categories of gambling expenditure are examined, namely: lottery tickets, lotto type ... activities in Australia poker machines and other gaming devices, lotteries, ...
SOCIOECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA ANDREW WORTHINGTON* School of Economics and Finance, Queensland University of Technology
KERRY BROWN & MARY CRAWFORD School of Management, Queensland University of Technology
DAVID PICKERNELL Business School, University of Glamorgan
Abstract Regression modelling is used to predict gambling patterns in Australia on the basis of the unit record files underlying the Australian Bureau of Statistics’ Household Expenditure Survey of 6,892 households. Eight categories of gambling expenditure are examined, namely: lottery tickets, lotto type games and instant lottery (scratch cards), TAB and related on course betting, poker machines and ticket machines, blackjack, roulette and other casino-type games, TAB-betting (excluding animal racing), club and casino broadcast gaming and gambling not elsewhere classified. Determining factors analysed include the source and level of household income, family composition and structure, welfare status, gender, age, ethnicity and geographic location. Apart from the determinants of expenditure varying widely across the different types of gambling activity, the results generally indicate that the source of household income is more important than the level of income and that household composition and regional location are likewise significant in determining gambling expenditure. Key words: Gambling expenditure, socioeconomic and demographic characteristics, cross-sectional analysis. JEL classification: C31, C35, D12, H22
Introduction In size and importance the Australian gambling industry has grown significantly over the last three decades. During this time there has been a fourfold increase in real gambling turnover, now more than $95 billion, in real gambling expenditure, currently some $821 per person, and in government revenue, at present accounting for some $3,850 million in gambling-related taxation or about 10 percent of State government revenues. At present, the main gambling activities in Australia poker machines and other gaming devices, lotteries, Totaliser Agency Board (TAB) betting, casino-type games and on-course totalisers, while the government revenues are obtained through fees and taxes on subscriptions, duties on gaming machines in clubs and hotels, taxes on turnover, excess and player loss, and licensing fees and charges on these activities. It is well recognised that all Australian governments have played a major role in this phenomenal growth in first, the legalisation, design and provision of gambling
*
Address for correspondence: Associate Professor Andrew C. Worthington, School of Economics and Finance, Queensland University of Technology, GPO Box 2434, Brisbane QLD 4001, Australia. Tel. +61 (0)7 3864 2658, Fax. +61 (0)7 3864 1500, email. [email protected]
2
ANDREW WORTHINGTON, KERRY BROWN & MARY CRAWFORD, DAVID PICKERNELL
activities, and second, the establishment of the differing revenue extraction devices. Accordingly, by some measures Australian are among the developed world’s most committed gamblers with per capita expenditure exceeding the US, Hong Kong and New Zealand and total expenditure of $10.8 billion greater than that spent in Australia on energy or household appliances and approaching the level set by alcohol. Unfortunately, the gambling industry and the public revenues it produces are accompanied by some undesirable socio-economic problems. For example, whilst gambling participation is voluntary, the pattern of expenditure may work to the relative detriment of low-income individuals and deepen the economic problems that must be addressed by other public support programs (Szakmary and Szakmary, 1995; Madhusudhan, 1996). For instance, there has been a steady increase in the percentage of household disposable income spent on gambling in Australia from less than one percent in the mid-1970s to currently in excess of three percent. Furthermore, there is evidence that certain forms of gambling have the capacity to create compulsive gambling and major addiction, attract criminal elements and foster corruption (Mason et al. 1989; Mikesell and Pirog-Good, 1990). For example, a recent survey of household gambling in Queensland identified 0.83 percent of the population (or 21,910 persons) as problem gamblers, 2.70 percent (71,227 persons) with moderate risk gambling, and a further 8.18 percent (215,824 persons) with low risk gambling (Gambling Policy Directorate 2001). Equivalently, a survey by the Productivity Commission (2001) concluded that problem gamblers constituted 15% of regular (non-lottery) gamblers, losing on average $12,000 per annum as compared to $650 for non-problem gamblers. Moreover, persons associated within these groups are very often associated with yet other social problems including alcohol and drug abuse, depression and suicide and losing time from work or study. Public policy in Australia has now begun to recognise some of these problems; largely in response to a community backlash against the expansion of gambling opportunities [a survey conducted by the Productivity Commission (2001) indicated that 70 percent of respondents believed gambling did more harm than good and 92 percent opposed the introduction of new gambling venues and machines]. Currently, at least some proportion of tax levies is spent on harm minimisation, the treatment of problem gambling and research into the costs and benefits of gambling. Similarly, smoking bans and limits on hours, and the restriction on access to Automated Teller Machines (ATMs) within gambling premises are also representative of a differing approach to policy in this area (Matterson 2003) as are requirements for the display of information about the ‘price’ and nature of gambling products,
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
3
the provision of information about the risks of problem gambling, controls on advertising and pre-commitment options, including self-exclusion arrangements (Productivity Commission 2001). Nevertheless, there is the ongoing need to address the concerns of equity and efficiency relating to the incidence of gambling-related taxation, and the important question of the socioeconomic burden of gambling expenditure itself. Borg and Mason (1988), Borg et al. (1991; 1993), Scott and Garen (1994), Davis et al. (1992), Jackson (1994), Hansen (1995), Cooper and Cohn (1994), Rodgers and Stuart (1995), Scoggins (1995), Szakmary and Szakmary (1995), Madhusudhan (1996), Layton and Worthington (1999) and Worthington (2001) have all examined the efficiency of gambling-related taxation as a means of fiscal extraction with an emphasis on maximising state revenue through the design of suitable tax structures and products. More recently, research has focused on the wider interpretation of demographic and socioeconomic incidence as a means of providing policy input into problem gambling and other undesirable consequences of the expansion in the gambling industry. Work in this area includes Garrett and Marsh (2002), Delfabbro and Winefeld (1999a, 199b), Jacques et al. (2000), Grun and McKeigue (2000) and Stanley and French (2003). The purpose of the present paper is to add to this literature the results of an analysis of gambling expenditure by Australian households. The Australian Bureau of Statistics’ (ABS) (2002) Household Expenditure Survey (HES) Confidentialised Unit Record File (CURF) is employed for this purpose. This survey focuses on both expenditure decisions by households, as well as the socioeconomic and demographic characteristics of these households. It thereby provides an important input into current economic policy regarding gambling behaviour in Australia and an indication of the benefits and costs associated with its use. The remainder of the paper is divided into three main areas. The first section explains the empirical methodology and data collection employed in the analysis. The second section discusses the results. The paper ends with some brief concluding remarks. Research method and data The analytical technique employed in the present study is to specify expenditures on various categories of gambling as the dependent variable (y) in a least squares regression with socioeconomic and demographic characteristics as explanatory variables (x). The variables used to estimate this model are detailed in Table 1. All data is obtained from the Australian Bureau of Statistics’ (ABS) 1998/99 Household Expenditure Survey Confidentialised Unit
4
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
Record File (CURF) and relate to a sample of 6,892 probability-weighted Australian households. The strength of this data is that it is a national survey and it provides expenditure data for different forms of gambling by a range of characteristics. However, it has two weaknesses when used to analyse gambling data (Productivity Commission 2001). First, with the rapid growth in the Australian gambling industry, the data from the Household Expenditure Surveys in 1993/94 and now 1998/99 predate some of the growth of Australian casinos and the expansion of gaming machines in Queensland, South Australia, Victoria and Tasmania. Second, the data is thought to understate the actual level of gambling expenditure in Australia. This is likely associated with the selective recall of gambling expenditures over the two-week survey period and the relatively uneven distribution of gambling expenditures throughout the year.
