UNR Economics Working Paper Series Working Paper No

Transcription

1 UNR Economics Working Paper Series Working Paper No Riverboat Casino Gambling Impacts on Employment and Income in Host and Surrounding Counties Karl R. Geisler, and Mark W. Nichols Department of Economics /0030 University of Nevada, Reno Reno, NV (775) Fax (775) July, 2013 Abstract: This paper uses crowd-sourced transaction data from a cross-section of the US to examine demand for marijuana. State and regional variations in consumption, price and quality are explored. Price elasticity of demand estimates range between -0.3 and Noticeable price differences are found between high, medium, and low quality marijuana, with high quality marijuana, at $13.57 per gram, 144% greater than low quality marijuana, at $5.55 a gram. Significant variation is also found by medical marijuana status and census region, although this variation depends critically on the quality of the marijuana. JEL Classifications: R11, R12, L83 Key words: Casino Gambling, Economic Impact, Spatial

2 Riverboat Casino Gambling Impacts on Employment and Income in Host and Surrounding Counties Karl R. Geisler Department of Economics University of Nevada, Reno Mark W. Nichols Department of Economics University of Nevada, Reno JEL Classifications: R11, R12, L83 KEYWORDS: Casino Gambling, Economic Impact, Spatial 1

3 Riverboat Casino Gambling Impacts on Employment and Income in Host and Surrounding Counties Abstract This paper explores how the opening of casinos in riverboat states (Illinois, Indiana, Iowa, Louisiana, Mississippi, and Missouri) in the mid-1990 s impacted county income and employment. Building on previous literature, this study extends the analysis beyond the county where the casino opens to explore the effects experienced by neighboring counties. Real per capita county income is found to increase in counties where casinos open while unemployment drops. Neighboring counties also experience significant changes in these variables, though to a lesser magnitude. Where casinos are opened adjacent to other counties that also have casinos, a competition effect is found that reduces the impacts below what either county would have experienced without competition. Neighboring counties that both have casinos experience increases in income and decreases in unemployment similar to counties without casinos that are adjacent to casino counties. 2

4 I. Introduction Over the last two decades, many jurisdictions have approved casino gambling with the intent to stimulate their local economies. This trend has continued, even as the gains to communities that result from opening casinos remain disputed. While the arguments for and against casinos take many shapes and forms, from cultural acceptability to site suitability, most of the claims made by economists revolve around the economic impacts of casino openings where none had previously existed. 1 As shown by Figures 1 and 2, the prevalence of casinos in the United States has dramatically increased from 1990 to One of the main arguments against using gambling as an economic development tool lies in the claim that casinos cannibalize other local businesses, resulting only in a redistribution of existing spending rather than adding production to the local economy. This idea that casinos change the mix of production and output while failing to stimulate new economic growth or improve the quality of life of those who live near the casino is a theme that rises time and again, especially with respect to non-tribal casinos. 3 Economic development theory agrees that this sort of re-allocation of resources cannot properly be classified as true growth. For economic growth, an increase in local production needs to occur. This can come either by producing a good for export, or by producing a good locally that substitutes for a good that had previously been imported (Felsenstein et al., 1999). In the case of casinos, this means attracting gamblers from other jurisdictions or retaining local gamblers to spend money locally rather than in other jurisdictions (Eadington, 1995). Riverboat states are used in this analysis for many reasons, one of the most important being the timing of when they legalized gambling. These states all legalized commercial casino gambling over a four-year period, and they all opened their first 1 Eadington (1995) advocates policy makers to consider more than economic factors when choosing whether or not to allow casinos, pointing out that if those economic benefits dissipate the community will still have to deal with problems associated with the casino. 2 Figure 2 is effective as of January 1, 2012 which is why Ohio, which opened casinos later that year, is not identified as a casino state. 3 An example of how findings differ by casino owner is found in Wenz (2008), who finds an improvement in quality of life associated with Indian casinos but not commercial casinos. 3

