SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Review of the Literature on Cross-Border Gaming
  5. Empirical Analysis
  6. Concluding Remarks
  7. References

Using panel data methods, for all the Spanish Autonomous Communities and for the period 1999–2009, the present study estimates the revenue of the casinos in each Autonomous Community on the basis of four variables of interest related to the taxation of gambling in each Community, the taxation in the neighboring Autonomous Communities and the expansion of online gambling, and certain economic and sociodemographic control variables. The estimation performed permits the verification of the hypotheses that the taxation of gaming in a region and the expansion of online gambling negatively affect the revenue accrued by the casinos located in that region, but does not offer evidence that the effect produced by the differences in taxation among neighboring regions is significant. The results obtained also confirm the expected impact on the casinos' revenue of the existence of “type” gamblers (young males, and tourists) and of the economic situation (income and unemployment).


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Review of the Literature on Cross-Border Gaming
  5. Empirical Analysis
  6. Concluding Remarks
  7. References

In Spain, as in other countries, there exists a plurality of types of gaming: lotteries, betting, raffles, competitions, etc. Gaming of a national scope is regulated and is the responsability of the central government. Gaming of a regional scope is regulated and is the responsibility of the regional governments (the Autonomous Communities). The operation of lotteries is reserved for the public administrations. For example, lotteries of a national scope are operated by the central government through the public entity State Lotteries and Betting (Loterías y Apuestas del Estado, LAE) and, through administrative authorization, by the National Organization of Spanish Blind People (Organización Nacional de Ciegos Españoles, ONCE). In the remaining gaming activities both public and private operators may intervene.

Until recently, the taxation of gambling in Spain was comprised of two taxes. The first is the tax on raffles, tombolas, bets, and random combinations, which taxes their authorization, celebration or organization; the second is the tax on games of chance, which taxes the organization or celebration of gaming in casinos and bingo halls and the use of recreational machines and machines of chance.

Both taxes constitute revenue for the Autonomous Communities. Initially, the regions did not have the authority to modify the regulation of these taxes, which was fixed at the central level, but did however enjoy responsibility for their administration. Nevertheless, the Autonomous Communities did utilize their powers widely to establish surcharges on those taxes, and similarly to establish their own taxes on gambling.

From 1997 onwards, the Autonomous Communities have enjoyed very wide authority in the regulation of taxes on gambling, in determining both the taxable base and the tax rates. This widening of regional regulatory powers has been utilized by the Autonomous Communities, which has logically led to the practical disappearance of the surcharges and own taxes on gambling. This legislative activity of the Autonomous Communities has produced a certain regional diversity in the taxation of gambling, the possible economic effects of which have not yet been studied.

In May 2011 a new tax came into force in Spain, the Tax on Gaming Activities (Law 13/2011), which taxes all gaming activities of a national scope (with the exception of lotteries), including online gambling. The Autonomous Communities have a share in the collection of that tax and can also exercise some regulatory authority over tax rates.

Regional taxes on gambling can affect the behaviour of gamblers in various ways. The objective of the present paper is precisely to empirically test the effects of these taxes on the behavior of gamblers in the Autonomous Communities. Interest is focused on determining whether the taxation of gaming in a region affects gaming in that region and, in line with the empirical literature on cross-border shopping, whether regional tax differences induce gamblers resident in one Autonomous Community to travel to another Community with lower taxation. The research will also attempt to determine the effect of online gambling—especially favored for not being subject to taxation until now—on gaming activities in the Autonomous Communities. This case is a special version of the phenomenon of cross-border gaming, in which the gambler does not move to another physical jurisdiction, but instead to a virtual one. Given these objectives, the empirical exercise performed is limited to gambling in casinos, as this is the activity which is most likely to produce the abovementioned behaviors of gamblers, as against, for example, gambling in bingo halls or on some recreational machines and machines of chance.

The paper is structured in four sections. Following this introduction, the second section offers a concise review of the applied literature on “cross-border gaming.” The third section undertakes an econometric exercise in which, through panel data methods, for all the Autonomous Communities and the Autonomous Cities of Ceuta and Melilla, and for the period 1999–2009, an estimation is made of the revenue of the casinos in each Community, on the basis of four variables of interest, related to the taxation of gaming in each Community, the taxation in the neighboring Autonomous Communities and the expansion of online gambling, and five control variables, of an economic (regional GDP per capita and the regional unemployment rate) and sociodemographic (travelers, male youths with higher education) nature. The estimation performed obtains evidence that the taxation of gaming in a region and the expansion of online gambling negatively affect the revenue accruing to the casinos located in it, but not that the effect produced by the differences in taxation among neighboring regions is significant. The paper ends with some concluding remarks.

