Hopkins, S.,Kidd, M. P. (1996). The determinants of the demand

Applied Economics ISSN: 0003-6846 (Print) 1466-4283 (Online) Journal homepage: http://www.tandfonline.com/loi/raec20 The determinants of the demand for private health insurance under Medicare Sandra Hopkins & Michael P Kidd To cite this article: Sandra Hopkins & Michael P Kidd (1996) The determinants of the demand for private health insurance under Medicare, Applied Economics, 28:12, 1623-1632, DOI: 10.1080/000368496327598 To link to this article: https://doi.org/10.1080/000368496327598 Published online: 01 Oct 2010 Submit your article to this journal Article views: 146 View related articles Citing articles: 38 View citing articles Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=raec20 Applied Economics, 1996, 28, 1623Ð 1632 The determinants of the demand for private health insurance under Medicare S A N D R A H O P K I N S * and M I C H A EL P K I D D § *School of Economics and Finance, Curtin University of T echnology, GPO Box U 1987, Perth W A 6001 and § Department of Economics, University of T asmania, GPO Box 252C, Hobart, T asmania 7001 , Australia and Department of Economics, University of Aberdeen, Dunbar Street, Old Aberdeen, AB24 3QY Since the introduction of Medicare in 1984, the proportion of the Australian population with private health insurance has declined considerably Insurance for health care consumption is compulsory for the public health sector but optional for the private health sector In this paper, we explore a number of important issues in the demand for private health insurance in Australia The socio-economic variables which in¯ uence demand are examined using a binary logit model A number of simulations are performed to highlight the in¯ uence and relative importance of various characteristics such as age, income, health status and geographical location on demand A number of important policy issues in the private health insurance market are highlighted First, evidence is provided of adverse selection in the private health insurance pool, second, the notion of the wealthy uninsured is refuted, and ® nally it is con® rmed that there are signi® cant interstate di€ erences in the demand for private health insurance I INTRODUCTION Medicare was established in Australia in 1984 It is a universal, tax-® nanced public health insurance system which provides a basic standard of care in public hospitals for all Australians and subsidizes the cost of medical care by private medical practitioners outside hospitals.1 Doessel (1992) suggests that a private health insurance policy is an imperfect substitute for Medicare which enables a policyholder to have access to alternative suppliers for the same type of treatment or care which is available under Medicare and subsidizes services not covered by Medicare.2 In the ® rst case, private health insurers compete with the public health insurer as well as each other; in the second case, the private health insurers compete with one another only The purchase of private insurance by an individual does not change the tax liability to the public health insurance system and therefore the private insurance policyholder has access to both the public and private health care sectors Prior to 1984, all health insurance was private and voluntary and consequently, 64% of the Australian population had private health insurance.3 Upon the introduction of Medicare in 1984, this percentage declined to 50 By 1990, the percentage had declined even further to 43.1 (Willcox, The hospital and medical bene® ts of the public health insurance system represent the main features of Medicare There are a number of additional bene® ts which are not mentioned here for example, optometry examinations `Basic’ or `supplementary’ private insurance supplements the hospital component of Medicare It allows patients to have choice of their own doctor in public hospitals and subsidizes the cost of private hospital care `Ancillary’ private health insurance covers the basic requirements of private hospital care in addition to subsidizing ancillary medical care sought outside the public hospital system such as physiotherapy, and dental care Medibank was introduced by the Labour Government in 1975 and provided universal cover for free public hospital treatment and medical bene® ts were covered at 85% of the government recommended schedule fee with a maximum $5 gap Medibank was progressively dismantled when the LiberalÐ National Party Government came to power in 1975 From September 1981 until the introduction of Medicare in 1984, Commonwealth government support for medical and hospital costs was con® ned to tax rebates for basic health insurance premiums and a Commonwealth subsidy towards medical costs covered by private health insurance Pensioners and the disadvantaged were covered by a government health insurance scheme 0003Ð 6846 Ó 1996 Routledge 1623 1624 1991) The decline in the demand for private health insurance raises important questions as to the reason why people purchase private health insurance when public health insurance is compulsory Willcox (1991) suggests that the three main considerations in the decision-making process about private health insurance are health status, cost and inadequacy of the public health insurance coverage Econometric analyses of the determinants of private health insurance in both Australia (Ngui et al., 1990; and Cameron and Trivedi, 1991) and the United Kingdom (Propper, 1989) as well as descriptive analyses of the Australian market (Australian Bureau of Statistics, 1990; Willcox, 1991) have nominated a consistent set of socio-economic and demographic factors as the major in¯ uences on demand A similar set of determinants in Australia and the United Kingdom is important as there are a number of similarities between the two countries in the public funding and provision of health care The United Kingdom like Australia, has a universal, tax-® nanced public health insurance system where excess demand is rationed in the public system by queuing, especially for elective surgery, rather than