In terms of the dependent variables, eight categories of household weekly gambling expenditure are employed. These are: Lottery tickets (LOT), lotto type games and instant lottery (scratch cards) (LTO), TAB and related on course betting (TAB), poker machines and ticket machines (POK), blackjack, roulette and other casino-type games (BLK), TAB-betting (excluding animal racing) (BET), club and casino broadcast gaming (CLB) and gambling not elsewhere classified (NEC). Whilst few studies have employed more than a single expenditure classification as the dependent variable, the definitions adopted are consistent with Scott and Garen’s (1994) and Kitchen and Powell’s (1991) respective analyses of lotteries in Kentucky and Canada, Thiel’s (1991) inquiry into Washington’s Lotto, Hansen’s (1995) study of Colorado instant lotteries, and Pugh and Webley’s (2000) research into UK lottery participation, amongst others. In terms of average weekly gambling expenditure, LTO is highest ($3.43), followed by POK ($2.14), TAB ($1.47), NEC ($1.07), BLK ($0.25) and BET ($0.08). By and large, the distributional properties of the gambling expenditures appear non-normal. All of the expenditures are positively skewed, ranging from 4.5806 (LTO) to 53.9838 (NEC), indicating the greater probability of larger than lower expenditures on gambling. The asymptotic sampling distribution of skewness is normal with mean 0 and standard deviation of
6 T , where T is the sample size. Since the sample size for all the return series is 6,892
then the standard error under the null hypothesis of normality is 0.0295: all estimates of skewness are significant at the .10 level or higher. The kurtosis, or degree of excess, in all
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
5
gambling expenditures is also large, ranging from 38.3473 for LTO to 3591.7130 for NEC, thereby indicating leptokurtic distributions. Given the sampling distribution of kurtosis is normal with mean 0 and standard deviation of
24 T = 0.0591 , then all estimates are once
again statistically significant at any conventional level. Finally, Jarque-Bera statistics and pvalues (not shown) are used to test the null hypotheses that the distribution of household gambling expenditures is normally distributed. All p-values are smaller than the .01 level of significance suggesting the null hypothesis can be rejected. The set of socio-economic and demographic variables upon which the household gambling expenditures are regressed are also included in Table 1. Whilst there is no unequivocal rationale for predicting the direction and statistical significance of many of these independent variables, their inclusion is consistent with both past studies of gambling-taxation and demographic/socioeconomic incidence and the presumed interests of policy-makers and other parties. However, the effects are expected to vary across the various categories of gambling expenditure. As an example, the Productivity Commission (2001) found that gaming machine players are slightly biased towards middle income earners and those aged between 18 and 24; racing punters are slightly biased towards males, middle income earners and those aged between 18 and 34; the profile of lottery gamblers reflects that of the general population with a small bias towards people aged between 50 and 64 and people with higher incomes; gamblers on casino table games have a strong bias towards males, singles, and those aged between 18 and 24; sports gamblers are strongly biased towards males, people aged between 18 and 24, people with higher incomes, and singles; and gamblers that play games privately for money are biased towards males, people aged between 18 and 24, and singles. Similarly, included in the Productivity Commission (2001: 14) report were findings drawn from the Burswood and Star City Casinos that “...casino patrons in general come from a wide range of backgrounds. All age groups are well represented and there is an even distribution between male and female casino patrons. The majority of casino patrons are married and come from a blue or white-collar background. Unemployed, home duties, students, pensioners and retirees are less represented” and though “There is no typical gambler although there may be a preponderance of type in certain forms of gambling which may relate to preference, cost and availability”. These included findings that: 60 per cent of customers were male; 39 per cent are from Asian backgrounds; 44 per cent are aged under 35, 21 per cent are aged 35-44 while 35 per cent are aged over 45; and mature singles and older couples are the most likely to visit the casino and young families are least likely.
