The impact of disaster exposure on the success of open space referenda

ABSTRACT

In response to losses of open space, local voter referenda have become a popular and effective means to raise funding for land conservation in the US. This paper provides empirical evidence on the effects of disasters on the success of voter referenda for open-space conservation. Our results show that voters become less supportive of subsequent ballot measures if their communities recently experienced a major disaster. The preferred specification reveals that a recent major disaster is associated with an 8.4% reduction in the odds of a ‘yes’ vote. We find further evidence that the effect is especially strong for referenda using tax increases to fund preservation. This disaster effect, however, is temporary rather than permanent. For disaster events occurring more than 12 months prior to a referendum, we find that the effect is no longer statistically significant.

KEYWORDS:

• Natural disasters
• open space conservation
• voter referenda
• environmental amenities

JEL CLASSIFICATION:

• Q38
• Q54
• Q58

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Trust for Public Land’s LandVote Database: www.landvote.org.

² See Section III of this paper for a more detailed discussion of this literature.

³ Some uncommon disaster types, like earthquake, drought, landslides and levee break, are dropped in this analysis due to few observations. As discussed later in this paper, we include these rare disasters in our specifications as a robustness check and find similar results.

⁴ We collected population data from the 2000 Census to construct population change from 2010 to 2000. Housing value and household income are adjusted to 2016 US dollars using the consumer price index.

⁵ Because presidential elections are held every four years in the US, we use election outcomes in year 2000 for years 2000 to 2003 for a county. A similar strategy is used for other election cycles.

⁶ It is possible that a jurisdiction held more than one referendum during this study period. For these cases, we consider each vote as a separate observation.

⁷ We use subscript ‘−t’ to indicate that the variable is constructed based on recent past disaster events.

⁸ These referenda span 37 different states.

⁹ Massachusetts and New Jersey are the most highly represented states, accounting for about half of the observations in the sample.

¹⁰ For example, variable ‘wildlife’ equals 1 if the referendum purpose includes wildlife preservation.

¹¹ We find similar estimation results when we reestimate the models using standard errors clustered at the county level.

¹² Since a logit transformation is used in the model, the estimated coefficient represents the effect of a one-unit change in the explanatory variable on the natural logarithm of the ratio of the proportion of ‘yes’ votes to the proportion of ‘no’ votes.

¹³ The regression models in Table 3 do not include the referenda variables listed in Table 1.

¹⁴ The underlying assumption is that if the emotional effect plays a significant role, the coefficients on PDD should also be negative and significant in the subset of bond issue referenda.

¹⁵ Here we use subscript ‘+t’ to indicate that the variable is constructed based on occurrence of future disaster events.

¹⁶ The sample size is reduced because of the absence of information on IHP assistance prior to 2003.

¹⁷ It is well known that the appearance of open space referenda is largely the outcome of the efforts led by large national conservation organizations such as the Nature Conservancy, the Trust for Public Land and the Conservation Fund. These organizations use screening tools to direct their efforts towards areas where referenda are most likely to pass, and towards areas where conservation is especially valuable.

¹⁸ New Jersey approved the Green Acres Planning Incentive Program and Massachusetts approved the Community Preservation Act to provide state matching funds to jurisdictions that adopt tax increases for open space conservation.

¹⁹ Note that we do not include county and year fixed effects in the first stage specifications, because the sample includes municipalities that never held a referendum.