5 casinos within a similarly short period (see Table 1). This timeframe has the advantage of being relatively recent but also long enough to allow for meaningful comparisons before and after the casinos open. Another advantage of analyzing riverboat states is that casinos were only allowed to open in limited locations, generally on or near a river or at an existing pari-mutuel racetrack. This limited introduction serves as a natural experiment, leaving many nearby counties without casinos to serve as valid economic controls while minimizing any biases that may be due to the selection of where casinos are located. Finally, riverboat casinos are representative of many jurisdictions that have recently legalized casino gambling in that they are primarily local or regional casinos, relying on import substitution as the driver of any economic growth. The rest of this paper proceeds as follows: section II reviews existing studies that examine the impact of casinos on employment and income; section III outlines the data and empirical model; section IV presents our empirical findings; and section V concludes. II. Literature Review Despite the controversy surrounding the economic impact of casino gambling, few academic papers on the subject have been published. In one of the earliest studies, Rephann et al. (1997) find positive employment and income effects, with riverboat casino counties experiencing larger employment effects than other casino counties. Evans and Topoleski (2002), using a difference-in-differences methodology similar to what is employed in our study, find that Indian casinos resulted in an increase in the number of jobs per adult and a 26 percent increase in tribal employment. Walker and Jackson (1998) and Walker and Jackson (2007) employ Granger causality analysis to casino revenue and per capita income. In the earlier study they use quarterly data over a 6 year period and find that casino revenues do Granger cause per capita income. In the latter study, they use annual data over a fifteen year period and find no relationship between casino revenue and per capita income. They attribute the different conclusions to differences in short-run versus long-run impacts. Garrett (2004) uses an ARIMA model for six counties to forecast household employment in the absence of casino gambling and compares this with actual household 4

6 employment after casinos opened. Garrett (2004) finds evidence of increases in employment when casinos are located in rural communities with the impact in urban communities much less pronounced. Wenz (2008), using difference-in-differences and propensity score matching, finds that casinos are associated with an increase in employment, but this increase is mainly attributed to Indian as opposed to commercial casinos. Wenz (2008) finds no significant increase in per capita income. Monchuck (2007) compares riverboat casino counties with other counties along waterways that were eligible to host a casino between 1995 and 2002 and finds that casino counties experienced slower growth in income but higher growth in employment. The above studies constitute the majority of published academic articles that examine the impact of casinos on employment and income. It is clear that no definitive conclusion can be made based on these studies. The current study focuses on riverboat states similar to Monchuk (2007) and Garrett (2004), and employs difference-indifferences similar to Evans and Topoleski (2002) and Wenz (2008). The current study expands on and contributes to this literature by examining a longer period (1984 to 2009, whereas Walker and Jackson (2007) only have data through 2005, and most of the other studies above end in 2000 or 2002). In addition, this study examines a casino s impact not only in the host counties but also its adjacent neighboring counties, and analyzes how changes in income and employment occur over time. III. Data and Empirical Methodology This study uses data on real per capita income, the unemployment rate, and labor force participation rate for all counties in Illinois, Indiana, Iowa, Louisiana, Mississippi, and Missouri between the years 1984 and Real per capita income is expressed in 2007 dollars using the GDP deflator. Since casinos began operating in these states between 1991 and 1995, our sample period should adequately capture pre- and postcasino trends. Figure 3 illustrates counties where a casino had operated as of

7 To estimate the impact that a casino opening had on real per capita income, the unemployment rate, and the labor force participation rate the following difference-indifferences (DID) model is estimated: Y i,t = β 0 + β 1 CasinoLegal t + β 2 CasinoCounty i + β 3 CasinoOpen i,t + γ 1 + γ 2 + ε i,t (1) Y i,t is the dependent variable and is either real per capita income, the unemployment rate, or the labor force participation rate for county i at time t. CasinoLegal t is dummy variable that indicates when state legislation was enacted to allow commercial casino gambling. This variable is used in the DID framework to capture any changes in Y i,t following casino legalization that would affect casino and non-casino counties alike. CasinoCounty i is a dummy variable that has a value of one for any county that had a casino at any time over the period of observation, separating the treated counties from the control counties. This variable accounts for any differences between casino counties and non-casino counties irrespective of whether a casino is actually open. CasinoOpen i,t is a dummy variable equal to one in the year the first casino opened in that county, and was compiled from various gaming control boards and through manual phone and internet searches. This variable is the treatment in the DID framework and captures the impact of the opening of the casino. 4 The vector γ 1 is made up of demographic variables that account for the shares of the population that are 65 and older ( Share65 i,t ), male ( ShareMale i,t ), and white ( ShareWhite i,t ). The vector γ 2 contains county and year fixed effects, which are not separately reported. Population and county demographic data are come from the US Census, county income data are from the Bureau of Economic Analysis, and The Bureau of Labor Statistics is the source for both the unemployment and labor force participation data. 5 Descriptive statistics for our key variables are summarized in Table 2. 4 Since casinos opened in different counties in different years, CasinoOpen i,t is our treatment variable rather than the interaction of CasinoLegal and CasinoCounty, as done in a standard DID model. See Imbens and Wooldridge (2009) for a full discussion. 5 Due to a change in how data were collected, the unemployment and labor force participation data used in this study begin in