copayments.4 In the paper, we explore a number of important issues in the demand for private health insurance in Australia The ® rst contribution of this paper is an examination in an econometric framework of the relationship between the demand for private health insurance and the determinants of the demand The econometric analysis enables the exploration of not only what is important in determining demand but also the relative importance of competing determinants In so doing, we are able to address the important socioeconomic di€ erences in the demand for health insurance, including the in¯ uence of spousal characteristics on the insurance decision We also consider two key issues in the health insurance policy debate The ® rst, the relationship between private health insurance and income, is important in managing the queue to public sector health facilities The second key policy issue, the relationship between private insurance and health status, is important in terms of the long-term viability of the private health insurance market Government regulation prohibits private health insurance companies from discriminating statistically between high and low-risk groups Low-risk groups tend over time to drop out of the market as they perceive that the premium is in excess of their probability of loss The third contribution is a consideration of the regional di€ erences in the demand for private health insurance The S Hopkins and M P Kidd total Australian ® gures hide the considerable interstate variation in the demand for private health insurance For example, in 1990, approximately 31.6% of the Queensland population had private health insurance compared with 50.4% in Victoria Our analysis updates and extends the papers by Ngui et al (1990) and Cameron and Trivedi (1991) Ngui et al used 1983 Australian Health Survey data and socio-economic and health status variables to explain the insurance demand decision Cameron and Trivedi used the 1977Ð 78 Australian Health Survey and the 1983 Australian Health Insurance Survey Their data thus, covered two separate regimes of health insurance management Both papers considered the determinants of health insurance choice preMedicare Our paper uses the 1989 Health Survey data and also extends their analyses in a number of important ways First, we include geographical location variables as important determinants of the health insurance decision in addition to individual health status and socio-economic variables We also incorporate spousal variables in additional to individual health status characteristics to emphasize the importance of family rather than just individual characteristics in the insurance decision Finally, in addition to the logit analysis, we undertake a number of simulations which provide important and interesting additional information on the magnitude of the impact of a change in one of the explanatory variables on the insurance decision I I EX P E C T E D U TI L I T Y G A I N FR O M PR I V A T E H E A LT H I N S U R A N C E The theory of insurance has been applied extensively to the health insurance decision (Arrow, 1963; Feldstein, 1973) Under conditions of consumer rationality and risk aversity, the decision to purchase insurance is made on the basis of expected utility gain Individuals or family groups weigh up both the direct and indirect costs of the insurance premium against the expected bene® ts from a private health insurance state Both the direct and indirect costs and bene® ts are discussed below One bene® t of using the private sector for treatment rather than the public sector, which is di cult to quantify and therefore model, is the quality di€ erences between the private and public sector.5 The purchase of private insurance confers access to both the private and public health sector, whereas non-purchase Approximately, 10% of the British population has private health insurance The considerably smaller proportion of the British population with private health insurance re¯ ects a long-established and stable public health insurance system, the National Health Service, and therefore a small private sector and the use of risk-rating rather than community-rating in premium setting by British private insurance companies Hospital output is generally considered to be a Lancastrian good (Rice, 1966) One of the attributes is the `hotel’ services or amenities The amenities quality tends to vary between the public and private sector in terms of the general physical environment The second attribute is the medical treatment received This characteristic is assumed to be constant between the private and public hospitals as the medical sta€ work in both sectors and the medical, nursing and paramedical sta€ are trained in the public sector 1625 Demand for private insurance under Medicare restricts access to the public health sector only.6 Unlike other types of insurance, the expected loss for which the insurance is purchased is only incurred if one chooses to consume private sector resources The consumption and insurance decisions are inextricably bound together Private health insurance, thus, removes the uncertainty of a ® nancial loss in the event that both ill-health occurs and the private health sector is used for treatment.7 Di€ erent individuals or families put di€ erent weight on the costs and bene® ts of private insurance purchase and therefore of consuming private sector services For some, the choice set is restricted not because of income but because they would not contemplate the consumption of the services which are made available under private insurance and would possibly derive disutility rather than utility from their consumption The motivation for their restricted choice is possibly ideological Another group of people who may not actually make a choice between private and public health insurance are employees for whom private health insurance membership is part of their employment contract Expected utility gains from the purchase of private health insurance are in the ® rst instance related to expected medical need Some individuals face greater risk