SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A

As states become increasingly reliant on taxable casino revenues to augment their budgets, questions concerning optimal casino location have entered into policy dialogues across the country. Notably, policy makers have become concerned with the presence and size of so-called “cannibalization effects” within the casino industry whereby casinos operating within overlapping markets capture one another's business. However, the size, significance, and underlying mechanics of these effects have received very little attention in the academic literature. Using a unique data panel for the Illinois region that spans over a decade, this paper develops a working framework for identifying the presence of intra-industry cannibalization effects for the riverboat gaming industry. Evidence suggests cannibalization effects do indeed exist and are largely a function of new casino development, not the expansion of pre-existing casinos. These effects also attenuate rather quickly with distance.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A

Since 1989, states have become increasingly reliant on taxable casino revenues to augment their budgets. In 2010 alone, casino tax revenues reached $7.6 billion across the 22 states with commercial casinos.1 The financial crisis of 2007–2008 and subsequent recession have only enhanced states' desires to harness casino revenues, with many large states (including Ohio, Florida, and Illinois) having recently considered legislation to legalize or expand commercial casino gaming. As states debate these issues, questions concerning optimal casino location have entered into policy dialogues across the country. Notably, policy makers and analysts are often concerned about the presence and size of the so-called “cannibalization effects” whereby casinos with overlapping market areas partially capture one another's business. While some states hope to avoid placing new casinos too close to a pre-existing stock, others wish to siphon revenues away from neighboring states. Thus, knowledge of these effects can be especially relevant to the political, if not fiscal, viability of a project when several desirable locations exist and the number of casino licenses is limited. The size and significance of these effects, however, have received only modest attention in the literature.

Most research to date concerning cannibalization effects has focused on topics related to the displacement of general tax revenues (Walker and Jackson 2011) and/or inter-industry cannibalization effects (Anders, Siegel, and Yacoub 1998; Popp and Stehwien 2002; Siegal and Anders 1999). Within the gaming industry itself, the literature has identified the effects of casino development on other gambling industries such as lotteries and racing (Elliott and Navin 2002; Seigal and Anders 2001; Walker and Jackson 2008). However, with a few exceptions, little has been done to measure cannibalization effects within the casino industry itself. Walker and Jackson (2008) provided the most recent analysis of these effects, finding that states' commercial casino revenues are adversely impacted by the presence of casino gaming in adjacent states. An earlier study by Thalheimer and Ali (2003) also finds that wagering at a casino declines when the adjusted number of competitors rises. These papers provide clear evidence that cannibalization effects in the industry likely exist. However, much still remains to be done on this particular topic; notably, a clearer understanding of the various channels through which cannibalization effects arise is needed. And, more must be done to directly account for the uneven distribution of competition and demand across space. Efforts along these lines will help both policy makers and economists address some questions of particular interest. For instance, does the threat of competition attenuate slowly or rapidly as the distance between competitors rises? Are cannibalization effects most severe within a particular distance from a casino? Are they sensitive to only a particular type of increase in competition? This paper works to address these very questions.

Using a rich panel of data covering every commercial casino in or around Illinois between 1994 and 2006, the findings reported here suggest that cannibalization effects are quite large and attenuate rather quickly with the distance between competitors. Special attention is also paid to the two primary channels through which cannibalization effects arise: 1) the patronage effect, which occurs when competitors influence a casino's patronage levels, and 2) a spending effect, which occurs when a casino's spending-per-patron changes in response to competition. Cannibalization appears to materialize primarily through the patronage effect, particularly when new casinos develop, as opposed to when pre-existing casinos simply expand.

The Illinois-Centric Market Area

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A

The sample used in this study covers every commercial casino operating in or around Illinois between 1994 and 2006. This Illinois-centric region includes all “contiguous” commercial casinos where contiguity is defined by a 75-mile ring. Thus, after including all Illinois commercial casinos, all commercial casinos within 75 miles of an Illinois casino are included, and all commercial casinos within 75 miles of these casinos, and so on. The states represented in this region were some of the first to adopt commercial gambling outside of Nevada and New Jersey after passage of the Indian Gaming Regulatory Act in 1988. Today, this region still constitutes a major hub for U.S. casino activity, accounting for approximately 35 percent of U.S. commercial casino revenues and 33 percent of state-collected casino tax revenues outside of Nevada and New Jersey in 2010 (American Gaming Association 2011).

