This article was originally published in a journal published by Elsevier, and the attached copy is provided by Elsevier for the author s benefit and for the benefit of the author s institution, for non-commercial research and educational use including without limitation use in instruction at your institution, sending it to specific colleagues that you know, and providing a copy to your institution s administrator. All other uses, reproduction and distribution, including without limitation commercial reprints, selling or licensing copies or access, or posting on open internet sites, your personal or institution s website or repository, are prohibited. For exceptions, permission may be sought for such use through Elsevier s permissions site at: http://www.elsevier.com/locate/permissionusematerial

Earmarking lottery proceeds for public goods: Empirical evidence from U.S. lotto expenditures Abstract Craig E. Landry a, Michael K. Price b, a East Carolina University, United States b University of Nevada-Reno, United States Received 18 February 2006; received in revised form 4 September 2006; accepted 29 November 2006 Available online 12 April 2007 This study examines motives for lottery play using a state-level panel of lottery expenditures. We find that expenditures per capita are greater in states that earmark proceeds for public goods. Further, we find that casino gambling only impacts lotto play in general fund states. 2006 Elsevier B.V. All rights reserved. Keywords: Lotteries; Public goods JEL classification: H41; H27 1. Introduction Economics Letters 95 (2007) 451 455 www.elsevier.com/locate/econbase Charitable lotteries are a popular mechanism employed by fund-raisers to finance the provision of public goods. Morgan (2000) shows that lotteries obtain higher levels of public good provision than other voluntary mechanisms. A fundamental insight from this work is that lotteries can outperform voluntary mechanisms even when conventional explanations, such as risk loving behavior or consumption of Lead authorship is shared. The editor Eric Maskin and an anonymous reviewer provided remarks that significantly improved the study. Thanks to Jens Schubert for outstanding research assistance. John List, Andreas Lange, Nicholas Rupp, and Ted McConnell provided excellent comments during the discovery phase that greatly improved the study. Corresponding author. Department of Resource Economics, Mail Stop 204, University of Nevada-Reno, Reno, NV 89557, United States. Tel.: +1 775 784 1679; fax: +1 775 784 1342. E-mail address: mprice@cabnr.unr.edu (M.K. Price). 0165-1765/$ - see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.econlet.2006.11.022

452 C.E. Landry, M.K. Price / Economics Letters 95 (2007) 451 455 gambling, are suppressed. While recent experimental evidence provides support for the superior performance of charitable lotteries, it is difficult to infer the underlying motivations for giving in these settings (see e.g., Morgan and Sefton, 2000; Landry et al., 2006; Lange et al., 2007). This raises an interesting question: might individuals in these experiments be responding only to the wagering opportunity without any consideration for the attendant public good provision? In an attempt to understand the data patterns observed in these studies and discriminate between alternative motives for lottery play, we provide complementary evidence from naturally occurring data. Using state-level panel data from 1990 to 2000, we explore the effect of earmarking lottery proceeds on per capita lottery expenditures. Empirical estimates provide a number of insights suggestive of the importance of linking lottery purchases with the provision of a public good (education) and at odds with alternate explanations for lottery play. First, our results suggest per capita lottery expenditures are greater when proceeds are earmarked for education, suggesting individuals account for the link with public good provision. Second, we find that casino gambling is a substitute for state lotto in general fund states but has no effect on expenditures in Table 1 Summary statistics state lotto and casino gambling 1990 2000 Per capita expenditures Earmark for education Multi-state lottery Casino gambling Arizona 57.64 (6.75) No 1994 2000 1995 2000 Connecticut 197.92 (44.23) No 1996 2000 1992 2000 Delaware 219.49 (136.82) No 1991 2000 1996 2000 Indiana 90.02 (11.46) No 1991 2000 1995 2000 Iowa 57.81 (5.49) No 1992 2000 1991 2000 Kentucky 114.68 (32.64) No 1993 2000 No Louisiana 70.19 (16.18) No 1995 2000 1996 2000 Maine 105.15 (18.06) No 1990 1991 No Maryland 187.44 (23.92) No No No South Dakota 123.29 (44.9) No Yes No West Virginia 85.08 (37.36) No 1991 2000 No California 63.33 (10.05) No 1992 2000 1995 2000 Florida 142.33 (9.34) Yes No Yes Georgia 201.82 (35.18) Yes No Yes Idaho 65.14 (10.41) Yes No No Illinois 119.81 (5.22) Yes Yes No Michigan 136.03 (18.85) Yes No 1996 2000 New Hampshire 116.84 (29.63) Yes No 1994 2000 New York 150.06 (39.54) Yes No No Ohio 176.85 (25.06) Yes No No Missouri 64.59 (19.56) 1992 2000 Yes 1994 2000 Montana 65.58 (110.58) 1990 1995 Yes No Oregon 261.41 (156.26) 1997 2000 1992 2000 No Texas 138.95 (24.79) 1997 2000 No No Vermont 104.79 (23.43) 1998 2000 No No Virginia 126.05 (18.24) 1999 2000 No No Note: Cell entries provide data on the average per capita lottery expenditures and other factors that may influence lottery purchases. All data on lottery revenues come from the U.S. Bureau of Census, State Government Finances, series GF, No. 3 annual.

