An econometric model of determinants of visitor use on western national forests

Abstract

The accuracy of visitor use data from the National Visitor Use Monitoring Program (NVUM) allows for testing the relationship between public land visitation and individual site characteristics and facilities. In an attempt to predict visitation on both BLM and USFS lands, forty National Forests in the Western US were chosen for their spatial and landscape resemblance to BLM lands. Using multiple regressions, facility and landscape characteristics have a statistically significant relationship with the four recreation types in NVUM data: Day use developed sites (DUDS), Overnight use developed sites (OUDS), General Forest Area (GFA), and Wilderness. Mean absolute percentage error (MAPE) of prediction calculated using ten out of sample National Forests for Wilderness was lowest at 69%, with OUDS, DUDS and GFA higher at 93%, 103% and 115% respectively. As an alternative method to estimate the predictive power, stepwise procedures were applied to all forty observations. These resulting models were used to construct a spreadsheet calculator that provides an annual visitation prediction for a USFS or BLM land.