Figure 1 maps out the boundaries of this region for 1994 and 2006. Clear concentrations of casino activity exist in three large population centers (Chicago, St. Louis, and the Quad-Cities) and a disproportionate number of casinos operate along the Mississippi River. As shown in Table 1, the industry experienced substantial growth within this time frame, making it an appropriate area of study for observing the types of spatial interactions that result in cannibalization. Over this time, the number of casino locations in the region grew nearly 63 percent from 16 to 26, while overall capacity, measured by annual average number of floor positions, rose 206 percent from 11,944 to 36,526. This growth occurred primarily in the population centers along the Illinois border near Chicago, St. Louis, and Iowa, although mostly outside of Illinois proper. In addition, few tribal casinos operated within 75 miles of any of the region's commercial casinos, thus mitigating one of the major limiting factors to casino research; tribal casinos are not required to publically disclose information on revenues or admissions.2

figure

Figure 1. (a) The Illinois-Centric Region—1994. (b) The Illinois-Centric Region—2006.

Download figure to PowerPoint

Table 1. Commercial Casino within the Illinois-Centric Region: 1994–2006
StateYear legal (First Boat)19942006
No. of CasinosPositionsAdmissions (mil.)Revenue (mil.)No. of CasinosPositionsAdmissions (mil.)Revenue (mil.)
  1. Note: Two casino pairs, Indiana's Gary Majestic Star 1/Majestic Star 2 and Missouri's Maryland Heights Players/Harrahs, are treated as a single casino. The two Gary casinos are merged because they share the same dock and, beginning in 2002, share the same turnstile. Garrett and Pakko (2010) also treated these two casinos as a single entity. The Maryland Heights casinos merged operations in 2000. Prior to this merger they shared a building.

IL1990 (1991)108,86020.4980.6911,02816.21,923.5
IA1989 (1991)41,6662.793.496,7907.2410.3
IN1993 (1995)0000410,08612.91,297.7
MO1993 (1994)21,4182.651.248,62223.7727.5
Total 1611,94425.71,125.22636,52660.04,359.0

Conceptual and Empirical Framework

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A

For the purposes of this paper, cannibalization is defined as the adverse impact one casino's operations have on another's sales. Here, casino c's sales consist of two components:

  • display math

where nc = number of patrons and sc = average spending per patron. To account for the influence of other casinos, nc and sc are treated as functions of intra-industry competition, ϕ. Taking the natural logarithm and differentiating, yields:

  • display math(1)

which decomposes the total cannibalization effect, dln(salesc), into a patronage effect, dln(nc), and a spending effect, dln(sc). Thus, while it is hypothesized that competition reduces total sales due to lost patrons, this decline has the potential to be mitigated or aggravated based on the direction of the spending effect. The spending effect will be positive if cannibalized patrons spend less than the average, and negative if they spend more than average. No formal hypothesis regarding the direction or trajectory of the spending effect is made at this time.

To test for these effects, a fixed-effects model of the following general form is estimated:

  • display math(2)

where yct is a proxy for dln(salesc), ln(nc), or ln(sc) in year t. The vector of explanatory variables, Xct, describes characteristics specific to a casino's internal operations and external business environment, such as a casino's size, the level of market demand, and the proximity of competing casinos.

Casino-level data come from the annual and monthly gaming reports published by each state's regulatory agency (for Illinois, Indiana, Missouri, and Iowa). These agencies report monthly and annual data on a commercial casino's monthly and annual adjusted gross receipts (AGRs). A casino's AGR in year t, agrct, can be represented as agrct = nct × sct × pct, where pct is the fraction of every wagered dollar retained by the casino. This is often referred to as a casino's takeout (or hold) rate and is the effective price of gambling $1 (Navin and Sullivan 2007; Suits 1979; Thalheimer and Ali 2003). Annual admissions count, admct, is also reported and serves as a proxy for casino patronage.3 With these variables, an approximation of equation (2) is written as:

  • display math(3)

where inline image approximates pct using a casino's average takeout rate for electronic gaming devices (EGD) only. This decomposition allows for competition's impact on a casino's sales, estimated by inline image), to be fully disaggregated into the patronage and spending effects, ln(admct) and inline image, respectively.