vulnerability than others due to their age, sex, pre-existing health status and marital status The probable distribution of future health states is based on present and past health states Medical need, generally, increases with age and is also higher for female gender Additionally, for many people, the purchase of private health insurance is a family rather than an individual decision, therefore, the demographic characteristics of the family unit are important An example is the impact that a dependent child has on the private health insurance decision A second group of factors which a€ ects the health insurance decision is what van de Ven and van Praag (1981) call material well-being The direct cost of private health insurance is the insurance premium itself An additional cost is the often considerable copayments which apply for the use of private sector facilities or doctor of your own choice in a public hospital There are two sources of utility gain These are the direct bene® t associated with the insurancesubsidized access to private sector treatment and the indirect bene® t of bypassing the public sector queue, which is largely for elective surgery, by entering the private sector The rationing of services in the public sector is done by either actual waiting time, for example, to consult with a doctor in an outpatients clinic or by queuing for service via a waiting list Most of the waiting time in the public sector is of the second type Both forms of time rationing, however, impose a cost on the individual and his/her family due to the likelihood that the medical problem and/or the associated discomfort may worsen during the waiting period and also due to the uncertainty of the timing of the medical intervention Ceteris paribus, the expected utility gain from bypassing the public sector queue would be greatest for people who place the highest value on the time of the individual as well as that of the family unit The value of time is most probably higher for those who are employed rather than unemployed or not in the labour force, and those on higher incomes rather than lower incomes The third group of factors, unlike expected medical need and material wellbeing, is speci® cally Australian and relates to the interstate di€ erences in the mix of private and public services The federal system in Australia means that the States and Territories have some autonomy over the organization and delivery of health services, but the overall structure is determined by the Commonwealth Government The autonomy of the States and Territories, however, has varied with changes in health ® nancing arrangements and the preferences of the political party in Government at the Federal level The autonomy has certainly diminished with the increasing focus on a national insurer, however the geographical di€ erences in quality and services available in the public sector have persisted Historical di€ erences in the interstate public provision of health services has led to di€ erences in the size and mix of services o€ ered by the private sector For example, the private bed to population ratio varies from 0.5 beds per 1000 in the Australian Capital Territory to 1.7 beds per 1000 in South Australia (National Health Strategy, 1991) The implication of the interstate di€ erences is that the queuing problem is more acute in some locations than others Therefore, by implication the expected utility gain from holding private health insurance is greater in some locations than others III DESCRIPTION OF THE DATA AND SAMPLE DER IVATION The focus of the current analysis is the decision to purchase private health insurance The data set utilized is the 1989Ð 90 Of course, it is possible that some individuals or family groups may self-insure In this case, they consume private sector facilities but not have private health insurance The data set does not allow us to identify cases of self-insurance Private health insurance in Australia covers the use of private health facilities as well as the admission by doctor of one’s own choice to a public hospital Ben-Akiva and Lerman (1985) suggest that coe cient estimates from a restricted choice set are uniformly greater in absolute value than those from the full choice set The interpretation of the results from the empirical analysis in Table focuses on the sign and signi® cance of the coe cient values only and therefore, this issue is not considered to be a problem here 1626 National Health Survey This is a representative sample of the Australian population which provides detailed information on a series of personal characteristics including age, education, state of residency, income, health status, consumption of health services and private health insurance status The raw data comprised 54 241 individual records In the initial sample, 37.7% have private health insurance and 35.2% not The remainder of the initial sample consist of individuals under the age of 15 who are not classi® ed with respect to insurance status After deletion of these individuals from the sample, 51.7% of eligible insurance holders have private health insurance.9 Many of those individuals under the age of 15 would be covered by the insurance policy of their parents and thus the higher percentage would appear to be appropriate Although the survey record data are at the individual level, individuals can be matched in terms of whether they belong to the same income unit and whether they are related by marriage (either de facto or de jure) An income unit is de® ned as consisting of a head of household plus his spouse and persons in the same family who are assumed to be dependent on the head including children under the age of 15 and full-time students living at home.1 The ability to link individual records is important as the decision to purchase private insurance is likely to be a joint decision This is con® rmed by the preponderance of family insurance policies rather than individual insurance policies held by households In the empirical analysis reported below, the income unit is the unit of observation and the head of the household