Review of the Literature on Cross-Border Gaming

  1. Top of page
  2. Abstract
  3. Introduction
  4. Review of the Literature on Cross-Border Gaming
  5. Empirical Analysis
  6. Concluding Remarks
  7. References

The literature, which attempts to empirically test for the existence of cross-border shopping produced by tax differences among jurisdictions, has a long tradition in the comparative sphere (especially in the U.S.: see Leal, López-Laborda, and Rodrigo, 2010), yet has hardly been touched upon in Spain. Leal, López-Laborda, and Rodrigo (2009) analyzed cross-border shopping produced by the differences generated among the Spanish regions in the Hydrocarbon Retail Sales Tax.

The empirical analysis of the effects of the existence of cross-border gaming on the revenue obtained by a specific jurisdiction due to the activity of gaming and on other related industries has been scarcely developed outside the U.S. Garrett and Marsh (2002) estimate the impact of cross-border shopping on the revenue from lotteries in the state of Kansas, for the year 1998. Estimating by ordinary least squares and correcting for spatial dependency, the authors confirm the existence of cross-border shopping for lotteries and calculate that this phenomenon represents for Kansas overall net losses of approximately 10.5 million dollars, 5.4 percent of its lottery sales.

In the same vein, Elliott and Navin (2002) evaluate, for 48 states of the U.S., and for the period 1989 to 1995, the role played by the existence of lotteries and the amount gambled on them in neighboring territories in explaining the decision to implant this form of gaming in a specific state and the importance acquired by private expenditure on it. The paper shows that the presence of lotteries in bordering states has a significant negative effect, although minor, on this expenditure.

Tosun and Skidmore (2004) analyze the data for lottery sales in the 55 counties of West Virginia throughout the period 1987–2000, with the objective of identifying the factors explaining their evolution. The results of the panel data estimation they perform show once more the importance of possible external competition (the introduction of new modes of gaming) in determining the revenue obtained from lotteries in a jurisdiction.

Skidmore and Tosun (2008) estimate the effects produced by the introduction of new lottery games, both in West Virginia and in its neighboring states, on the revenue obtained by the commercial sector of the former (as a proxy of its sales). The structure utilized in that study is that of panel data and the period of analysis extends from 1987 to 2001. The authors show the positive relationship between lottery sales and retail activity, although this is explained by what happens in the internal counties and not by what occurs in the border counties. The introduction of new lotteries in the neighboring states reduces retail activity in the border counties of West Virginia.

Walker and Jackson (2008) also explicitly consider the influence of cross-border consumption when studying the relations of complementarity or of substitution, which occur between lottery games, casinos, and horse and greyhound racing in the states of the U.S., during the period 1985 to 2000. In that study, the econometric estimation technique utilized is that of seemingly unrelated regression equations, with a panel data structure. The results from the paper show that state revenue from casinos decreases with the presence of this type of gambling in adjacent states.

Some recent papers contradict earlier results. Garrett and Coughlin (2007) study the income elasticity of demand for lottery tickets for a set of counties in the states of West Virginia, Iowa, and Florida, between the late 1980s and 2005. The authors conclude that the introduction of lotteries in neighboring states does not significantly change the income elasticity of demand in the adjoining state.

Using data from 48 of the U.S. in the period 1977–2006, Stitzel and Pjesky (2011) conclude that revenue from lotteries in one state is not affected by the introduction of a lottery in a neighboring state. Nevertheless, the study confirms the hypothesis that state lotteries are complementary and that, when a state introduces a lottery, all states with lotteries, neighboring or not, experience significant increases in their revenue.

Knight and Schiff (2010) develop a theoretical model that predicts that if cross-border gaming is substantial, the negative relationship between lottery sales and prices must be stronger in states with lower populations and densely populated border regions. Estimated using panel data, with weekly information from 1995 to 2008 for the U.S., the empirical results confirm the theoretical predictions.

In short, the literature has proven the relevance of cross-border gaming, but it has not studied the influence that the existence of different gambling taxes in different jurisdictions may have on that behavior. This last approach is precisely the one adopted in the following section of this paper.

An analysis of cross-border gaming would be incomplete if it ignored the fact that the spectacular expansion of the activities performed over the Internet widens the options available to gamblers. So the gambling on the Internet provides an additional alternative to traditional gaming in the region of the residence or in the border ones. It would appear that the literature has not focused on this problem by taking as the research subject the activity of gaming, but it has done so for sales taxes. Goolsbee (2000), Alm and Melnik (2005), and Goolsbee, Lovenheim, and Slemrod (2007) analyze the effects of sales taxes on the decision of an individual to purchase in his or her locality or to do so on the Internet. They find a positive and significant relationship between residency in a locality with high taxation and the probability of purchasing on Internet: Goolsbee estimates a tax elasticity in the range of 2 to 4, while Alm and Melnik calculate an elasticity of around 0.5.