Explanatory variables

A casino's effective size is measured using the variables monthsct and posct, which account for the months in a year a casino is operated and its number of floor positions, respectively. Floor positions are measured using inline image, where inline image and inline image are the annual average number of EGD and table games. This gives a relatively high weight to table games (i.e., blackjack, craps, etc.) in order to account for the larger number of players they can accommodate at any given time.4 Together, monthsct and posct influence a casino's annual capacity multiplicatively, as the annual impact of position count depends directly on the number of months a casino is open (and vice versa). Thus, to estimate equation (2) in linear form with ln(yct) as the dependent variable, the natural logarithm of monthsct and posct are used. It is anticipated that both of these variables will have a positive impact on sales.

The spatial distribution of market demand and competition also helps to explain the level of wagering at a casino. Locations of a certain size and income will have the most impact on wagering demand when there is only a short distance between them and a casino, while competitors will pose the greatest threat when they are comparatively close to significant sources of a casino's market demand.5 To account for these spatial interactions, locations within a casino's market area are first assigned weights based on their relative size and distance from the casino, whereby larger weights indicate a greater level of significance (all else being equal). For variables that measure population, income, or competition, their respective distributions vis-à-vis these location weights will determine their effective importance to a particular casino. Using zip codes to define locations, weights are calculated as inline image, where dkc is the distance between zip code k's centroid and casino c's address, and sckt is k's share of total employment within c's market area. wckt ranges from zero to one and declines in dkc to account for the attenuative effects of distance on casino accessibility; wckt also rises with k's relative size (approximated by sckt) to account for any heterogeneity in the spatial distribution of economic activity. The distance–decay parameter ρ < 0 is the rate casino accessibility declines as distance from the casino rises, all else being equal.6

Market demand surrounding a casino is calculated using spatially weighted employment and income levels, empct = EctΣkwckt and incomect = Σkwckt × inckt, which proxy for the market size and average incomes within 75 miles of c. The variables Ect and incomekt measure total employment and income-per-employee in zip code k, respectively. Data from the U.S. Census' Zip-Code Business Patterns (ZBP) program are used to calculate wckt, empct, and incomect. ZBP reports select annual economic data for all U.S. zip codes, thus allowing for the economic activity surrounding a casino to vary freely across both time and space. Unfortunately, ZBP does not report natural measures of market size and purchasing power, such as population and household income. Instead, zip-coded employment and payroll-per-employee are used as proxies for a location's relative size and income.7

Demand for casino wagering is also expected to depend largely on price, which is measured using a casino's EGD hold rate, inline image. Controlling for inline image is particularly important when studying cannibalization because wagering is inversely related to price (Kearney 2005; Suits 1979; Thalheimer and Ali 2003), and price has also been found to respond negatively to competition (Navin and Sullivan 2007). Thus, failing to control for prices likely introduces an upward bias into the estimates of cannibalization.8

To measure the intensity of same-industry competition, the degree to which a casino's market area overlaps any competitors' market areas must be accounted for. To this end, a measure of same-industry competition, comp_posct, is calculated as

  • display math(4)

where inline image is the demand-adjusted size of competitor c′ taken from c's perspective, and posc′t is c′'s annualized number of positions. The term inline image adjusts the effective size of c' up or down based on its relative access to zip codes of particular importance to c. Only contested zip codes within both casinos' market areas (i.e., within 75 miles of each) enter into the calculation of comp_posct. Because higher values for comp_posct raise the likelihood that a casino's patrons will substitute toward a closer or larger competitor, comp_posct is expected to have an overall negative impact on a casino's wagering, particularly through its impact on admissions.

The last set of variables, limitct and excursionct, measure the average value of statewide betting limits and the portion of a year that the casino's state had boarding restrictions, respectively. Boarding restrictions (or simulated excursions) require riverboats to undock for a period of time, thus limiting boarding. Once a boat redocks, all patrons are required to leave and reboard. These restrictions have gradually been eliminated in favor of continuous (i.e., dockside) boarding policies.9

Descriptive statistics for each variable are provided in Table 2.