6
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
The first four variables relate to household structure. These represent households composed of a single person living alone (PRS), a couple with no other usual residents (CPL), a couple with children (CPC) and a lone parent with children (LNP). As a rule, it is usually found that single persons and lone parents with children are overrepresented in most studies of gambling incidence. A positive coefficient is hypothesised when gambling expenditure is regressed against PRS and LNP. The next four variables correspond to the gender of the household reference person (SEX) and their age (ATF, AFS, and AMS) [see, for instance, Kitchen and Powells (1991), Pugh and Webley (2000), Hira and Monson (2000) and Götestam and Johansson (2003)]. All other things being equal, households headed by females generally spend less on gambling than male household heads while many categories of gambling are biased towards participation by younger gamblers, including BLK, TAB and NEC; middleaged gamblers tend to focus on LTO and LOTT; and older gamblers are concentrated in expenditures relating to POK and CLB. The following eight variables all relate to the ethnic background of the household. It is generally observed that different ethnic groups may be focused towards particular types of gambling as against others (Worthington 2001). The variables specified include whether the household head (and spouse of the household head) was born in northwest, southern or eastern Europe (EHH and ESP), North Africa and the Middle East (MHH and MSP), southeast, north-east, southern and central Asia (AHH and ASP) or the Americas or SubSaharan Africa (OHH and OSP). Finally, four variables are included to reflect additional dimensions of household structure and characteristics. These are: the number of usual residents in the household (NUR), the proportion of female spenders in the household (FEM), the proportion of dependents in the household (DEP) and the proportion of retirees in the household (RET). The hypotheses are that households with a higher NUR are generally associated with larger expenditures on gambling, whereas high levels of FEM, DEP and RET are associated with lower expenditures. The next group of variables relate to both the level of weekly household income (INC) and the sources of this income, and follow the work of Kitchen and Powells (1991) and Pugh and Webley (2000), amongst others. For the former, the level of expenditure on gambling products is posited to increase with income, though at a diminishing rate. In the case of the latter, Scott and Garen (1994) and Worthington (2001) have discussed the purported impact of welfare recipiency on gambling expenditures. It is posited that even after holding household income constant, certain groups of welfare recipients may engage in a disproportionate
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
7
amount of gambling expenditure. The qualitative (dummy) variables included to test this hypothesis are firstly whether the household in question derives the larger portion of its income from self employment (SEL), superannuation, investments and other private income (INV) and age, disability, unemployment, education, sickness and other pensions and benefits (PEN). A measure of relative socio-economic disadvantage (SOC) is also included to proxy additional dimensions relating to each household’s socioeconomic position. The final set of information comprises several dummy variables reflecting each respondent’s regional location: namely, New South Wales (NSW), Victoria (VIC), Queensland (QLD), South Australia (SA), Western Australia (WA) and Tasmania (TAS). The justification behind these measures is that the characteristics of the gambling industry and its products vary substantially across state borders, largely due to the differing pace of legalisation and liberalisation of state-regulated gambling. For example, New South Wales has had gaming machines for over 40 years. In contrast, this form of gambling has only relatively recently been introduced in Victoria, South Australia and Queensland while Western Australia has no gaming machines outside of casinos. However, while New South Wales has a disproportionate number of machines, Victoria has a disproportionate share of gaming machine revenue since the imposition of a cap of gaming machine numbers in 1996 has seen machines relocated to sites with higher turnover. Similarly, Tasmania has had a casino for over 25 years, whereas casinos are still a relatively recent development in New South Wales and Victoria. Nonetheless, New South Wales has a disproportionately higher share of casino expenditures, largely through a greater share of international visitors. However, on a per capita basis the Northern Territory has much higher expenditures on casino-type games. Moreover, the characteristics of gambling products, and hence the demand for them, vary greatly across the various states and territories. These typically include the price of the product, the size of the prize, the odds of winning, the extent to which odds can be changed by skill, the accessibility of the product, the experiences associated with the venue and the social acceptability of the activity. The signs and significance of the coefficients on regional location will therefore necessarily depend on the interaction between numbers of factors. Empirical findings The estimated coefficients, standard errors and p-values of the parameters for the least squares regressions are provided in Table 2. Also included in Table 2 are F-statistics for the null
8
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
hypothesis that all the slope coefficients are jointly zero and Akaike Information Criterion (AIC) and Schwartz Information Criterion (SIC) as a guide to model selection. Sixteen separate models are estimated. The estimated coefficients, standard errors and p-values employing the full set of demographic, socioeconomic and regional characteristics as predictors for LOT, LTO, TAB, POK, BLK, BET, CLB and NEC are shown in columns 1 to 3, 7 to 9, 13 to 15, 19 to 21, 25 to 27, 31 to 33, 37 to 39 and 43 to 45. The results of estimations for a refined specification for each category of gambling expenditure based on forward stepwise regression using an F-criterion are shown in the remaining columns. Six variables (excluding the constant) are stepped into the LOT and TAB models on this basis, sixteen for LTO, seven for POK, four for BLK, three for BET and CLB and two for NEC.