Ballard and Lee (2007) combine the analysis of the effects of taxes on Internet shopping with the analysis of cross-border shopping, obtaining two basic results. Firstly, that a resident of a county with high tax rates is more likely to purchase on the Internet than a resident of a county with low rates. And secondly, that the resident of a county bordering another with a lower tax rate or a narrower taxable base is less likely to purchase on Internet, ceteris paribus. For Ballard and Lee, these results show that purchases in the jurisdiction of residence, in the neighboring jurisdiction and on the Internet are substitutive.

Empirical Analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Review of the Literature on Cross-Border Gaming
  5. Empirical Analysis
  6. Concluding Remarks
  7. References

The objective of this section is to empirically determine whether gaming in an Autonomous Community is conditioned by the taxation of that activity in that region and in the neighboring regions and by online gambling. As explained in the introduction, given the objectives of the study, the empirical exercise is limited to gambling in casinos, as this is the activity in which it is most probable that gamblers can respond to a tax change in one region by traveling to another.

Some figures on gaming in casinos in Spain

Gaming in casinos is characterized by the existence of a wide set of games, which can only be played in these establishments: French roulette, American roulette, poker, etc. Furthermore, only these establishments can house machines of type “C” or of chance (slot machines).1

In proportion to the total of the amounts gambled per inhabitant/year in games of chance,2 gambling in casinos is the least important, touching 11 percent in recent years, as against the 16.5 percent of the total expenditure made in bingo halls and the approximately 72.5 percent spent on type “B” machines. However, its magnitude in absolute terms reached 39.5 euros per inhabitant in the year 2010, a figure practically identical to the average values of the quantities invested in ONCE lotteries (39.6 euros per inhabitant).

In 2010, 39 casinos existed in Spain and, except for Castile-La Mancha and Navarre, all the Autonomous Communities had at least one installed on their territories, an indicator of the popularity of this type of gaming. Figure 1 shows the supply of casinos in Spain, and also the number of “C” type machines available in them as a whole. The map (Figure 2) reflects the distribution of casinos among Autonomous Communities.

figure

Figure 1. Casino Supply in Spain.

Source: Ministry of the Interior (1999–2010).

Download figure to PowerPoint

figure

Figure 2. Number of Casinos in Each Autonomous Community (2010).

*In the Valencian Community, as well as the casinos, there are three “appendices rooms” in different locations of the region.

Source: Ministry of the Interior (1999–2010).

Download figure to PowerPoint

Figure 3 approximates the level of the activity of the casinos. It shows the evolution of total revenue and the number of visits to the casinos.3 While visitors increased by 38.9 percent between 1999 and 2010, revenues did so by 21.3 percent, effectively half. Similarly, the trend presents disparate behavior. Visits fell in the early years of this century but from 2004 onwards increased once more. The volume of total revenue had been increasing until 2007, showing in the last three years a clear reverse, before returning in 2010 to the levels at the start of the decade.

figure

Figure 3. Evolution of Activity in Casinos.

Source: Ministry of the Interior (1999–2010).

Download figure to PowerPoint

The sources of ordinary revenue are four-fold: gaming (roulette, poker, etc.), “C” type machines, tips, and entrance tickets. Table 1 reflects the evolution of the overall revenue figures, as well as each of their components. Figure 4 presents the distribution of the turnover of the casinos by each of the distinct concepts. According to the average of the period studied, revenue from gaming constituted the most significant element, 59.7 percent. Secondly, the collection from “C” type machines represents a notable percentage of total revenue (25.4 percent). Finally, tips and entrance tickets yield 14.2 percent and 0.7 percent, respectively.

figure

Figure 4. Internal Distribution of Casino Revenue.

Source: Ministry of the Interior (1999–2010).

Download figure to PowerPoint

Table 1. Casino Revenue (in Millions of Euros)
Year199920002001200220032004200520062007200820092010
Source: Ministry of the Interior (1999–2010).
Gaming196.37218.41234.99257.80262.80283.27329.54329.73330.70289.02245.42227.00
“C” machines59.9075.0185.0796.27101.95118.26141.19148.68153.40146.10131.57126.99
Tips59.8864.2965.9665.8166.2467.5375.9271.3872.8561.0049.4031.27
Entrance tickets3.753.913.423.272.872.883.233.122.952.602.322.56
Total319.90361.62389.44423.15433.86471.94549.88552.91559.90498.72428.71387.82

With regard to the evolution of the revenue corresponding to each source, in the last decade, there has been a significant recomposition of the relative share of the different items in total revenue. Thus, gaming revenue lost almost three percentage points of its relative share between 1999 and 2010. As Table 1 shows, the revenue generated by this component in 2010 is lower than that of 2001. Similarly, the two components of least quantitative importance, the sums received for tips and entrance tickets, have also seen their share reduced, by 10.7 and 0.5 percentage points, respectively. As reflected in Table 1, in 2010, the revenue corresponding to these two concepts is lower than that for 1999.

