to correct remaining errors in data sets produced by the 1979-1992 MRFSS telephone and intercept surveys. Using recently developed methods, we found and corrected telephone survey data errors in the accounting of total number of residential households and total sampled non-fishing households by county, as well as intercept survey data errors related to species codes, length/weight relationships, accounting of group catches, and other illogical or out-of-range variable values. The data corrections had relatively minor effects on the state level estimates of effort and finfish catches. Previously distributed intercept data sets and tables should be disregarded and should be replaced by the revised data sets. All of the MRFSS effort and catch estimates for 1980-1993 were re-calculated using the corrected data sets and improved estmation methods described in the estimation section.
Since its inception, the basic MRFSS effort has been enhanced through the cooperative participation of Federal and state agencies. Table 2 outlines Federal and state involvement in the MRFSS intercept and telephone household surveys during 1990-1991 (this table will be updated in the future to reflect the historical series as well as the Pacific ). Either additional questioning or additional sampling was conducted in every Pacific, Atlantic and Gulf coast state at some point during the 2-year period. States participate in the MRFSS survey effort in various ways.
These cooperative efforts provide participants with needed data while taking advantage of the cost savings and methodology of a major survey. Headboats are monitored by the Partyboat Logbook Survey conducted by the Southeast Fisheries Center of the NMFS. That survey collects and provides data on partyboat (headboat) catches throughout the Southeast Region. The Texas Parks and Wildlife Department collects and provides data to NMFS in lieu of NMFS conduct of the MRFSS in Texas. Interviews that would have been conducted by the MRFSS in Texas are reallocated to the remaining states. Table 2. Federal and state add-ons to MRFSS sampling, 1990-1991.
Intercept Survey Telephone Survey
State 1990 1991 1990 1991
ME NH MA S S RI S S CT S S NY S S NJ S S Q Q DE S S S S MD VA NC S S S S SC GA FL AL MS LA TX
Note: S = State funded sample size increase F = Federal funded sample size increase Q = Additional telephone survey questions