C.E. Landry, M.K. Price / Economics Letters 95 (2007) 451 455 453 Table 2 Per capita lotto expenditures in states that change designation Avg expenditure general fund Avg expenditure education Percent change Missouri 40.34 (1.23) 69.98 (17.27) 73.7% Montana 32.53 (1.97) 105.25 (164.19) 223.5% Oregon 166.33 (106.26) 427.81 (28.25) 157.2% Texas 134.63 (29.68) 143.26 (22.43) 6.4% Vermont 98.31 (24.58) 122.07 (3.8) 24.2% Virginia 123.71 (19.52) 136.59 (2.84) 10.4% Note: Cell entries provide average per capita expenditures, by designation, in states that switch the allocation of lottery proceeds from the state's general fund to education. The final column reports the percentage change in average expenditures generated by the switch. Standard errors are in parentheses. Cell entries can be read as follows: average per capita lotto expenditures in Missouri when proceeds were designated to the state's general fund (education) were $40.34 ($69.98). states that earmark. If bettors purchase lottery tickets solely for a love of gambling, we would expect casino gaming to crowd out expenditures in both general fund and earmark states. Finally, we find that per capita lottery expenditures are increasing in percentage of proceeds allocated to the public account. If the primary motivation for state lotto play was a love of gambling, variation in allocations to the public account should have no effect on expenditures. 2. Data and empirical results During the period 1990 2000, we observe twelve states in the U.S. that allocated lottery proceeds to the state's general fund, nine states that earmarked funds for education and six states that switched the allocation of proceeds between these categories. To test the importance of the link between lottery purchases and public good provision, we compare annual per capita lottery expenditures when proceeds are earmarked for education versus the state's general fund. Table 1 provides a state-level summary of our data. Average per capita lottery expenditures in states that designate lottery proceeds for education (the state's general fund) are $132.51 ($116.99) respectively, with this $15.52 difference statistically significant at the p b 0.05 level. Table 2 delineates average per capita expenditures by allocation in the six states that switch the designation of lotto proceeds. One clear data pattern noted in the table is a ubiquitous increase in annual expenditures when a state allocates lotto proceeds to education. To complement these unconditional insights, we estimate a series of linear regression models that explicitly control for observable and unobservable differences across states. Specifically, we estimate a linear regression model of the natural log of per capita lottery expenditures as: exp jt ¼ Z jt b þ a j þ e jt where exp jt is natural log of per capita lottery expenditures for state j in year t, Z is a vector of characteristics including a dichotomous indicator for earmarking of lottery proceeds for education, and the α j 's are state fixed effects. To account for unobservable temporal heterogeneities, standard errors are clustered by year. Empirical estimates presented in Column 1 of Table 3 provide results consistent with the unconditional insights: per capita lottery expenditures are approximately 60% higher in states that earmark lottery proceeds ð1þ