Only the revenue from “C” type machines has experienced a positive variation in the weight of internal composition. This revenue has continously increased its percentage of the total distribution of turnover, rising from 18.7 percent in 1999 to 32.7 percent in 2010. In absolute terms, the fall in revenue in the last three 3 years shown in Table 1 has been noticeably lower than that experienced by gaming revenue.

Specification, estimation, and results

To correctly propose the econometric specification, it is important to start by identifying the ways in which the regional taxation of gaming can affect the behavior of gamblers. The prizes arising from gaming in casinos depend on the bets or bids and are the same throughout Spain, and thus are not affected in whatever case by taxation. Type “C” machines return to gamblers in the form of prizes a specific percentage of the sums gambled, which may vary among Autonomous Communities; these prizes may therefore be conditioned by the taxation of gaming.

It is consequently fitting to propose as a hypothesis that the taxation of casinos and “C” type machines is passed on to gamblers, at least partially, via two mechanisms: firstly, the prizes from the slot machines, and the secondly, the prices of the complementary services provided in the casinos: entrance tickets, restoration, etc. Consequently, if a Community raises its taxes on gaming, this increase will be passed on, in part, to prizes and prices, encouraging a reduction of gaming in that Community and, as a result, in the revenue of its casinos, and—ceteris paribus—an increase in gaming and in the revenues of casinos in neighboring Communities. Furthermore, and above all in recent years, gamblers enjoy another gaming alternative, which does not require their physical presence in casinos: online gambling, neither regulated nor taxed in Spain until the passing of Law 13/2011, of May 27.4 In accordance with the literature on cross-border shopping reviewed in second section, it is possible that the nontaxation of online gaming could be one of the causes explaining the expansion detected of this activity in Spain in the last 5 years and, as a result, the reduction of presence-based gambling, subjected to taxation. If this were the case, the result would be a displacement of gamblers in the region to a virtual jurisdiction with a more favorable fiscal treatment.

In accordance with the above, the econometric exercise will estimate, using panel data methods, for all the Autonomous Communities and the Autonomous Cities of Ceuta and Melilla and for the period 1999–2009, the revenue of the casinos in each region on the basis of the taxation of gaming in each Autonomous Community and in the neighboring Autonomous Communities, a variable representative of the expansion of online gambling, and various economic and sociodemographic control variables, which the literature has shown to be explanatory factors of the evolution undergone by gambling in casinos.

Concretely, the following specification is assumed:

  • display math

where the subindex i represents each of the Autonomous Communities and Autonomous Cities with casinos on their territory (as stated earlier, all the Communities except Castile-La Mancha and Navarre) and the subindex t, the year considered. Next, a description is given of the variables introduced in the previous specification.

Casino represents the total real revenue generated by the casinos, and includes revenue from gaming in the strict sense, those obtained by type “C” machines, installed exclusively in these establishments, revenue from entrance tickets, and tips. The information has been obtained from the Memoirs and Reports on Gaming compiled annually by the Ministry of the Interior.

Rate is a tax variable intended to capture the effect of gaming taxes in each Autonomous Community on the revenue of the casinos located in its territory. As explained earlier, different taxes on gaming converge in a casino, and thus, an aggregate tax rate, which captures the level of joint taxation borne by this economic activity, is constructed. Concretely, this aggregate tax rate is equal to the weighted sum5 of the maximum marginal tax rate applied to gaming in casinos and of the specific rate applied to “C” type machines (relativized with regard to the average annual collection per machine). Regional surcharges in force throughout the period are also included in the aggregate tax rate. By way of example, Table 2 shows taxation on casino games and “C” type machines in the Autonomous Communities in 2009.

Table 2. Taxation of Gaming in Casinos in 2009
Autonomous communityCasinosSlots
Initial rateTop rateFee per machine (1 gambler)
Source: Our elaboration.
Andalusia20%58%4,623.89€
Aragon20%55%5,460€
Asturias20%55%5,400€
Balearic Islands22%61%4,946€
Basque Country20%35%4,025€
Canary Islands20%55%4,207.08€
Cantabria24%60%5,600€
Castile and Leon20%55%5,265€
Catalonia20%55%1,316€
Extremadura20%55%4,803€
Galicia22%60%5,460€
La Rioja24%60%5,872€
Madrid22%45%5,400€
Murcia25%55%5,300€
Valencia20%55%5,950.5€

It is expected that greater taxation produces a reduction in the prizes of machines or an increase in the price of other services connected to the activity of gaming and, consequently, a reduction in the revenue of the casinos of the Community.