8 IV. Empirical Results A. Host County Impacts The simple DID model given by equation (1) is estimated for each of the three dependent variables to find the impacts on the host counties where the casinos opened. These results are reported in Table 3 and robust standard errors are used to account for heteroskedasticity. CasinoLegal t accounts for changes that occurred after casino legalization for all counties, irrespective of whether a casino existed. Real per capita income, unemployment rates, and labor force participation rates were all experiencing downward trends. As shown by the coefficient on CasinoCounty i, casino counties, on average, had higher incomes, higher labor force participation rates, and lower unemployment rates than non-casino counties, though only the first two of these are statistically significant. Casino counties had over 10% more income, on average, than the average non-casino county. 6 The coefficient on CasinoOpen i,t, the variable of most interest, demonstrates that per capita county income increased $917 (3.9%) as compared to non-casino counties, the unemployment rate drops a half a percentage point, and the labor force participation rate increases by one percent. The results for income support the findings of Rephann (1997) and Walker and Jackson (1998), who found a positive association between casinos and income, but contrast with Wenz (2008) and Walker and Jackson (2007) who found no significant impacts on income. The increase in the labor force participation rate, coupled with the decrease in the unemployment rate, suggests that introducing a new commercial casino has a net positive effect on jobs in the county rather than simply cannibalizing jobs from existing firms. This does not imply that no cannibalization takes place, but rather that any cannibalization that does occur is more than offset by the gains from the new casino. The results for unemployment and labor force participation are consistent with Garrett (2004) and Wenz (2008) who also found positive employment impacts associated with casinos. 6 Recall from Table 2 that the average county income in the sample was $23,772, with non-casino counties averaging $23,548 and casino counties averaging $25,935. 7

9 B. Impacts on Neighboring Counties The analysis above demonstrated increases in per capita income and employment in the host counties. Here we present results for neighboring counties to determine whether the impact of casinos on per capita income, unemployment, and the labor force participation rate spread beyond their host communities. Casinos may employ individuals from nearby counties in addition to those who live in the county in where the casino is located. The importation of labor across county lines would mean that some of the gains in terms of lower unemployment and higher incomes may accrue to nonresidents of the host county. In their study of Native American casinos, Evans and Topoleski (2002) note how the importation of labor from nearby communities benefited the non-native population. Here we analyze whether the opening of commercial casinos benefits nearby (i.e., adjacent) counties. In order to analyze whether neighboring counties are impacted by the opening of a casino, two additional variables are added to the basic DID equation given in equation (1). NeighborCounty i is a dummy variable equal to one when a county is a neighbor to a casino county, zero otherwise, where neighbor counties are defined using queencontiguity. 7 NeighborCounty i identifies any difference between counties next to casino counties and those counties located further away. Figure 4 illustrates the sample of casino and neighbor counties. For each neighbor county, a second variable, NeighborOpen i,t, is a dummy variable equal to one when the casino opens in the adjacent casino county, zero otherwise. This captures the impact of the opening of the casino on the neighbor counties. The new equation for estimation becomes: Y i,t = β 0 + β 1 CasinoLegal t + β 2 CasinoCounty i + β 3 CasinoOpen i,t + β 4 NeighborCounty i + β 5 NeighborOpen i,t + γ 1 + γ 2 + ε i,t (2) Results from estimating equation (2) are reported in Table 4. Results for the host county are similar to those found above. Specifically, the coefficients on CasinoLegal t 7 Queen-contiguity means that any common border between two counties makes them neighbors, even if the shared border occurs only at a common corner point. Only 24 such borders occur in the 354 riverboat casino county borders. 8