All models were initially estimated using conventional least squares assumptions, however White’s heteroskedasticity test of the null hypothesis of no heteroskedasticity was rejected (statistic and p-value in brackets) for the models specifying LOT (1.6124, 0.0107), LTO (2.7023, 0.0000), TAB (6.0390, 0.0000), BLK (1.7878, 0.0023) and BET (6.0399, 0.0000) as regressands. Accordingly, the estimates for POK, CLB and NEC incorporate the usual ordinary least squares standard errors, whereas the standard errors for LOT, LTO, TAB, BLK and BET incorporate White’s corrections for heteroskedasticity of unknown form. Most of the estimated models in the full specifications and all of the models in the refined specifications are highly significant, with tests of the hypotheses that all of the slope coefficients are zero rejected at the 1 percent level or lower using the F-statistic. However, in all instances the lower values of AIC and SIC indicate that in terms of the trade-off between comprehensiveness and parsimony the refined specifications for each type of gambling expenditure are preferred to the full specification. In the interests of brevity, discussion of the individual estimated coefficients will only be made in reference to these refined models. The results in these models also appear sensible in terms of both the precision of the estimates and the signs on the coefficients. To test for multicollinearity, variance inflation factors (VIF) are calculated and presented in Table 1. As a rule of thumb, a VIF greater than ten indicates the presence of harmful collinearity. Amongst the explanatory variables the highest VIFs are for NUR (5.4746), PRS (4.8860), DEP (4.7912) and CPC (4.5500). This suggests that multicollinearity, while present, is not too much of a problem. The first model discussed is that determining gambling expenditure on lottery tickets (LOT). The estimated coefficients for PRS, CPC, AFS, DEP, NSW and SA are significant at the 10
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
9
percent level of significance or lower. This suggests that persons living alone and households with a higher proportion of dependents expend less on lottery tickets, while couples with children, households headed by a person aged between 50 and 69 years and NSW and South Australian households expend relatively more. The greatest marginal effects on Australian household’s expenditure on lottery tickets are the proportion of dependents in the household and whether that household is located in NSW. These results differ greatly to the relatively more complex model for lotto type games and instant (scratch) lotteries (LTO). In that regression, the estimated coefficients for PRS, CPC, ATS, AFS, AMS, ASP, UR, DEP, RET, SEL, INV, PEN, SOC, NSW, SA and WA are all significant at the .10 level or lower. The results indicate, in addition to those presented for LOT, that those households with heads between thirty and forty years, with a spouse born in Asia, and with the major portion of household income sourced from self-employment, investments and superannuation and pensions or with a lower level of relative socioeconomic disadvantage spend proportionately less on lotto and instant lottery type games. As a rule, the numbers of estimated coefficients in the refined specification for the remaining forms of gambling expenditure are lower than that found for lottery tickets and lotto type games and instant lottery. For TAB on-course and related betting gambling expenditure is relatively higher for households with a head aged between fifty and sixty-nine years (AFS) and those with a greater proportion of members who are retired (RET) and on a slightly higher income (INC), whereas expenditures are lower for households with heads born in Europe (EUR) or with a greater portion of members who are female (FEM) or dependents (DEP). For poker machine (POK) expenditures are positively associated with the number of usual residents (NUR) and households deriving a greater proportion of their income from investments and superannuation (INV) and negatively associated with a high proportion of dependents (DEP), pension and government benefit income sources (PEN), a relatively lower level of socioeconomic disadvantage (SOC) and those households living in Western Australia (WA). Blackjack, roulette and other casino games (BLK) are lower for couples with children (CPC) and pensioners and others on government benefits (PEN) and higher for households with an Asian born head, while TAB-betting (excluding animal racing) is linked with couples with children (CPC) and lone parents (LNP) though with a relatively lower level of dependents (DEP). Finally, club and casino broadcast gaming (CLB) and other gambling (NEC) is positively associated with couples (CPC), while CLB is negatively associated with
