13

be expected if casino gross revenues came from gamblers outside the area, all hiring took place from among the pool of local unemployed, and, except for the 15 percent of gross revenues mandated for state taxes, all revenues were spent locally. A net export multiplier of 2 was used to show the predicted number of direct and induced jobs. These are shown as the descending solid lines starting at the month of casino opening." Figure 4, part (b) shows the same information for casinos in the northern part of the

?

state. The date of casino opening in each market is easily noted by the point where the descending solid
line begins.

What does this data tell us? With the possible exception of Alton, there is no discernible pattern to
the data after casino opening compared to before, even though the plain solid lines show that the casinos
were capable of making a noticeable impact on the job market. In the case of Metropolis in Massac
County, for example, total county employment was 5,411 on the month before opening, and the casino
itself hired nearly 800.

Several possibilities can explain the fact that little or no impact on unemployment can be seen. One
is that most hiring did not take place from the pool of local unemployed: If jobs were taken by people
from outside the area or by individuals who left other jobs for casino employment, we would not observe
a reduction in the number of unemployed. Another explanation might be that the casinos attracted large
sums of money to the communities involved, but also removed large sums of money so that the local net
expenditure increase was negligible. A combination of these explanations also might be responsible.

In the case of Alton we note that before introduction of the casino, unemployment looked much like
state unemployment, sometimes greater, sometimes smaller than the state level. After introduction of the
casino, however, Alton unemployment stays uniformly below the state benchmark. Although the
reduction does not match the potential drop in unemployment, the consistent movement in one direction
signals an effect for further investigation. We will quantify this effect below, after we look at the
relationship between casinos and employment.

Months

Months

Months

10000

Months

14

B. Employment

Figure 5 charts the same information applied to the number of people employed. As before, the
figures show employment in the local labor market relative to state employment scaled to be the same
size as the local market on the month before casino opening.

We refer first to Aurora and Joliet in part (b). Casinos opened in Joliet in June 1992 (month 66) and
in June 1993 (month 78) in Aurora. Both are communities near the Chicago suburbs with extremely
profitable casinos. In addition, both are frequently cited by the gambling industry as examples of
economic development. For both locations there is no noticeable effect of casino opening. If anything,
both areas had growing employment relative to the state that seems to have leveled out when the casinos
were opened (Joliet) or eighteen months before (Aurora). Because both Joliet and Aurora were growing
before and after casino introduction, it is important to distinguish the effect of the casino from growth
that would have happened anyway. The great size of the potential employment increase from the casinos
in Joliet reflects the presence of the large Chicago market nearby from which the casino draws its
revenues and which gives the casino its tremendous profitability.

Figure 5 presents a similar story for Alton, Peoria, Rock Island, and E. St. Louis (St. Clair County)
farther away from Chicago. The figure shows that Peoria employment was growing relative to the state
before the introduction of its casino in November 1991 (month 59) while Rock Island's and E. St. Louis'
employment were declining. In the case of Peoria, growth relative to the state appears to stop at about
the same time the riverboat was introduced while the trend in Rock Island and E. St. Louis appear to have
abated before the casino. In all four cases there is no obvious positive employment impact of the casinos
starting with the month of opening. This is consistent with the observation that a large majority of the
gamblers in Peoria were from Illinois and for Rock Island with the fact that Iowa had competing casino
riverboats.

34

FIGURE 4: POTENTIAL AND ACTUAL UNEMPLOYMENT EXPERIENCE RELATIVE TO STATE BASELINE

15

In contrast to the other cities, E. Dubuque (Jo Davies County) and Metropolis (Massac County)

appear to show a positive impact of the casinos on employment. In both cases, employment is noticeably
higher after introduction of the casinos than before, though the effect in each case is a fraction of the
potential impact. This is consistent with what one would expect if substantial casino profits and non-
payroll expenditures are spent outside the area in question. Large inflows matched by nearly as large
outflows means that the casino will have a smaller impact on the economy.

C. Regression Evidence

Figures 4 and 5 show that casinos had the potential in principle to reduce unemployment, increase
employment, or both, but in practice had little or no effect except in one or two cases.

To quantify the impact, linear regressions were run to separate the statistical and economic effects of
casinos on unemployment and employment. Seasonally unadjusted county employment and
unemployment data were collected starting in October 1987 and ending in July 1994. With two month
lags in variables, this allowed the effective sample to begin in January 1988, 46 months before the first
Illinois casino began operation in September 1991. From September 1991 to July 1994 eight more boats
covered by the sample began operation.

For each casino, the data captures at least three years before and one or more years after the boat
began operation. The degree of riverboat activity was measured by real monthly revenues collected,
adjusted by the monthly consumer price index CPI-U. Regressions were run for all riverboats in Illinois
that entered before July 1994. Two regressions were run for each region: one for employment and one
for unemployment for a total of sixteen regressions.

For each regression, the final analysis was based on 90 observations and 82 degrees of freedom. To
allow for market dynamics, the number of people employed was represented by two lags of past
employment, current and two lags of state employment, and the size of casino revenues collected. A

similar regression was run for the number of people unemployed.

16

Table 5 reports the results of the regressions. Adjusted R2 are high in most cases, ranging from .99 to

.88 for employment and from .95 to .71 in the case of unemployment. The only exception is the
regression for unemployment in Jo Davies County, where the adjusted R2 was a relatively low 561. The
Durbin-Watson coefficients are around 2 as they should be for autocorrelation to be absent.35

Casino riverboat sales account for only part of economic activity in the counties in question. Using
lagged dependent and state explanatory variables to factor out the general trends maximizes the
likelihood of identifying a significant relationship between the casino revenue variable and the level of
the economic activity. Even though Figure 5 shows that the potential impacts are large relative to the
difference in the state and local trends, the regressions verify that the extent to which the variation in
local employment was explained by the casino is small.

Lagged local employment, state employment, and lagged state employment all clearly play an
important role in explaining local employment. The coefficients on lagged employment, state
employment, and lagged state employment were significant at the 5 percent level or better in every
regression. The relevant t-statistics ranged from 2.15 to over 27. Twice lagged state employment was
statistically significant in only one regression, suggesting that two lags are sufficient explanatory
variables.

In only two cases do the regressions reveal a significant relationship between the presence of casinos
and employment. The t-statistic for the casino revenue variable in East Dubuque (Jo Davies County) was
3.64 and was 4.89 for Metropolis (Massac County). Both are areas that showed a positive impact in
Figure 5. No regressions showed a significant impact of casinos on reducing unemployment except
Alton, an area that showed a probable, though small, impact in Figure 5. The Alton riverboat was the
first boat introduced, beginning operation in the aftermath of the 1990-91 recession. This may help to
explain its unique position as the only location where casino operation reduced the number of
unemployed. Interestingly, the employment regression for Alton did not show a significant impact of the

Madison County

35 $5 35 35 35 53 33 35 35 33 33 33 55 53 33

Notes: Statistically insignificant coefficients are shaded.