10 are nearly identical to those reported in Table 3, and the two significant coefficients for CasinoCounty i keep the same signs, though for real per capita income the coefficient falls to the 10% level of significance. Similarly, as evidenced by the coefficient on CasinoOpen i,t, opening a casino increases per capita income in the host county by $800 (3.3%), reduces the unemployment rate by approximately half a percent, and increases the labor force participation rate by nearly a percent. All of these coefficients remain statistically significant at the 1% level. The coefficients on NeighborCounty i suggest that neighbor counties are statistically indistinguishable from non-neighbor, non-casino counties when it comes to real per capita income, unemployment rates, and labor force participation rates. The coefficient on NeighborOpen j,t suggests that neighbor counties also experience positive economic gains, although of a smaller magnitude than host counties. When a casino opens in an adjacent county, its non-casino neighbor counties experience average increases in real per capita income of $617 (2.6% more than non-neighbor, noncasino counties) and decreases in the unemployment rate by nearly two tenths of a percent, both significant at the 1% level. 8 Non-casino neighbor counties also experience an increase in the labor force participation rate of more than two tenths of a percent, though this is only significant at the 10% level. These findings support Garrett (2004), who hypothesized that casinos would employ workers from nearby counties. Rephann et al. (1997) also found evidence of jobs going to individuals residing outside of the host county. Working with tribal casinos, Ha and Ullmer (2007) found that casinos not only reduce tribal poverty, but that neighboring counties also experience increased economic activity. C. Impacts on Neighboring Casino Counties One of the motivations for states to legalize casino gambling is to minimize expenditures their citizens make in other states, thereby keeping gambling dollars instate. Some previous work has been done with regard to how casinos compete geographically, though the focus of this work has been on the impacts felt by casino 8 Non-casino, non-neighbor counties had an average real per capita income of $22,869. 9

11 patrons, the casinos themselves, and the response of regulation to increased competition. Consumers seem to benefit the most from increased competition, as casinos are pressured to offer more non-gaming amenities while increasing their payouts. Looking at a single metro area, Navin and Sullivan (2007) found an increase in the number of casinos resulted in a larger payout rate in the St. Louis market. Nichols (1998) found that commercial gambling revenue in Iowa fell as a result of the introduction of riverboat casinos in neighboring Illinois, where some of the new casinos in Illinois opened directly across the river from existing casinos in Iowa. Illinois, in turn, experienced a decline in revenue as a result of increased competition in Indiana and Missouri. More recently, declines in casino revenue in Atlantic City have been attributed to expansion of casinos in Pennsylvania, as noted by McGowan (2009). As indicated by Figure 5, casinos in most states are located on state borders, with Iowa being the main exception. As also shown in Figure 5, this has resulted in some casino counties being adjacent to other casino counties. This raises an obvious question. Does opening a casino in a county adjacent to an existing casino have the same impact as opening a casino without any competitors nearby? To answer this question, our basic DID analysis is expanded to include the dummy variable NeighborCasinoCountyOpen i,t, which has a value of one starting in the year a casino opens in a county that is adjacent to neighbor county with an open casino, zero otherwise. This allows for geographically isolated casino counties to experience different returns from casino counties near other casino counties. The DID model to be estimated now becomes: Y i,t = β 0 + β 1 CasinoLegal t + β 2 CasinoCounty i + β 3 CasinoOpen i,t + β 4 NeighborCounty i + β 5 NeighborOpen i,t + β 6 NeighborCasinoCountyOpen i,t (3) + γ 1 + γ 2 + ε i,t Results from estimating equation (3) are presented in Table 5. Coefficients on CasinoLegal t, CasinoCounty i, CasinoOpen i,t, NeighborCounty i, and NeighborOpen i,t are similar to those reported in Table 4 based on the estimation of equation (2). The interpretation of the coefficient on the new variable, NeighborCasinoCountyOpen i,t, is the 10