10
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
Victorian households (VIC) and NEC with lower levels of relative socioeconomic disadvantage (SOC). Concluding remarks and policy recommendations The present study uses regression analysis to investigate the determinants and incidence of gambling expenditures in Australian households. The current paper extends empirical work in this area in at least three ways. First, and as far as the authors are aware, it represents the first attempt to test the purported socioeconomic and demographics determinants of gambling expenditures across all states in Australia using the latest Household Expenditure Survey (HES). The evidence provided suggests that, on average, participation in manifest gambling activities such as lotteries, Lotto and instant lotteries, TAB and on-course betting, and is strongly influenced by demographics such as age, ethnicity and household composition. However, these effects vary widely across the many types of gambling activity. Second, the study analyses in detail the posited linkage between expenditures on gambling and implied tax incidence. The results indicate that the incidence of gambling-related taxation is only mildly regressive; that is, gambling expenditures as a percentage of income decline as income increases; and that this incidence is only statistically significant in the case of TAB on-course and related betting. This has at least some implications for the use of gamblingrelated taxation as a means of fiscal extraction. However, factors other than income level are also at play in determining gambling expenditures. More particularly, rather than the level of income a more pertinent factor in determining the level of gambling expenditure in Australia is the primary source of this income, whether salaries and wages, self-employment, investments and superannuation, and pensions and other government benefits. Perhaps the most striking influence on gambling expenditure in Australian households is the role of household structure and composition and regional location. In the former case, this says much about the demands of families and the ability to engage in certain types of gambling. For example, couples with children are likely to have higher expenditures in gambling products that can be consumed at home such lottery tickets and lotto type games and instant lotteries than blackjack, roulette and other casino type games, though higher dependency ratios is general are associated with lower levels of expenditure. One factor at play here may be the lack of child-minding facilities at casinos for instance. The other interesting feature is the close link between certain forms of gambling expenditure and regional location. This, of course, bears much relation to the role of the states in legalising
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
11
and regulating gambling activity. For example, NSW households spend relatively more on poker machines and relatively less on lotto type’s games and instant lotteries, while Western Australian households are the reverse. This indicates that governments can do much to change the pattern of gambling expenditure by modifying both the characteristics of the gambling activities and the goods consumed in common with these activities. References Australian Bureau of Statistics (2002). 