Relative rate is a variable, which measures the ratio between the average aggregate rate of the bordering Communities and the rate of the region itself. It is expected that an increase in this ratio will be accompanied by displacements of gamblers in these adjacent regions and, from there, an increase in the revenue of the casinos located in the Autonomous Community is produced.

Intro * Difference rate is a variable, which results from multiplying Intro, a dummy that takes the value of 1 in the year in which a bordering Community establishes, for the first time, a casino, and Difference rate, a variable, which expresses the difference between the aggregate rate of the Community itself and that of the neighboring region that has established the casino. The review performed of the literature produces the expectation of a negative effect of this variable: the bigger the difference between rates, the lower will be the incentive of gamblers in the region with new casinos to move to regions with existing casinos.

Online is a dummy variable, which takes the value of 1 from 2006 on. Repeatedly, firms in the conventional gaming sector have stated that the uncontrolled expansion of online or Internet gaming, which, having not been regulated until the approval of Law 13/2011 had not paid taxes, has seriously affected their turnover in recent years. Diverse statistics and reports indicate 2006 as the year from which this expansion occurred.6 A negative sign in the estimated value of the coefficient corresponding to this variable is therefore expected.

In turn, X is an i × m matrix of control variables, which includes the groupings listed below.

Economic situation variables

GDPpc captures regional GDP per capita, deflated according to the Consumer Price Index series. The data come from the National Statistics Institute (Instituto Nacional de Estadística, INE).

Unemployment captures the regional unemployment rate, measured by the data provided by the Economically Active Population Survey (Encuesta de Población Activa, EPA).

A priori, the amounts gambled will be influenced by macroeconomic variables such as the evolution of GDP or the unemployment rate. It is expected that while a positive evolution of GDP encourages greater gaming in casinos, the growth of regional unemployment produces the opposite effect.

Sociodemographic variables

Young males captures the percentage of the total regional population represented by males of between 20 and 44. The data source is the INE.

Educated males captures the percentage of the regional population of 16 or over constituted by males with higher education. The information also proceeds from the INE.

Travelers captures the regional number of travelers, calculated by nights spent in hotel establishments, campsites, tourist apartments, and rural tourism, as a percentage of the national total. The data were obtained from the Tourism Accommodation Occupancy Surveys (Encuestas de Ocupación en Alojamientos Turísticos).

This set of variables is aimed at capturing the standard profile of gamblers in casinos: The literature reviewed led to the choice of variables, which capture age, gender, education, or the number of tourists who visit a region. Consequently, an increase in the value of these variables will be associated with greater revenue obtained by the casinos.

Table 3 shows the principal descriptive statistics of the variables used in the model.

Table 3. Descriptive Statistics of the Variables Used in the Analysis
 MeanStandard deviationMinimum valueMaximum valueCoefficient of skewnessCoefficient of kurtosis
Source: Our elaboration.
Casino (in millions of €)31.7334.830.24131.751.373.77
Rate0.450.070.240.922.3917.46
Relative rate0.940.3001.83−1.597.57
Intro * Difference rate−0.010.07−0.560.08−7.1152.78
Online0.410.49010.381.14
GDPpc20,693.63,858.8213,286.2629,833.80.442.42
Unemployment0.110.050.050.271.073.69
Young males0.210.010.180.230.082.19
Educated males0.210.050.120.371.143.97
Travelers0.060.060.00040.180.762.24

The database employed is a nonbalanced panel, as not all the Autonomous Communities considered have at least one casino throughout the period. The estimation has been performed using the fixed-effects procedure since, from a theoretical point of view, this appears to be the most advisable option when there is a panel available of all the regions comprising a country and not merely a selected sample of them.7

Elsewhere, a prior analysis has been made of the correlations existing among the explanatory variables, to rule out possible problems of multicolinearity in the estimations. The construction of the fiscal variables (Rate, Relative rate, Intro * Difference rate) could a priori give rise to some doubt on this point. The highest value of these tax correlations is that presented by Rate and Relative rate and is equal to −0.35, which appears to indicate the nonexistence of the problem alluded to.

Given the nature of the data, in addition to diagnosing eventual problems of serial autocorrelation and of heteroskedasticity, it is necessary to deal with the possible existence of spatial dependence. With regard to the detection of problems of serial autocorrelation, the use of a test proposed by Wooldridge (2002) specifically for panel data, on the basis of the estimation of a first-differences model robust to serial correlation, leads to reject the null hypothesis of an absence of autocorrelation and suggest the presence of a first-order autoregressive process (F[1,16] = 12.892, Prob > F = 0.0024).

In turn, the performance of the modified Wald test leads to reject the null hypothesis of homoskedasticity and propose, alternatively, for the data structure employed here the presence of heteroskedasticity among groups (χ2[17] = 3,876.51, Prob > χ2 = 0.0000).