12 estimate of the additional impact of opening a casino when another casino is open in an adjacent county. Thus, the impact of opening a casino for geographically isolated casino counties is β 3, the coefficient on CasinoOpen i,t, whereas the impact of opening a casino next to another casino county is β 3, the coefficient on CasinoOpen i,t, plus β 6, the coefficient on NeighborCasinoCountyOpen i,t. 9 The results in Table 5 suggest that when a casino opens in a county with no competing casinos nearby per capita county income rises by $956. In contrast, when opening in a county that has a casino operating in an adjacent county the impact on per capita income is reduced by $350 dollars, to $606. This value is statistically indistinguishable from the $656 increase in per capita county income experienced by non-casino neighbor counties, calling into question the argument for casino expansion as a means to increase local income if a casino is already operating in an adjacent county. The lack of significant explanatory power in the unemployment rate and labor force participation rate estimations of NeighborCountyCasinoOpen i,t provide further support. With no further reduction in unemployment rates or gains in labor force participation rates, communities without a commercial casino, but with a commercial casino in an adjacent community (county in this case), may want to think twice before allowing one, as any income and employment impacts may have already been realized. Of course, income and employment are not the only reasons that states legalize casinos and the fight over tax revenues is often a key driver. These results do suggest, however, that the spatial distribution of casinos is an important determinant of the overall economic impact of a casino. V. Timing of Impacts One drawback to the DID approach is that it does not allow for the effects of the structural change to fluctuate over time. While the change could be a static improvement 9 NeighborCasinoCountyOpen i,t is essentially an interaction between CasinoOpen i,t, NeighborCounty i,t, and CasinoCounty i,t and is a dummy variable identifying if the neighbor county is also a casino county with an open casino. Thus, the change in Y i,t from opening a casino, Y i,t / CasinoOpen i,t is β 3 + β 6. We don t account for a neighbor casino county with a separate dummy variable in the regression because it is perfectly collinear with NeighborCounty i,t and CasinoCounty i since a neighbor casino county must, by definition, be both. 11

13 in the local economy, it is also possible that the impact of opening a casino changes over time or occurs with a lag. It could be that a new casino spurs other growth by instigating a virtuous cycle of economic development. It is also possible that the positive impacts of a casino opening could erode over time as the novelty wears off or as competition increases in nearby jurisdictions. One approach to determining if short-run variation exists in a casino s impact lies in creating lead and lag indicator variables around the casino opening. 10 Since casinos opened in different years in different counties, these values will vary by county and by year, though non-casino counties will have zeros for all of these variables. In order to capture changes in the values of per capita income, unemployment, and labor force participation, the following equation is estimated: Y i,t = β 0 + β 1 Lead2Open i,t + β 2 LeadOpen i,t + β 3 CasinoOpen i,t + β 4 LagOpen i,t + (4) β 5 Lag2Open i,t + β 6 Lag3Open i,t + β 7 Lag4Open i,t + β 8 Lag5Open i,t + γ 1 + γ 2 + ε The two lead variables are included to account for any difference between casino and non-casino counties prior to the casino opening. CasinoOpen i,t corresponds to the year the casino initially opens. The various lag variables allow for variation in the dependent variable up to and including five years after the initial opening. Results from estimating equation (4) are presented in Table 6. The timing of the impacts shown in Table 6 help to validate the DID framework used in the above sections while providing additional evidence as to how these changes take place over time. Initially, the only significant impact of a new casino is an increase in the labor force participation rate, though by the second year after the casino opens statistically significant changes occur in real per capita county income and unemployment rates. These changes grow over the next few years, and at five years after the casino opens statistically significant changes have occurred for all of the dependent variables: per capita county income increases, unemployment rates fall by about half a percent, and the labor force participation rate has increased by about a percent. This shows that the long-run impacts found in the DID 10 In the DID framework this is identical to including CasinoOpen*t, where t represents the number of years a casino has been open. See Eissa and Liebman (1996) for an applied example. 12

14 framework do not occur immediately upon the casino opening. Instead, the gains to the local economy build to the levels found in the previous section over a number of years. VI. Conclusion This study finds that communities where new commercial casinos open experience positive changes in key economic indicators. These counties see real per capita income increase by $800 to $900, the unemployment rate decrease by approximately half a percent and the labor force participation rate increase roughly one percent. These changes do no occur immediately, but rather, occur over time, taking up to five years to be fully realized. The economic gains, however, are not limited to the host counties. Neighboring counties that do not have casinos experience positive externalities, with increases in per capita income of more than $600, decreases in their unemployment rates by approximately two tenths of a percent, and increases in their labor force participation rates by nearly that same amount. These findings partially support Garrett (2004), who proposed that in order to find enough workers with the necessary skills casinos would likely have to hire workers from outside of the area, leaving local unemployment rates unchanged. Here we find evidence that workers are drawn from outside the area, as evidenced by the decreasing unemployment rate in surrounding counties, but also find a decrease in the unemployment rate in the host county. 11 Competition from other casinos, however, diminishes the economic impact for the host community. When casinos open in counties next to other casino counties, the effects of competition reduce the impacts of having a casino to approximately the levels experienced by non-casino counties that border casino counties. This suggests that 11 This conclusion obviously depends on how outside the area is defined. Here we define outside the area as a neighboring county. If outside the area were defined as outside the city where the casino is located, it is possible that unemployment in the city remains unchanged while the county unemployment falls if workers live in the county but not in the city. Our data do not allow this distinction. 13