Table 2. Summary Statistics for 1994–2006Thumbnail image of

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A

Baseline estimates

Baseline estimates of equation (2) are summarized in Table 3, column A. Beginning with measures of size, column A1 shows that floor positions and the number of months open positively impact annual wagering. Because these parameter estimates each exceed one, the market for casino gambling does not appear to have been oversaturated during the time frame of this study; markets having reached the point of saturation are likely to experience disproportionately small increases in revenues when capacity levels rise. Columns A2 and A3 point to the patronage effect as being the primary channel through which increases in size raises revenue. Spending-per-patron changes very little.

Table 3. Baseline Estimates of Casino CannibalizationThumbnail image of

Local sources of market demand, as well as prices, also appear to impact spending, although through different channels. Raising demand-adjusted employment by one standard deviation increases wagering by over 200 percent; this occurs primarily through the patronage effect and suggests that population density is a very important determinant of casino demand. The impact of average earnings is far less straightforward, with rising incomes seem to yield fewer, yet higher-spending, patrons. These effects are countervailing, leaving income with little real overall impact on wagering. Altogether, these estimates indicate casinos to be most attracted to high-density areas with plenty of potential gamblers, as is reported by Wenz (2008), while local income levels may not weigh too heavy in the calculation of an optimal location. A patron's demand curve also appears to be downward-sloping, with the estimate for inline image in column A3 being both negative and large in magnitude. At its mean, the price elasticity of demand for spending-per-patron is estimated as −0.99 (= −14.494 × 0.068), indicating that the demand for wagering is unity elastic. This is nearly identical to Thalheimer and Ali's (2003) elasticity estimate for casino wagering-per-capita and is about twice the magnitude of that for state lotteries, as reported by Kearney (2005). Mandated gambling restrictions also appear to detract from local demand, a rather intuitive finding given their intent.10

From column A1, a one standard deviation rise in demand-adjusted competition reduces wagering by approximately 10 percent, providing partial support to the cannibalization hypothesis. The patronage and spending effects may both play an important role here, with increased competition leading to fewer and lower-spending patrons. However, these point estimates are each rather imprecise, which limits the level of certainty that can be put behind them. To address this imprecision, it must be noted that variation in comp_posct relies on fluctuations in the number of competing casino positions and the number of competing casino locations. This is an interesting and important distinction to make because most of the relevant cannibalization literature to date have focused strictly on the presence of competing casinos and not necessarily on their size (Nichols 1998; Thalheimer and Ali 2003; Walker and Jackson 2008). Thus, it is important to ask, is it an increase in the number of competing casinos that drives cannibalization? Or, is it an increase in a competing casino's size? While both factors surely matter when assessing a casino's exposure to competition, it is unclear which factor, if either, is most important. A complete analysis of cannibalization should surely account for both.

To examine this issue further, comp_posct is decomposed into its two major components: 1) the number of competing casinos, comp_casinoct, and 2) the average number of positions per competing casino, avecomp_posct. In similar fashion to comp_posct, comp_casinoct is a demand-adjusted measure of the number of competing casinos:

  • display math(5)

where casinoc′t is the portion of year t (i.e., number of months out of 12) that casino c′ was in operation. Here, inline image is the demand-adjusted number of competing casinos within 75 miles of c.11 avecomp_posct measures the average size of a demand-adjusted casino, which is calculated as inline image.12

Column B of Table 3 re-estimates equation (2) after replacing comp_posct with comp_casinoct and avecomp_posct. From here, it is clear that it is the number of competing casinos that drives cannibalization, not an average competitor's size. A one standard deviation rise in the number of demand-adjusted competitors reduces wagering by nearly 27 percent, most of which comes about through the patronage effect. Spending-per-patron also declines, although imprecisely, suggesting that the marginal patron may be a relatively large spender or that gamblers respond by distributing their fixed budgets across multiple venues, thus reducing spending-per-patron at any one location. This gives some important context to Thalheimer and Ali's (2003) earlier findings that spending-per-capita falls with increases in competition. Any reduction in a casino's spending-per-capita is likely tied to the patronage effect, not the spending effect. The results for avecomp_posct are more mixed. Although the estimators are positive and somewhat large for both wagering and patronage, each lacks the precision needed to warrant much discussion at this time without a more in-depth analysis. The impact of average competitor size will be revisited in the section Extensions using a spatial band analysis.