454 C.E. Landry, M.K. Price / Economics Letters 95 (2007) 451 455 Table 3 Fixed effects regression models per capita lotto expenditures Model A Model B Model C Model D earmark Model E general fund Earmark indicator 0.47 (0.12) 0.44 (0.12) 0.33 (0.12) % to prizes 0.04 (0.01) 0.04 (0.01) 0.03 (0.01) 0.06 (0.02) 0.02 (0.01) % to public account 0.05 (0.01) 0.05 (0.01) 0.04 (0.01) 0.05 (0.02) 0.03 (0.01) Casino gambling 0.09 (0.03) 0.09 (0.05) 0.03 (0.03) 0.16 (0.07) Multi-state lottery 0.33 (0.06) 0.19 (0.06) 0.004 (0.11) 0.20 (0.07) Per capita income 0.00005 (0.00001) 0.00004 (0.00001) 0.0001 (0.00001) Population 0.0002 (0.00004) 0.0001 (0.00004) 0.0002 (0.0002) # of obs 277 277 277 122 155 R-squared 0.99 0.99 0.99 0.99 0.99 Denotes statistical significance at the p b 0.05 level. Note: Cell entries provide parameter estimates for a fixed effects model of annual per capita lottery expenditures. Standard errors are in parentheses and have been clustered around individual year effects. for education. 1 Further, a 1% increase in the percentage of lottery revenues allocated to the public account leads to a 5% increase in expenditures. Combined, these results suggest the importance of linking lottery purchases with public good provision. If the sole motivation for state lottery play was the wagering opportunity, per capita expenditures should be independent of the designation and magnitude of proceeds allocated to state budgets. Our data set is sufficiently rich to determine whether lotto play is motivated by a love of gambling by measuring the impact of casino gaming (a substitute form of gambling) on lotto purchases. Model B in Table 3 provides evidence that casino gaming is a substitute for state lotto play. The presence of legalized casino gambling generates an approximate 9.4% reduction in per capita lotto expenditures with this difference significant at the p b 0.05 level. However, this substitution is incomplete: it only offsets a portion of the approximate 55% expected gain associated with earmarking. Empirical estimates in Model C highlight that this crowding result is robust to alternate specifications that include state demographics such as per capita income and population. Exploring this result a level deeper, Models D and E highlight an important asymmetry in the impacts of casino gaming. In general fund states, the presence of legal casinos is associated with an approximate 17% reduction in annual per capita lotto expenditures with this difference significant at the p b 0.05 level. For states that earmark, the introduction of casino gaming generates a reduction in per capita lottery expenditures of only 3%. However, this reduction is not significant at any meaningful level. If state lotto purchases are driven by a love of gambling, casino gaming should crowd out expenditures in both general fund and earmark states equally. Hence, our results are at odds with such alternate theories. 3. Conclusions This study provides empirical evidence from state lotto play that allows us to discriminate between alternate theories for lottery purchases. Empirical results suggest per capita lotto expenditures are greater when proceeds are earmarked for education. Further, we find that casino gambling is a substitute for state 1 We calculate the marginal effects for indicator variables by 100{exp(β) 1} as suggested in Halvorsen and Palmquist (1980).

C.E. Landry, M.K. Price / Economics Letters 95 (2007) 451 455 455 lotto play in general fund states but has no effect in states that earmark. Finally, we find that per capita lottery expenditures respond to variations in the allocation of proceeds to the public good. Combined, these results suggest wagering individuals respond to attendant public good provision when purchasing state lotto tickets. References Halvorsen, R., Palmquist, R., 1980. The interpretation of dummy variables in semilogarithmic equations. American Economic Review 70, 474 475. Landry, C.E., Lange, A., List, J.A., Price, M.K., Rupp, N.G., 2006. Toward an understanding of the economics of charity: evidence from a field experiment. Quarterly Journal of Economics 121, 747 782. Lange, A., List, J.A., Price, M.K., 2007. Using lotteries to finance public goods: theory and experimental evidence. International Economic Review. Morgan, J., 2000. Financing public goods by means of lotteries. Review of Economic Studies 67, 761 784. Morgan, J., Sefton, M., 2000. Funding public goods with lotteries: experimental evidence. Review of Economic Studies 67, 785 810.