15 planning at the regional level could lead to more efficient outcomes than the current system that encourages inter-state competition. The finding that neighboring casino counties have less of an impact on income and unemployment than more geographically isolated casinos supports the importance of opening a casino aimed at exporting the gaming experience in order to get the most benefit. Where casinos are opened primarily to serve as mechanisms for import substitution, fewer gains occur. The gains demonstrated herein on key economic indicators are just the beginning of an understanding as to how casinos impact host and surrounding communities. Policy makers also need to consider potential downsides to permitting new casinos. As evidenced here, this should be done for host and surrounding communities. Using the framework established in this study to explore how problem gambling, for example, changes in non-casino neighbor counties could lead to better guidance for policy makers considering casinos as economic development tools. This approach could also lead to better regional planning when it comes to the location of casinos. As found above, the benefits to casino counties drop when neighbors also open casinos, and the new casino county does not gain statistically higher income per capita or lower unemployment rates than they would have otherwise experienced. The placement of casinos is a critical component of their overall economic impact and suggests that the tit-for-tat strategy of casino expansion adopted by states over the last few years is likely not socially optimal. 14

16 References Calcagno, Peter, Douglas Walker, and John Jackson "Determinants of the probability and timing of commercial casino legalization in the United States," Public Choice 142(1): Eadington, William R Economic development and the introduction of casinos: Myths and realities. Economic Development Review 13(4): Evans, William N., and Julie H. Topoleski The Social and Economic Impact of Native American Casinos. NBER Working Paper No Eissa, Nada, and Jeffrey B. Liebman Labor Supply Response to the Earned Income Tax Credit. Quarterly Journal of Economics 111(2): Felsenstein, Daniel, Laura Littlepage, and Drew Klacik Casino Gambling as Local Growth Generation: Playing the Economic Development Game in Reverse? Journal of Urban Affairs 21(4): Garrett, Thomas A Casino Gaming and Local Employment Trends. Federal Reserve Bank of St. Louis Review 86(1): Ha, Inhyuck Steve, and James Ullmer "The economic effects of Harrah's Cherokee casino and hotel on the regional economy of Western North Carolina." Journal of Economics and Economic Education Research 8(2): Imbens, Guido W., and Jeffrey M. Wooldridge "Recent Developments in the Econometrics of Program Evaluation." Journal of Economic Literature 47(1): McGowan, Richard The Competition for Gaming Revenue: Pennsylvania v. New Jersey. Gaming Law Review and Economics 13(2): Monchuk, Daniel C. (2007). People Rush in, Empty Their Pockets, and Scuttle Out: Economic Impact of Gambling on the Waterways. The Journal of Regional Analysis and Policy, 37(3): Navin, John C., and Timothy S. Sullivan "Do Riverboat Casinos Act as Competitors? A Look at the St. Louis Market." Economic Development Quarterly 21(1): Nichols, Mark W Deregulation and Cross-Border Substitution in Iowa s Riverboat Gambling Industry. Journal of Gambling Studies 14(2): Rephann, Terance J., Margaret Dalton, Anthony Stair, and Andrew Isserman "Casino gambling as an economic development strategy." Tourism Economics 3(2): Walker, Douglas, M. and Jackson, John, D. (1998). New Goods and Economic Growth: Evidence from Legalized Gambling. Review of Regional Studies 28(2): Walker, Douglas M. and Jackson, John D. (2007). Do Casinos Cause Economic Growth? American Journal of Economics and Sociology 66(3): Wenz, Michael "Matching estimation, casino gambling and the quality of life." The Annals of Regional Science 42(1):

