The Texas Parks and Wildlife Department mailed 2 different survey questionnaires to 5,000 randomly selected households to estimate the number of residents fishing and the ways they fished in Texas from 1 September 1978 through 31 August 1979. The questionnaires in this survey were identical except one sub-group (1000) asked household members questions regarding their possession of a valid fishing license. Differences in responses to these 2 questionnaires were used to assess response biases. Estimates of total license sales (a known statistic) from survey data showed a strong positive bias (license sales were overestimated by 62% ). The mean number of fishermen per household was significantly greater (P < 0.05) in the group which was not asked about the possession of fishing licenses. Return rates and responses to questions regarding all other fishing activities were identical for the different survey questionnaires. Consequently, the number of resident fishermen in Texas (2.47 ± 0.40 million) was estimated using data from questionnaires that had asked about license possession and adjusting the estimate downward by 62%, whereas estimates of other fishing activities were made using all questionnaires from persons claiming they had fished. Our study shows that loading a survey with a question that permits the estimation of a known statistic can be a useful device to evaluate the effect of response bias.