One potential explanation for the sizable differences in estimated impacts of comp_casinoct and ave_composct is that the opening of a new casino directly alters the spatial distribution of consumers' gaming alternatives, whereas an expansion of a pre-existing casino does no such thing. A casino's locational advantage is thereby potentially most sensitive to new competitors who can encroach on its market area, as opposed to a mere expansion of pre-existing competitors. This is not to say that casino size is unimportant; the estimates for posct are large and significant. Rather, a regional expansion in casino gambling will yield fewer cannibalization effects if the expansion is relegated to pre-existing casinos.

Extensions using a spatial band analysis

Casino accessibility is not likely to attenuate as smoothly or completely as the distance–decay function, eρd, suggests. In order to relax this parametric constraint, the unadjusted level of competition is measured for geographically concentric bands surrounding a casino, replacing comp_casinoct and ave_compposct with a series of band-specific analogs. This allows the rates of cannibalization to vary in magnitude as distance from a casino changes, thus providing a more clear and accurate estimate of spatial attenuation. The first set of variables, inline image, measures the annualized number of competing casinos between d0 and d1 miles from casino c, and the second set, inline image, measures the unadjusted number of annualized positions per competitor within the same set of bands. Three d0-to-d1 bands are used for the estimates reported in Table 4, as determined by structural breaks in the distribution of distances between casinos, per Figure 2. These particular sets of distances are: [0, 10], (10, 25], and (25, 75]. Unadjusted measures of aggregate employment, emp0,75, and income-per-employee, income0,75, are used in place of the demand-adjusted variables from the section Explanatory variables. Summary statistics are reported in Table 2.

figure

Figure 2. Distance Distribution of Competing Casinos within 75 Miles.

Major structural breaks in the distance distribution of competitors are identified at the 10- and 25-mile marks. A less explicit break is visible at the 60-mile mark.

Download figure to PowerPoint

Table 4. Estimates of Casino Cannibalization Using Spatial BandsThumbnail image of

Estimates of equation (2) using the band-specific variables are reported in Table 4, column A. The parameters for inline image are mostly small in magnitude and lacking in precision, providing further support that expansions in casino size have relatively little impact on cannibalization. However, total wagering and patronage react positively and significantly to the size of nearby competitors (i.e., within 10 miles). This suggests that shopping externalities may generate localized benefits, particularly if increased capacity at one casino helps to improve foot traffic elsewhere.13 Although externalities of this type are more commonly associated with exceptionally dense markets such as Las Vegas or Atlantic City, it is not unreasonable to expect that the riverboats in this sample, especially those that operate along opposite flanks of the Mississippi River, can benefit from agglomeration economies of this type as well. It does not appear, however, that heavier foot traffic necessarily translates into increased wagering; an expansion of competitors within the 0-to-10 mile band is associated with a decline in spending per patron, suggesting that, although consumers may visit more casinos as competitors cluster, they may spend less at each individual venue.

Any interesting shopping externality is quickly countervailed, however, once one considers the large and significantly negative impact of an additional competitor. Within the 0-to-10 mile band, one new competitor is associated with a reduction in wagering of approximately 37 percent, most of which comes about through lost patronage. Indeed, the patronage effect is large and significant across all three bands, but attenuates rapidly beyond the 25-mile point; up to 25 miles, admissions fall by nearly 30 percent for each new competitor. This rate falls to only 6 percent for distances beyond 25 miles. The impact of lost patrons does appear to be aggravated by a negative spending effect within the first 10 miles, but is then mitigated by a positive spending effect within the 10-to-25 mile band. Although a complete explanation is not entirely clear, it is possible that two countervailing behaviors are behind this pattern. For example, the negative spending effect within the 0- to 0-to-10 mile band may be attributed to gamblers distributing their limited budgets across multiple nearby casinos, thus reducing spending-per-patron for any one casino. However, this effect may be offset if low-spending patrons (who are not willing to travel very far) are the first to be cannibalized, thus putting upward pressure on a casino's spending-per-patron. This latter effect is most likely to dominate the former when competitors locate farther away from one another, as the incentive to distribute spending across multiple casinos will be lower as well.