Finally, in this task of prior diagnosis, and with an approximation similar to that followed by Garrett and Marsh (2002) and Tosun and Skidmore (2004), the possible presence of spatial dependency in the present exercise is tackled. The spatial dependency models were introduced by Cliff and Ord (1981) and Anselin (1988) and consider the direct influence the regions or neighboring areas have on a territory, or the possible spillover effects or externalities generated among regions. The nonconsideration of this contemporary autocorrelation could lead to the obtaining of biased and inconsistent estimations of the coefficients of the specification.

For the detection of spatial autocorrelation, a set of tests is performed, constructed for the spatial error and spatial lag models (Anselin et al. 1996), on various panel cross sections. This in turn requires the construction of a matrix of spatial weights, W. The spatial error model assumes the existence of an autoregressive process in the error term, ε = λWε + υ, where ε is a vector n × 1 of errors, υ is a vector n × 1 of independent and identically distributed (iid) errors, W is a matrix n × n of spatial weights, which can adopt distinct configurations and λ is a scalar, which is interpreted as the coefficient of non-observed spatial correlation. Logically, the error terms are not spatially correlated in the case that λ is equal to zero.

In turn, the spatial lag model responds to the following characterization: y = ρWy +  + ε. In this specification, y represents the dependent variable, ρ is a scalar, which represents the spatial correlation coefficient, X is the matrix n × k of exogenous variables. Once again, if ρ proves to be zero, the existence of the abovementioned correlation can be rejected.

When constructing the matrix W, and following the characterizations habitually found in the literature, this study tests both the binary matrix, where wij = 1 if the observations i, j (with i ≠ j) correspond to regions, which share borders and wij = 0 otherwise, and also with the contiguity matrix, whose elements wij* are defined as inline image, where wij = 1 if the observations i, j (with i ≠ j) share a common geographical area (in the present case, for each region j, the set of Communities that share a border with the region is considered as a common geographical area) and wij = 0 otherwise.

Having performed the construction of both matrices W, the calculation of the Robust Lagrange Multiplier test permits testing of the null hypothesis of the presence of coefficients of spatial autocorrelation (λ, ρ) equal to zero.8 This hypothesis is rejected for the two models of spatial dependency when using the contiguity matrix (spatial error model: Robust Lagrange Multiplier = 9.09, p-value = 0.003; spatial lag model: Robust Lagrange Multiplier = 5.89, p-value = 0.01), and for the spatial error model when the binary matrix is used (spatial error model: Robust Lagrange Multiplier = 2.71, p-value = 0.1; spatial lag model: Robust Lagrange Multiplier = 0.59, p-value = 0.44).9

Given the proven existence of serial and spatial autocorrelation and also the presence of heteroskedasticity, and similarly to the approach made by Tosun and Skidmore (2004), the present study performs a panel-corrected standard error estimation, described in Beck and Katz (1995), which basically achieves an estimation of the matrix of variances and covariances of the parameters that is asymptotically efficient.10

Table 4 shows the results obtained with this estimation technique.11 The table makes it clear that the taxation of gaming in the region itself (Rate) has a significant negative influence on the revenue of its casinos. Empirical confirmation is also made of the negative and significant impact of online gambling (Online) on presence-based gaming in casinos; this is doubtless due to various reasons, among them the greater comfort of online betting and its nontaxation in the years included in the estimation.12 The new regulation of online gaming, which incorporates its taxation, will reduce, as long as its application is effective, the attraction of this activity with regard to gaming in casinos.

Table 4. Results of the Estimation Performed
 Estimated coefficient (p-value)
Rate−17.98 (0.028)
Relative rate−3.70 (0.421)
Intro * Difference rate−0.80 (0.815)
Online−3.74 (0.077)
GDPpc0.002 (0.002)
Unemployment−0.31 (0.029)
Young males289.11 (0.016)
Educated males−27.88 (0.301)
Travelers486.31 (0.004)
N = 167
R2 = 0.9682
Wald χ2(17) = 6,340.60
Prob > χ2 = 0.0000

However, differences in taxation with the neighboring Communities (Relative rate) do not significantly affect this revenue: An increase in the ratio between the average aggregate rate of the adjacent Communities and the rate of the region itself does not appear to have any influence on the turnover of the casinos in the region.13 Consequently, the results obtained indicate that individuals do not move among regions as a result of the lower fiscal cost of gaming in some territories, but they do confirm that presence-based gaming is being replaced by online gambling. There does not appear to exist, therefore, physical but instead virtual cross-border gaming.

Moreover, the opening of the first casino in a neighboring Autonomous Community (Intro * Difference rate) does not have a significant impact, not even strictly occasional (limited to the year of opening), on the revenue obtained by the casinos in the region itself.