1998-99 Household Expenditure Survey Australia, Confidentialised Unit Record File (CURF) Technical Paper, 2nd ed. ABS cat. No. 6544.0.30.001. Beenstock, M. Goldin, E. and Haitovsky, Y. (2000). What Jackpot? The Optimal Lottery Tax”. European Journal of Political Economy 16, pp. 655-671. Borg, M.O. and Mason, P.M. (1988). “The Budgetary Incidence of a Lottery to Support Education”. National Tax Journal 41, 75–85. Borg, M.O. Mason, P.M. and Shapiro, S.L. (1991). “The Incidence of Taxes on Casino Gambling: Exploiting the Tired and Poor”. American Journal of Economics and Sociology 50, 323–332. Borg, M.O. Mason, P.M. and Shapiro, S.L. (1993). “The Cross Effects of Lottery Taxes on Alternative State Tax Revenue”. Public Finance Quarterly 21, 123–140. Cook, M. McHenry, R. and Leigh, V. (1998) Personality and the National Lottery, Personality and Individual Differences 25, 49-55. Cooper, S. and Cohn, E. (1994). “Equity and Efficiency of State Lotteries: Review Essay”. Economics of Education Review 13, 355–362. Davis, J.R. Filer, J.E. and Moak, D.L. (1992). “The Lottery as an Alternative Source of State Revenue”. Atlantic Economic Journal 20, 1–10. Delfabbro, P.H. and Winefeld, A.H. (1999) “Poker Machine Gambling: An Analysis of Within-Session Characteristics, British Journal of Psychology 90, 425-439. Delfabbro, P.H. and Winefeld, A.H. (1999) “The Danger of Over-Explanation in Psychological Research: A Reply to Griffiths”, British Journal of Psychology 90, 447-450. Farrell, L. and Walker, I. (1999). The Welfare Effects of Lotto: Evidence from the UK”. Journal of Public Economics 72, 99-120. Gambling Policy Directorate (2001) Short Analysis of Queensland and Australian Gambling Statistics, Queensland Government Treasury. Garrett, T.A. (2001). “An International Comparison and Analysis of Lotteries and the Distribution of Lottery Expenditure”. International Review of Applied Economics 15, 213-227. Garrett, T.A. and Marsh, T.L. (2002) The Revenue Impacts of Cross-Border Lottery Shopping in the Presence of Spatial Autocorrelation, Regional Science and Urban Economics 32, 501-519. Garrett, T.A. and Sobel, R.S. (1999) “Gamblers Favour Skewness, not Risk: Further Evidence from United States’ Lottery Games”. Economics Letters 63, 85-90. Götestam, K.G. and Johansson, A. (2003). “Characteristics of Gambling and Problematic Gambling in the Norwegian Context: A DSM-IV-Based Telephone Interview Survey”. Additive Behaviours 28, 189-197. Grun, L. and McKeigue, P. (2000) “Prevalence of excessive Gambling Behaviour Before and After Introduction of a National Lottery in the United Kingdom: Another Example of the Single Distribution Theory”. Addition 95, 959-967. Hansen, A. (1995). “The Tax Incidence of the Colorado State Lottery Instant Game”. Public Finance Quarterly 23, 385–398. Hira, T.K. and Monson, K.W. (2000) A Social Learning Perspective of Gambling Behaviour Among College Students at Iowa Sates University, USA”. Journal of Consumer Studies and Home Economics 24, 1-8. Jackson, R. (1994). “Demand for Lottery Products in Massachusetts”. The Journal of Consumer Affairs 28, 313– 325.
12
SOCIO-ECONOMIC AND DEMOGRAPHIC DETERMINANTS OF HOUSEHOLD GAMBLING IN AUSTRALIA
Jacques, C. Ladouceur, R. and Ferland, F. (2000) “Impact of Availability on gambling: A Longitudinal Study”. Canadian Journal of Psychiatry 45, 810-816. Kitchen, H. and Powells, S. (1991). “Lottery Expenditures in Canada: A Regional Analysis of Determinants and Incidence”. Applied Economics 23, 1845–1852. Ladouceur, R. (1996) “The Prevalence of Pathological Gambling in Canada”. Journal of Gambling Studies 12, 129-142. Layton, A.P. and Worthington, A.C. (1999). “The Impact of Socioeconomic Factors on Gambling Expenditure” International Journal of Social Economics 26, 430–440. Madhusudhan, R.G. (1996). “Betting on Casino Revenues: Lessons from State Experiences”. National Tax Journal 49, 401–412. Matterson, H. (2003) “A Push of the Panic Button on Gambling”. The Weekend Australian June 28-29, p. 30. Productivity Commission (1999) Public Inquiry into the Australian Gambling Industry: Final Report. Productivity Commission: Canberra. Pugh, P. and Webley, P. (2000). “Adolescent Participation in the U.K. National Lottery Games”. Journal of Adolescence 23, 1-11. Rodgers, W.M. and Stuart, C. (1995). “The Efficiency of a Lottery as a Source of Public Revenue”. Public Finance Quarterly 23, 242–254. Scoggins, J.F. (1995). “The Lotto and Expected Net Revenue”. National Tax Journal 48, 61–70. Scott, F. and Garen, J. (1994). “Probability of Purchase, Amount of Purchase, and the Demographic Incidence of the Lottery Tax”. Journal of Public Economics 54, 121–143. Scott, F.A. and Gulley, O.D. (1995). “Testing for