Abstract
Recent journalism and scholarship have noted a years-long decline in Americans' participation in rural forms of outdoor recreation such as hunting. While some attempt has been made to understand these declines few have analyzed the causes of these changes in a theoretically rigorous empirical manner. This study addresses this issue in two empirical approaches. First, we analyze survey data on hunting and various theoretical predictors from the General Social Survey. Second, we statistically analyze changes in hunting license acquisitions at the state level for a period of several years. We empirically test the Videophilia hypothesis (CitationPergams & Zaradic, 2006) as one explanation positing a substitution effect involving electronic forms of indoor entertainment. We find evidence that the switch to certain kinds of electronic entertainment as well as the growth in urban living explain the decline in hunting and discuss the implications of these findings and future directions for research.
Notes
1. The USFW (2006) service reports that wildlife watching has increased 12% (1996 to 2006), with “at-home” participation (e.g., bird-watching in one's neighborhood) explaining virtually all of the growth (away from home or outside of one's neighborhood wildlife watching actually declined over the period).
2. Much of our data are more widely available at the state level than at lower levels for a useful period of time; we are limited to this period as state-level data on computer/Internet usage is not available prior to the late 1990s.
3. “Do you (or does your [husband/wife]) go hunting?” Yes = 5,302 respondents (20% of the total sample).
4. No question asks about comparable use of computers, although WWW use suggests access to a computer.
5. The GSS does not provide controls for fuel, travel or other costs that are salient to hunting; we thus employ cost controls in the state-level analysis where we obtained more direct cost-related data.
6. We use panel-corrected standard errors as a method to accommodate for both heteroscedasticity and spatial autocorrelation, while the inclusion of a lagged dependent variable accommodates serial autocorrelation.
7. The inclusion of a lagged dependent variable on the right hand-side of the equation also is a conservative test of the remainder of the theoretically guided independent variables as it absorbs much of the predictive variance.
8. We employ the broader measure of TV subscribers as it captures viewership not indicated by cable TV such as both network TV and satellite networks. We did not find state-level time-series data on dish satellite viewership.
9. We extrapolated the earliest value (2000) back to the beginning of our sample—1997; thus, it is relatively time-invariant but yet is the best currently available.
10. Manufacturing work hours are the most complete data series available from the source.
11. We thank an anonymous reviewer for this suggestion.
12. Both the fatality and ownership variables have essentially the same effects in the analyses and neither influence the videophilia findings discussed later in the article. We did find shotgun and rifle fatality data for a single year (2007), and perhaps because it was for a later year than the sample period (1997–2002), we found no effect of the shotgun/rifle proxy in a separate analysis.
13. We thank an anonymous reviewer for this suggestion.
14. We dummy coded states as South, Midwest, Northeast, and West. West is the reference category.
15. VIF scores never exceed 10 for any of the models, yet gun ownership and the lag of hunting license VIFs are 8 and 6, respectively, likely because the two are highly correlated. The time-invariance of the gun ownership variable may contribute to the multicollinearity. Removing them does not diminish the significance of any of the results—in fact, the cost of agricultural land becomes positive and significant. However, removing the lag is ill-advised given the fairly low VIF and its importance as a conservative test of the hypotheses (as well as the importance of controlling for gun ownership/regulation).
16. Neither the cable TV nor the general TV variables are significant when entered separately or together into the models. For the sake of brevity, we report only the effects of the combined index here.
17. We also tried proxies for rural land-supply and not just dollar value (e.g., agricultural-use land, forestland, pasture/rangeland, public land in the form of national/state parks from CitationNatural Resources Conservation Service [2009]), but all failed to attain significance possibly indicating that diminishing land access (e.g., suburbanization's destruction of farmland) is not threatening the sport. The variables may also not have been significant because much of the land area covered by the proxies is not utilized for hunting, at least most of the time.
18. In fact, large R-squares are typical of pooled, cross-sectional designs given that more information is gleaned (across time) from each unit in the analysis and factored into the explained variance. Cross-sectional designs, by contrast, only tap variance across multiple diverse units and do not delve more deeply into each unit's contributions across a larger time frame.