Column B of Table 4 re-estimates the model after separating the 25-to-75 mile band into two separate bands (treating the 60-mile marker as another natural break point). The basic patterns are unchanged. How does this set of estimate compare to that from Table 3? Comparing the Akaike or Bayesian information criteria (AIC and BIC) from column B in Table 3 and column A in Table 4, the band analysis offers the slightly better fit, although these differences appear to be minimal. Of greater interest, perhaps, is what these estimates say about the true value of distance decay, ρ, after controlling for the spatial distribution of economic activity. The estimates for inline image in columns A2 and B2 of Table 4 make it possible to approximate the share of a casino's patrons who, but for the presence of a new competitor, would have resided beyond a particular distance, d*, from the casino. This is measured as inline image, where inline image is the parameter estimate for inline image. If casinos always locate at the midpoint of each band, and consumers only travel to the closest venue, competitors operating in the (d0,d1) ring will cannibalize anyone residing beyond inline image miles from the casino. The rate of distance decay within d* miles, inline image, is estimated using inline image such that inline image. The term inline image adjusts for the relative distribution of employment beyond d*, where De(d*) and Da(d*) are cumulative distributions of employment and land area up to d*, respectively.14 As the relative distribution of employment beyond d* rises, the presence of cannibalized patrons having had lived beyond d* becomes attributed more to a higher concentration of economic activity beyond d* and less to a slow rate of distance decay in accessibility.

Although clearly rough approximations, the values presented in Table 5 suggest the rate of distance decay begins large but declines rapidly (and stabilizes) as distance from a casino rises. The weighted average value for ρ ranges from −0.08 to −0.1, which are larger in magnitude (but not exceptionally) than the initial value of −0.03. As a sensitivity check, column B of Table 3 was re-estimated using ρ = −0.09. The estimates are similar in sign and precision to those reported in Table 3; however, the magnitudes are considerably larger in some instances.15 Thus, the results in Table 3 can be interpreted as lower bounds of the cannibalization effect.

Table 5. Approximating the Distance–Decay Parameter, ρ
ModelBand|eβ-1|d*α(d*)De(d*)ρd*
  1. Note: Weighted averages are calculated by assigning a weight of De(d*)/Σd*De(d*) to each ρd*. This weight measures ρd*'s relative representation of the 75-mile market area used to calculate each ρd*.

Table 4, column A20,100.2972.501.010.03−0.48
10,250.3278.751.050.19−0.12
25,75 0.05925.001.331.00−0.08
Weighted average ρ:  −0.1
Table 4, column B20,100.3032.501.010.02−0.47
10,250.3278.751.050.12−0.12
25,600.06321.251.250.51−0.10
60,75 0.04633.751.661.00−0.06
Weighted average ρ:  −0.08

Conclusion and Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A

This paper develops an analytical framework for studying casino cannibalization effects across space. Using a unique panel of Illinois-area casinos, these effects are found to be large, driven by the presence of new, not bigger, competitors, and attenuate quickly with distance. It is hoped that the techniques and observations reported here will contribute to future analyses on this subject. In particular, extending this research to data sets with broader geographic coverage and denser agglomerations should be a direction for future researchers. While the results presented here provide weak evidence that agglomeration economies may exist in the casino industry, they are unlikely to be externally valid for densely concentrated markets such as Las Vegas or Atlantic City where shopping externalities are believed to be substantial. In destination markets such as these, the addition of new casinos may very well add to the region's appeal, thus promoting further agglomeration.

Another potentially fruitful avenue of research involves the incorporation of spatial econometric methods to control for well-defined and spatially correlated unobservables. Casinos' locations, even if regulated by the state, are certainly not random. The fixed-effect estimators in this paper, along with the host of explanatory variables, likely absorb much of relevant spatial autocorrelation that otherwise may go unobserved. However, a spatial econometric analysis that relies on a well-designed model for spatial dependency may yield additional insights.

