Lotto nothing? The budgetary impact of state lotteries

Abstract

Lottery revenues are often touted as an independent revenue source for states. Using 32 years of state financial data, the fallacy of such thinking is demonstrated. Being the first to control for the self-selection of being a lottery state, it is found that overall tax revenues decline with increased lottery sales. Moreover, it is discovered that this decline is driven by a decrease in revenues from general sales and excise taxes, which is only partially offset by increases in income tax receipts. Such findings are attributed to a combination of behavioural and political responses following the lottery's implementation.

Acknowledgements

We thank Geoffrey Jehle as well as participants at the 2002 Meetings of the National Tax Association and Southern Economics Association for their comments and insight. Finally, we would also like to recognize funding from the Ford Scholars Program at Vassar College for making this project possible.

Notes

 See Miyazaki et al. (1998) for a thorough survey on the numerous studies looking at the regressivity of the lottery.

 See, for instance, http://www.stickerzone.com/14lottaxonpe.html (11/4/2002).

 In general, public and private expenditures can be complements or substitutes. For example, the marginal utility of automobile spending is likely to increase with public spending on roads, but is likely to decrease with public spending on public transit.

 It is interesting to note from Equation 3 the direct effect of a change in lottery revenue is tβ on tax revenue, which is smaller than the (1 − α − (1 − t))β effect on government spending.

 Although an implicit tax exists on the lottery, the distinction is made between net lottery revenues and tax revenues throughout the paper.

 Note that the elasticity with respect to per capita lottery revenue is equivalent to one with respect to aggregate lottery revenue as long as population remains constant; that is, (dt/dl) = (dt/dL), so Condition 7 holds for Equation 8.

 Technically, this conclusion requires there to be no income effects. Nonetheless, the degree of substitutability (or complementarity) will be related to the size of u _xg . That is, for sufficiently negative u _xg , x and g will be substitutes, and conversely for complements.

 Because some states will have zero tax revenue for various tax instruments, 0.1 is added to all dollar figures prior to taking the natural log.

 Recall that Filer et al. (1988) and Caudill et al. (1995) have shown that states with more high income individuals tend to adopt lotteries. On the other hand, Alm et al. (1993) and Erekson et al. (1999) show that in times of financial crisis, states will look to lottery revenue as a substitute for tax revenue. In either case, the intuition holds, and any selection effect must be corrected if we are to apply the results more generally.

 Comparing this result to estimates using per capita values (Fink et al., 2003) strengthens this result. Per capita estimates show an insignificant impact of lottery revenue on tax revenue. However, the OLS estimates continue to be higher than the corrected estimates.

Also, keep in mind that many states enacting lotteries are located in the Northeast and Midwest, regions that have been losing population during the time frame of this study. Hence, this may be a case of the denominator decreasing quicker than the numerator. In other words, the people who remain in the state are paying more in taxes, but because there are fewer of them left than previous years, the totals in the aggregate are decreasing.

 Results of the full model are available upon request.

 The change in expenditures (E) can be imputed by

where expenditures are approximated by sales tax revenue (S) divided by the sales tax rate (t _s ). Thus, the elasticity can be transformed by the proportionate change in lottery revenue (ΔR/R) to yield the change in expenditure. Assuming average 1999 levels, a 6% sales tax, and a US$1 increase in lottery revenue gives ΔE = − 0.059·(3545/0.06) · (1/316) ≃ US$11.

 SUR was also applied using lagged lottery revenues with similar results.

 In dealing with taxes with multiple brackets, an alternative approach is to create an effective tax rate. This method forces the researcher to make assumptions on family size, income, etc. The average tax rate described above is less problematic in that regard.

 In unreported results, the previous year's tax rate is also included. The coefficient on lottery revenue remains essentially unchanged.