17 16

18 17

19 18

20 19

21 Table 1: Timeline of Riverboat Casinos Legalizations and Intial Openings Year Legislation Passed Month and Year of First Casino Opening State Iowa 1989 April 1991 Illinois 1990 Sept Mississippi 1990 Aug Louisiana 1991 Oct Missouri 1993 May 1994 Indiana 1993 Dec Adapted from Calcagno, Walker & Jackson (2010) Variable Table 2: Descriptive Statistics of Key Variables Full Sample Casino Counties Non-Casino Counties Real Per Capita Income 23, , , ( ) ( ) ( ) Unemployment Rate ( ) ( ) ( ) Labor Force Participation Rate ( ) ( ) ( ) Share ( ) ( ) ( ) ShareMale ( ) ( ) ( ) ShareWhite ( ) ( ) ( ) Means for key varibles, with standard deviations in parentheses, for the full sample, casino counties, and non-casino counties. 20

22 Table 3: Simple Difference in Difference with County Fixed Effects Independent Variables Real Per Capita Income Dependent Variable Unemployment Rate Labor Force Participation Rate CasinoLegal *** *** * ( ) ( ) ( ) CasinoCounty *** *** ( ) ( ) ( ) CasinoOpen *** *** *** ( ) ( ) ( ) Share *** *** ( ) ( ) ( ) ShareMale *** *** *** ( ) ( ) ( ) ShareWhite *** ( ) ( ) ( ) Constant *** *** *** ( ) ( ) ( ) County Dummies? Yes Yes Yes Year Dummies? Yes Yes Yes R n = n = n = Significance indicated by * at the 10% level, ** at the 5% level, and *** at the 1% level. 21

23 Table 4: Difference in Difference with Neighboring County Effects Dependent Variable Independent Variables Real Per Capita Income Unemployment Rate Labor Force Participation Rate CasinoLegal *** *** * ( ) ( ) ( ) CasinoCounty * *** ( ) ( ) ( ) NeighborCounty ( ) ( ) ( ) CasinoOpen *** *** *** ( ) ( ) ( ) NeighborOpen *** *** * ( ) ( ) ( ) Share *** *** ( ) ( ) ( ) ShareMale *** *** *** ( ) ( ) ( ) ShareWhite * *** ( ) ( ) ( ) Constant *** *** *** ( ) ( ) ( ) County Dummies? Yes Yes Yes Year Dummies? Yes Yes Yes R n = n = n = Significance indicated by * at the 10% level, ** at the 5% level, and *** at the 1% level. 22

24 Table 5: Difference in Difference with Neighboring County Casino Effects Dependent Variable Real Per Capita Independent Variables Income Unemployment Rate Labor Force Participation Rate CasinoLegal *** *** * ( ) ( ) ( ) CasinoCounty * *** ( ( ) ( ) NeighborCounty ( ) ( ) ( ) CasinoOpen *** *** *** ( ) ( ) ( ) NeighborOpen *** *** ( ) ( ) ( ) NeighborCasinoCountyOpen *** ( ) ( ) ( ) Share *** *** ( ) ( ) ( ) ShareMale *** *** *** ( ) ( ) ( ) ShareWhite * *** ( ) ( ) ( ) Constant *** *** *** ( ) ( ) ( ) County Dummies? Yes Yes Yes Year Dummies? Yes Yes Yes R n = n = n = Significance indicated by * at the 10% level, ** at the 5% level, and *** at the 1% level. 23

25 Table 6: Testing for Timing of Impacts from Casino Openings Dependent Variable Real Per Capita Independent Variables Income Unemployment Rate Labor Force Participation Rate Lead2Open ** ( ) ( ) ( ) LeadOpen ( ) ( ) ( ) CasinoOpen * ( ) ( ) ( ) LagOpen *** ( ) ( ) ( ) Lag2Open * ** *** ( ) ( ) ( ) Lag3Open *** *** ( ) ( ) ( ) Lag4Open ** ** *** ( ) ( ) ( ) Lag5Open *** *** *** ( ) ( ) ( ) Share *** *** ( ) ( ) ( ) ShareMale *** *** *** ( ) ( ) ( ) ShareWhite *** ( ) ( ) ( ) Constant *** *** *** ( ) ( ) ( ) County Dummies? Yes Yes Yes Year Dummies? Yes Yes Yes R n = n = n = Significance indicated by * at the 10% level, ** at the 5% level, and *** at the 1% level. 24