Almost all the control variables prove significant and with the sign expected, that is to say positive in the case of regional income (GDPpc) and of the variables determining the conventional profile of the clients of such establishments, namely Young males and Travelers, and negative in the case of the effects captured by the variable Unemployment. The only variable associated to this conventional profile that does not appear to have a direct incidence upon the revenue figures is that which captures the percentage of the regional male population of 16 or over with higher education (Educated males).

Concluding Remarks

  1. Top of page
  2. Abstract
  3. Introduction
  4. Review of the Literature on Cross-Border Gaming
  5. Empirical Analysis
  6. Concluding Remarks
  7. References

The aim of the present paper has been to empirically test the effects of regional taxation on gaming in the Spanish Autonomous Communities. The focus has been on determining whether taxation in one region affects gaming in that region and, in line with the empirical literature on cross-border shopping, whether regional tax differences induce gamblers resident in one Autonomous Community to travel to another Community with lower taxation. The paper has also attempted to identify the effect of online gaming, treated especially favorably from the tax point of view, on gaming in the Autonomous Communities. This is a special manifestation of the phenomenon of cross-border gaming, in which the player does not move to another physical jurisdiction, but instead to a virtual one.

The estimation performed has permitted verification of the hypothesis that the taxation of gaming in a region negatively affects the revenue accruing to the casinos located in it, but, by contrast, does not offer evidence that the effect produced by the differences in taxation between neighboring regions is significant. The results obtained also indicate that the expansion of online gambling has had a significant and negative impact on the turnover of the casinos in the region, although it must be remembered that, due to the lack of availability of other alternatives, a dummy variable has been employed as a proxy for the expansion of online gambling, which is not fully satisfactory. The estimation also confirms the expected impact on the revenue of casinos of the existence of “type” players (young males and tourists) and of the economic situation (income and unemployment).

These results have very important implications for economic policy. Regional governments and legislators need to be aware of the consequences derived from the fiscal measures they propose or adopt in the sphere of gaming. The present research suggests that an increase in regional tax rates leads to a decrease in casino revenues. Thus, a tax increase will have consequences in terms of collection (of other taxes too, such as VAT or income taxes) and may additionally affect economic activity and employment in the region. These results have been obtained for gaming in casinos, but can surely be extended to other modes of gaming. Similarly, the methodological approach followed here can easily be applied to other geographical areas, whether national (e.g., federal countries) or supranational (e.g., the European Union).

Notes
  1. 1

    Type “A” (or recreational) machines grant the user a time of use or of game, but do not give prizes. Type “B” (or recreational with scheduled prize) machines grant a time of use or of game and, eventually, in accordance with the program of the game, a cash prize. Type “C” (or chance) machines, in exchange for a specific bet, grant a time of use or of game and, eventually, a prize that always depends on randomness.

  2. 2

    The share of expenditure on games of chance represented approximately 58 percent of the total amounts played in 2010. LAE lotteries represented around 35 percent and ONCE lotteries approximately 7 percent.

  3. 3

    In accordance with the annual Memoirs and Reports on Gaming (Memorias e informes anuales del juego), it is more correct to speak of visits to casinos more than of visitors since there exist habitual players who attend such establishments frequently.

  4. 4

    The activity performed by the European Commission to analyze the impact of online gambling on the Single European Market can be consulted at http://ec.europa.eu/internal_market/services/gambling_en.htm.

  5. 5

    The weighting has been performed according to the relative importance of collection from games and machines in the set of Autonomous Communities.

  6. 6

    See, for example, the Telecommunications Market Commission, Comisión del Mercado de las Telecomunicaciones (2000–2010), Orange Foundation, Fundación Orange (2001–2010).

  7. 7

    Additionally, the Hausman test rejects the hypothesis of the nonexistence of statistically significant differences in the coefficients of the time-varying explanatory variables in the fixed effects and random effects models (χ2[8] = 35.43, Prob > χ2 = 0.00). Moreover, the estimation of the fixed effects of the proposed specification permits the performance of a Wald test, which confirms the joint significance of the individual effects (χ2[16] = 2,387.59, Prob > χ2 = 0.0000).

  8. 8

    In STATA, this diagnostic task is performed by the “ado” files provided by Pisati (2001).

  9. 9

    The additional performance of the Breusch-Pagan LM test confirms for this exercise the presence of a problem of contemporaneous correlation among the regions: χ2(136) = 201.38, Prob > χ2 = 0.0002, and therefore, the null hypothesis of cross-sectional independence in the residuals of the fixed effect regression model is rejected.

  10. 10

    This method is considered as an alternative to the use of feasible generalized least squares (FGLS), also used when iid errors cannot be assumed. Beck and Katz (1995) show that the FGLS method produces a calculation of the standard errors, which underestimates the true variability of the coefficients obtained using it.