Lastly, the findings reported here do have interesting policy implications for the growing number of states and local governments considering a casino expansion. In markets structured similar to that of the Illinois-centric region, cannibalization is likely limited to geographic areas no more than 25 miles from the nearest casino and will come about almost entirely from lost patronage. The development of new casinos, and not the expansion of pre-existing ones, will also have the largest impact on another's operations. More importantly, however, it needs to be stressed that cannibalization may only be a second-order determinant of wagering. The size of a market is by far the most important factor, followed by a casino's own size. Thus, if a market is not yet saturated, a mere expansion of pre-existing operations, especially those in large markets, may prove to be the least costly and most timely alternative to the building of new casinos. This approach also runs the least risk of resulting in a cannibalization of nearby in-state gaming venues.

Notes
  1. 1

    This equates to an average just below 2 percent of total 2010 net tax receipts for these 22 states. Net tax receipts are calculated as gross tax receipts less intergovernmental transfers received. Commercial casinos in the American Gaming Association report include racetrack casinos. Data sources: American Gaming Association (2011) and U.S. Census Bureau (2011).

  2. 2

    Two tribal casinos, the Meskwaki in Iowa and the Potawatomi in Wisconsin, are located just within the boundaries of the region for at least a portion of the 1994–2006 period. However, any measurement error introduced due to their absence in the data is expected to be relatively minor. The Meskwaki casino completed a major expansion in July 2006 and did not fall within 75 miles of an Illinois-centric commercial casino until Iowa's Riverside casino opened in August of 2006. The Potawatomi casino had a major expansion in June 2008 and was relatively small for most of the sample period beforehand (see Thompson, Gazel, and Rickman 1995). Data on expansion dates were taken from CasinoCity.com.

  3. 3

    Admission count is often significantly larger than the true visitor count. This is because a single patron may be “double-counted” upon re-entering a boat multiple times during a single visit, especially when excursion restrictions are in place. The presence of state-mandated excursion boating is accounted for in later regressions.

  4. 4

    This technique tracks closely with that used by the states to account for and regulate casino capacity.

  5. 5

    To clarify this latter point, it is helpful to consider the Grand Victoria Casino in Elgin, IL. Elgin rests along the western fringe of the Chicago metropolitan area, so that most patrons visiting Elgin come from its east. As such, the Grand Victoria faces the greatest threat from a competitor locating 15 miles to its east as opposed to one that locates 15 miles to its west, where the communities are predominately rural and yield few patrons to begin with.

  6. 6

    Thalheimer and Ali (2003) were the first to treat ρ as a rate of distance decay in casino accessibility. For this study ρ = −0.03, which is based on the Illinois Gaming Board's (1997: p. 6–11) observation that approximately 80 percent of a casino's visits are from patrons living within a 50-mile radius (i.e., .2 = eρ50 [RIGHTWARDS ARROW] ρ ≈ −0.03). Another report published by Gazel and Thompson (1996) indicates that 85 percent of visitors to Illinois casinos come from within 50 miles, implying ρ ≈ −0.038. However, the spatial distribution of population is not accounted for when estimating ρ.

  7. 7

    For all Illinois zip codes in 2000, the correlations between employment and population and payroll-per-employee and household income are 0.59 and 0.39, respectively (Source: ZBP: 2000, Census of Population and Housing: 2000).

  8. 8

    Treating inline image as an explanatory variable does increase the potential for a spurious correlation with inline image and inline image. To check for this, inline image and inline image were replaced with real data on EGD wagering and EGD wagering-per-admission, which require no transformation using inline image. The estimates were nearly identical, suggesting little-to-no spurious correlation is present in the results reported below. These estimates are available upon request.

  9. 9

    Iowa introduced continuous boarding in April 1994. The other states adopted continuous boarding much later (Illinois, July 1999; Missouri, November 1999; Indiana, August 2002).

  10. 10

    Boarding limits have the anticipated effect of raising admission counts, indicating many patrons simply re-board after a boat docks.

  11. 11

    It can be shown that the general structure of comp_casinoct is similar to that used by Thalheimer and Ali (2003) to measure competition. However, comp_casinoct takes advantage of much finer levels of geographic variation (i.e., zip codes), whereas Thalheimer and Ali relied on county-level data.

