The analyses described thus far in the report have focused on estimating the impacts of changes in lottery sales with respect to casino proximity. We have provided county-level sales data for casino-hosting and control counties, and we have used regression results to simulate the impact of a new casino within different travel times to a “representative” or “average” Maryland zip code.
We next use the econometric results presented in Tables 10 and 13 to provide specific monetary estimates of the aggregate dollar impacts of casinos on Maryland lottery sales during our sample period. First we provide estimates for the total annual impacts of all four existing casinos in Maryland. We provide these estimates in monetary and percentage terms relative to the most recent full year of sales (from March 2013 to February 2014, the end of our data sample).14 We calculated the estimates for each lottery game type separately and sum the games to arrive at a total estimated effect. Our estimate is based on the “simple” or aggregated model presented in Table 10. We calculate the Δ Gravity from the first to last month of the sample and multiply this change by the coefficient estimate for Δ Gravity from Table 10. This calculation yields the change in monthly lottery sales due to the changes in gravity occurring as Maryland casinos opened across the state. Table 14 shows the annual estimates. The overall impact is estimated to be -$40.3 million per year, or 2.31% of recent annual sales, as shown in the right-most column in the table. This can be interpreted to mean that Maryland lottery sales would have
14 To be precise, we should be performing these calculations based on lottery sales in the twelve months prior to each casino’s opening. That is unfortunately not practical here, since the four casinos opened at different times. Since lottery sales declined, our use of the most recent twelve months of sales (after most
been $40 million greater per year in the absence of the casinos. This is our estimate of the annual impact for 2013 and subsequent years, assuming no other significant changes occur in the market that would affect lottery sales.15
Table 14. Estimated annual impact of all Maryland casinos on lottery sales, by game type ($ millions)
Instant Monitor Pick Multi-St Other Total, All
Next we calculated a 95% confidence interval for our estimated effects. This calculation uses the standard errors of the regression coefficient estimates to calculate a range of figures within which we can be 95% certain that the actual value lies.16 This range for the total annual impact of Maryland casinos on lottery sales is between -$27.9 million to -$52.6 million.
Next we estimate the impact of each existing Maryland casino on lottery sales by type of game. To derive these estimates we use the coefficient estimates from Table 13. The incremental impact of each casino is calculated by adding the estimated Δ Gravity effect plus the incremental
15 In Appendix A we explain that the recent large Multi-State jackpots (MegaMillions and Powerball) may have propped up Multi-State sales. As a result, the actual negative impact of casinos is probably larger.
16 A more detailed explanation of the calculation and interpretation of confidence intervals can be found
impact on each casino, as shown in the estimates in Table 13.17 The resulting estimated impacts from each casino are shown in Table 15. We briefly discuss the results for each casino.
Table 15. Estimated annual impacts of individual Maryland casinos on lottery sales, by game type ($ millions)
Instant Monitor Pick Multi-St Other Total, All Games Actual Last Year Sales (3/2013 – 2/2014) $471.2 $472.5 $515.3 $223.2 $58.63 $1,740.8 Hollywood Casino Impact $3.10 ($1.60) ($0.53) ($2.84) ($0.38) ($2.25)
0.66% -0.34% -0.10% -1.27% -0.66% -0.13%
Ocean Downs Impact ($0.42) ($2.21) ($0.52) ($1.67) $0.18 ($4.80)
-0.09% -0.47% -0.10% -0.75% 0.03% -0.28%
Maryland Live Impact ($12.05) ($20.23) ($3.74) ($7.85) $2.77 ($41.09)
-2.56% -4.28% -0.73% -3.52% 4.73% -2.36%
Rocky Gap Impact $0.63 $1.06 $0.20 $1.00 $0.15 $3.04
0.13% 0.22% 0.04% 0.45% 0.25% 0.17%
Total Estimated Impacts ($8.74) ($22.97) ($4.59) ($11.36) $2.56 ($45.10)
-1.85% -4.86% -0.89% -5.09% 4.36% -2.59%
The Hollywood Casino is estimated to have a slightly positive impact on Instant sales in the state (0.66%). However, Hollywood does not seem to have had a statistically significant impact on either Monitor or Pick games. However, the table shows negative impacts on Multi-state and Other games types. The overall estimated effect of the Hollywood Casino is a loss in lottery sales of roughly $2.25 million per year, or 0.13% of total lottery sales in the state.
The results for Ocean Downs are similarly “mild.” This casino had no significant impact on Instant or Other games, but it had modestly negative impacts on Monitor, Pick, and
Multi-state games, so that the overall impact is a loss in lottery sales of $4.8 million per year, or 0.28%
of total State Lottery sales.
The Maryland Live casino is shown to have had the largest impacts on lottery sales. All of the estimated impacts are negative and significant, except for Other games, which is positive.
The largest negative impact has been on Monitor games; Maryland Live is estimated to have caused a reduction in sales of just over $20 million per year. The total negative impact of Maryland Live is estimated at around $41 million per year.
Lastly, Rocky Gap is estimated to have modestly positive impacts on lottery sales, to a tune of 0.17% of annual sales, roughly $3 million per year. In our earlier analysis, we also found a positive impact on lottery sales, and suggested that this might be due to the location of Rocky Gap, its attraction of visitors from other states, and perhaps its effects on employment in the local area.
When we sum the individual casinos’ effects from this analysis, the casinos are estimated to have caused a reduction in total lottery sales of about 2.6%, or $45 million ($41 million of which is attributable to Maryland Live). This is very close to the estimated -$40 million impact from the aggregate analysis presented in Table 14. The fact that these two estimates are very close, even though they are derived from different models, suggests that these estimates are indeed reliable.