  11. 11

    As mentioned earlier, the estimation has been performed considering the existence of regional fixed effects. It has also been tested with the introduction of temporal effects, but the performance of an F-test using the distinct restricted values of R2, which emerged with the alternative estimations, suggests that the time effect is not significant in the specification employed here (F[10,131] = 0.0012; Prob > F ≅ 1.00).

  12. 12

    Alternative specifications of the variable Online have been experimented with, using dummies to reflect distinct moments of the appearance of this effect, but evidence of statistical significance is only found if the effect began in 2006.

  13. 13

    Alternative constructions of this variable of differential taxation with neighboring territories have also been employed, maintaining the result presented in the text. For example, an alternative variable has been considered, expressed as the ratio between the rate in the adjoining Community with lower taxation and the rate of the Community itself. Once more, the nonsignificance of the coefficient estimated for this variable confirms the nonpresence of the cross-border effect.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Review of the Literature on Cross-Border Gaming
  5. Empirical Analysis
  6. Concluding Remarks
  7. References
  • Alm, J., and M. Melnik. 2005. Sales taxes and the decision to purchase online. Public Finance Review 33(2): 184212.
  • Anselin, L. 1988. Spatial econometrics: Methods and models. Dordrecht: Kluwer Academic Publishers.
  • Anselin, L., A.K. Bera, R. Florax, and M.J. Yoon. 1996. Simple diagnosis tests for spatial dependence. Regional Science and Urban Economics 26: 77104.
  • Ballard, C.L., and J. Lee. 2007. Internet purchases, cross-border shopping, and sales tax. National Tax Journal 60(4): 711725.
  • Beck, N., and J.N. Katz. 1995. What to do (and not to do) with time-series cross-section data. The American Political Science Review 89(3): 634647.
  • Cliff, A., and J.K. Ord. 1981. Spatial processes, models, and applications. Londres: Pion.
  • Comisión del Mercado de las Telecomunicaciones. 2000–2010. Informe sobre el comercio electrónico en España a través de entidades de medios de pago. http://www.cmt.es/informes-de-comercio-electronico (accessed November 2013).
  • Elliott, D.S., and J.C. Navin. 2002. Has riverboat gambling reduced state lottery revenue? Public Finance Review 30: 235247.
  • Fundación Orange. 2001–2010. eEspaña. Informe anual sobre el desarrollo de la sociedad de la información en España. Madrid: Fundación Orange. http://www.fundacionorange.es/fundacionorange/analisis/eespana/e_espana10.html (accessed November 2013).
  • Garrett, T.A., and C.C. Coughlin. 2007. Inter-temporal differences in the income elasticity of demand for lottery tickets. Working Paper 2007-042C, Federal Reserve Bank of Saint Louis.
  • Garrett, T.A., and T.L. Marsh. 2002. The revenue impacts of cross-border lottery shopping in the presence of spatial autocorrelation. Regional Science and Urban Economics 32(4): 501519.
  • Goolsbee, A. 2000. In a world without borders: The impact of taxes on internet commerce. The Quarterly Journal of Economics 115(2): 561576.
  • Goolsbee, A., M. Lovenheim, and J. Slemrod. 2007. Playing with fire: Cigarettes, taxes and competition from the internet. Policy Paper 07-002, Stanford Institute for Economic Policy Research.
  • Knight, B., and N. Schiff. 2010. Spatial competition and cross-border shopping: Evidence from state lotteries. NBER Working Paper 15713.
  • Leal, A., J. López-Laborda, and F. Rodrigo. 2009. Prices, taxes and automotive fuel cross-border shopping. Energy Economics 31(2): 225234.
  • Leal, A., J. López-Laborda, and F. Rodrigo. 2010. Cross-border shopping: A survey. International Advances in Economic Research 16(2): 135148.
  • Ministry of the Interior, Ministerio del Interior. 1999–2010. Memorias e informes anuales del juego en España. http://www.mir.es (accessed November 2013).
  • Pisati, M. 2001. Tools for spatial data analysis. Stata Technical Bulletin 60: 2137.
  • Skidmore, M., and M.S. Tosun. 2008. Do new lottery games stimulate retail activity? Evidence from West Virginia counties. Journal of Regional Analysis and Policy 38(2): 4555.
  • Stitzel, L., and R. Pjesky. 2011. Lottery revenue and cross-border shopping: A nation-wide analysis. Journal of Business and Public Affairs 5(1): 15.
  • Tosun, M.S., and M. Skidmore. 2004. Interstate competition and lottery revenues. National Tax Journal 57(2): 163178.
  • Walker, D.M., and J.D. Jackson. 2008. Do U.S. gambling industries cannibalize each other? Public Finance Review 36: 308333.
  • Wooldridge, J.M. 2002. Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press.