  12. 12

    avecomp_posct is set to zero in instances when there are no competing casinos within 75 miles.

  13. 13

    For example, a competitor's expansion may involve the addition of a hotel as well as the provision of a broad range of entertainment services, all of which may help foster the development of a destination market. The data set used here does not contain information on hotel rooms, restaurants, or other non-gambling services provided.

  14. 14

    Da(d*) = π(d*)2/π(75)2.

  15. 15

    The estimates for comp_casinoct and ave_composct are reported in Table A1 of the Appendix.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A
  • American Gaming Association. 2011. State of the States: The AGA Survey of Casino Entertainment. http://www.americangaming.org/ (accessed November 2013).
  • Anders, G.C., D. Siegel, and M. Yacoub. 1998. Does Indian casino gambling reduce state revenues? Evidence from Arizona. Contemporary Economic Policy 16(3): 347355.
  • Garrett, T., and M. Pakko. 2010. The Revenue Performance of Casinos after a Smoking Ban. Federal Reserve Bank of St. Louis Working Paper #2009-27b. Revised March 2010. http://research.stlouisfed.org/wp/2009/2009-027.pdf (accessed November 2013).
  • Gazel, R.C., and W.N. Thompson. 1996. Casino Gamblers in Illinois: Who Are They? Report to The Better Government Association. Chicago, IL.
  • Illinois Gaming Board. 1997. 1997 Annual Report. Springfield, IL.
  • Kearney, M.S. 2005. State lotteries and consumer behavior. Journal of Public Economics 89(11–12): 22692299.
  • Navin, J.C., and T.S. Sullivan. 2007. Do riverboat casinos act as competitors? A look at the St. Louis market. Economic Development Quarterly 21(1): 4959.
  • Nichols, M. 1998. Deregulation and cross-border substitution. Journal of Gambling Studies 14(2): 151172.
  • Popp, A., and C. Stehwien. 2002. Indian casino gambling and state revenue: Some further evidence. Public Finance Review 30(4): 320330.
  • Siegal, D., and G. Anders. 1999. Public policy and the displacement effects of casinos: A case study of riverboat gambling in Missouri. Journal of Gambling Studies 15(2): 105121.
  • Siegal, D., and G. Anders. 2001. The impact of Indian casinos on state lotteries: A case study of Arizona. Public Finance Review 29(2): 139147.
  • Suits, D. 1979. The elasticity of demand for gaming. The Quarterly Journal of Economics 93(1): 155162.
  • Thalheimer, R., and M. Ali. 2003. The demand for casino gaming. Applied Economics 35(8): 907918.
  • Thompson, W., R. Gazel, and D. Rickman. 1995. The economic impact of native American gaming in Wisconsin. Wisconsin Policy Research Institute Report 8(3): 151.
  • U.S. Census Bureau. 2011. 2010 Annual Survey of State Government Finances. G10-ASFIN. Washington, D.C.
  • Walker, D.M., and J.D. Jackson. 2008. Do U.S. gambling industries cannibalize each another? Public Finance Review 36(3): 308333.
  • Walker, D.M., and J.D. Jackson. 2011. The effect of legalized gambling on state government revenue. Contemporary Economic Policy 1: 114.
  • Wenz, M. 2008. The spatial evolution of casino gambling. CityScape 10(3): 203227.

Sources of Data

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A

Appendix A

  1. Top of page
  2. Abstract
  3. Introduction
  4. The Illinois-Centric Market Area
  5. Conceptual and Empirical Framework
  6. Results
  7. Conclusion and Discussion
  8. References
  9. Sources of Data
  10. Appendix A
figure

Figure A1. Market Area for Casino c1 with Competition.

Casino c1's market area is depicted using gray-shaded rings with darker shades indicating a higher share of c1's total patronage (assuming ρ < 0 and an even distribution of population across space). The market area for c2 is the dotted circle and all contested patrons reside where market areas intersect. The solid black line identifies all equidistant locations within a contested market area, and d* is the shortest distance between c1 and all equidistant locations. Contested patrons closer to c2 are cannibalized.

Download figure to PowerPoint

Table A1. Baseline Estimates Using ρ = −0.09.Thumbnail image of