Efficiency in Lotto Markets”. Economic Inquiry 33, 175–188. Stanley, R.E. and French, P.E. (2003). “Can Students Truly Benefit from State Lotteries: A Look at Lottery Expenditures Towards Education in the American States”. Social Science Journal 40, 327-333. Szakmary, A. and Szakmary, C.M. (1995). “State Lotteries as a Source of Revenue: A Re-examination”. Southern Economic Journal 61, 1167–1181. Thiel, S.E. (1991). “Policy, Participation and Revenue in Washington State Lotto”. National Tax Journal 44, 225–235. Vasche, J.D. (1990). “The Net Revenue Effect of California’s Lottery”. Journal of Policy Analysis and Management 9, 561–564. Volberg, R. (1996) “Prevalence of Problem Gambling in the United States”. Journal of Gambling Studies 12, 111-128. Worthington, A.C. (2001) “Implicit Finance in Gambling Expenditures: Australian Evidence on Socioeconomic and Demographic Tax Incidence” Public Finance Review 29, 326–342.
TABLE 1 Dependent and independent variable definitions and descriptive statistics Variable description Expenditure on selected classifications of gambling Lottery tickets Lotto type games and instant lottery (scratch cards) TAB on course betting and related Poker machines and ticket machines Blackjack, roulette and other casino-type games TAB-betting (excluding animal racing) Club and casino broadcast gaming Gambling not elsewhere classified Household structure Person living alone Couple, no other usual residents Couple with children Lone parent with children Sex of household reference person Age of household reference person Thirty to forty-nine years Fifty to sixty-nine years More than seventy years Country of birth of the household reference person North-west, southern and eastern Europe North Africa and Middle East South-east, north-east, southern and central Asia Americas, Sub-Saharan Africa Country of birth of spouse of the household reference person North-west, southern and eastern Europe North Africa and Middle East South-east, north-east, southern and central Asia Americas, Sub-Saharan Africa Number of usual residents in the household Proportion of female spenders in the household Proportion of dependents in the household Proportion of retirees in the household Weekly household income from all sources Principal source of household income Self employed Superannuation, investments and other private income Age, disability, unemployment, education, sickness pensions Index of relative socio-economic disadvantage State/territory of enumeration New South Wales Victoria Queensland South Australia Western Australia Tasmania
Notes: Dependent variables are Australian dollar weekly gambling expenditures on LOT, LTO, TAB, POK, BLK, BET, CLB and NEC. The control for household structure (PRS, CPL, CPC, LNP) is mixed and other families. The control for sex of the household reference person (SEX) is male. The control for the age of the household reference person (ATF, AFS, AMS) is twenty-nine years or less. The control for the country of birth of the household reference person (EHH, MHH, AHH, OHH) is Australia and Oceania. The control for the country of birth of the household reference person’s spouse (ESP, MSP, ASP, OSP) is Australia and Oceania. The proportion of female spenders (FEM), dependents (DEP) and retirees (RET) in the household is in reference to the number of usual residents (NUR). The control for principal source of household income (SEL, INV, PEN) is salaries and wages. The index for the index of relative socio-economic disadvantage is in deciles ranked from a higher level to a lower level of socioeconomic disadvantage. The control for the state/territory of enumeration (NSW, VIC, QLD, SA, WA, TAS) is the Australian Capital Territory and the Northern Territory. The critical values for skewness and kurtosis at the .05 level are 0.0578 and 0.1156, respectively.
Notes: (a) Estimated coefficients, standard errors and p-values are given for a full and a refined specification for the dependent variables, LOT, LTO, TAB, POK, BLK, BET, CLB, NEC. The full specification models are obtained by including all the independent variables in Table 1; the refined specification models are obtained by using forward stepwise regression using an F-criterion. AIC – Akaike Information Criteria, SIC – Schwartz Information Criteria.
