Nonresponse Weighting Adjustment in a Telephone Survey of Recreational Fishing Activity

In the United States, information about recreational saltwater fishing effort is generally collected through the Coastal Household Telephone Survey (CHTS), a random-digit-dial (RDD) survey of coastal county households.  Like many RDD surveys, CHTS response rates have declined in recent years, increasing the potential for nonresponse error in survey estimates.  Because little is known about nonrespondents, the CHTS accounts for unit nonresponse through a simple stratum cell-weighting adjustment that assumes random nonresponse.  This may be an invalid assumption as previous studies have demonstrated that avid anglers are more likely to respond to fishing surveys than non anglers.  Here we describe efforts to enrich the CHTS sample, and subsequently define more meaningful weighting cells, by matching the sample to external data sources, specifically, databases of licensed saltwater anglers.  We assess the extent to which CHTS estimates are subjected to nonresponse error resulting from differential nonresponse, the success of matching and the impact of the revised weighting approach on survey estimates.