is the representative decision maker The con® guration of the dependent variable recognizes that insurance purchase is a joint decision by classifying an income unit as having insurance if either the head or the spouse indicated that they have private health insurance.1 The ® rst step in de® ning the sample was to select heads of households or their spouses.1 Several restrictions were imposed upon this sample The ® rst and most important restriction was to delete those households in which either the head or the spouse have a health care card This restriction led to the deletion of 7306 households Other deletions from the sample were for cases where the head of the household was aged 18 or less, the head of household was still at school and if any information relating to one of the key variables was missing S Hopkins and M P Kidd The empirical analysis focuses on the private health insurance decision We were unable, however, to distinguish between the types of private health insurance.1 It is possible that the determinants of the decision to purchase private health insurance varies across type of policy In the public release sample of the National Health Survey, however, approximately 50% of records are uncoded with respect to the type of insurance policy purchased Furthermore, the type of policy is generally considered to be a secondary rather than a primary choice IV EC ONOMETRIC ESTIMATES Estimates of the model of the probability of the purchase of private health insurance are presented in Table The model is estimated using a maximum likelihood logit estimator The speci® cation is based on the assumption that an individual’s or household’s decision to purchase insurance is determined by the expected utility gain In the analysis reported in Table 1, the variables attempt to capture the three major determinants of the insurance purchase decision of health status, material wellbeing and geographical location The table contains parameter estimates for two separate speci® cations of the econometric model The ® rst speci® cation is similar to that of Ngui et al in that the focus is on the characteristics of the head of the income unit only The second speci® cation, in which the ® rst is nested, includes variables which capture spousal income, and information on spousal smoking, hospitalization and doctor visits These results are reported in the last two columns of Table The pattern of coe cient signs and the signi® cance of the variables does not vary signi® cantly across the two speci® cations A likelihood ratio test of the null hypothesis underlying the nested speci® cation in column one gives a chi-squared statistic of 380 Thus, the null hypothesis that the coe cients on spousal characteristics are jointly zero is rejected This indicates that the second model speci® cation which includes the role of the spousal variables is the preferred model Details of the variables and the de® ned default group in the model are reported in the Appendix The usual partial derivative interpretation is not appropriate in the discussion of the coe cient estimates In the binary model, however, the marginal e€ ect of a change in This statistic of 51.7% contrasts with the statistic of 43% reported by Willcox (1991) which represents the percentage of the Australian population who have private health insurance 10 The data set follows the standard Australian Bureau of Statistics practice of treating the male as the head of the household 11 The proportion of income units in which either the husband or the wife has individual private cover is very small Thus, it is unlikely that the econometric results are sensitive to the current de® nition of the dependent variable Some preliminary speci® cation checks con® rmed this result 12 There were 25 619 heads and 12 726 spouses of whom 47% and 60.5%, respectively, had private health insurance Since spousal information was merged with head of household information, the initial sample size was 25 619 13 The two main types of private health insurance Ð basic or supplementary Ð are discussed in footnote 1627 Demand for private insurance under Medicare Table L ogit estimates for demand for private health insurance Variable Constant age (25Ð 34) age (35Ð 44) age (45Ð 54) age (55Ð 69) age (70 + ) smoker drvisit1 drvisit2 drvisit3 drvisit4 hosp smokers drvisit1s drvisti2s drvisit3s drvisit4s hosps gender marital status depkid degree trade diploma other income incomes nsw vicmet vic qldmet qld samet sa wamet wa tasmet tas nt act Coe cient estimate Asymptotic standard error 0.757** 0.075 0.052 0.059 0.063 0.071 0.307 0.036 0.059 0.049 0.055 0.057 0.061 Ð Ð Ð Ð Ð Ð 0.051 0.051 0.050 0.059 0.046 0.049 0.181 1.44E Ð 0.069 0.052 0.084 0.077 0.075 0.080 0.128 0.073 0.124 0.123 0.095 0.193 0.123 - 0.063 - - - - 0.109* 0.312** 0.700** 1.867** 0.578** 0.349** 0.336** 0.323** 0.251** 0.361** Ð Ð Ð Ð Ð Ð 0.516** 0.969** 0.108* 0.280** 0.024 0.277** 0.067 2.7E - 5** Ð 0.143* 0.099* 0.218** 0.708** 0.547** 0.736** 0.457** 0.159** 0.456** 0.775** 0.203* 0.264 0.303** Coe cient estimate - 0.773** - 0.083 - - - - 0.110* 0.333** 0.751** 1.870** 0.512** 0.327** 0.323** 0.309** 0.248** 0.365** 0.466** 0.383** 0.332** 0.233** 0.320** 0.410** 0.520** 0.500** 0.074 0.244** 0.023 0.216** 0.087 2.7E - 5** 1.6E - 5** 0.191** 0.079 0.247** 0.691** 0.535** 0.752** 0.486** 0.180** 0.500** 0.807** 0.237** 0.280 0.314** Asymptotic standard error 0.075 0.053 0.060 0.063 0.072 0.307 0.038 0.059 0.049 0.055 0.057 0.062 0.057 0.086 0.079 0.088 0.094 0.067 0.051 0.089 0.052 0.060 0.046 0.049 0.182 1.46E 2.26E 0.070 0.052 0.085 0.078 0.076 0.080 0.128 0.073 0.124 0.124 0.095 0.194 0.124 6 **indicates p < 0.001 *indicates p < 0.005 Summary statistics Number of observations = Prediction L (0) L (b ) - 2[L (0) - L (b )] = - 2[L (b a) - L (b b )] = 16 472 72.7% 22 383 19 682 5402 (50.9) 380 (18.5) 73.5% - 22 383 - 19 492 5782 (50.9) where L (0) and L (b ) are the log likelihood value for a model with an intercept only and the intercept and all covariates respectively L (b a) and L (b b) are the log likelihood values for the ® rst set of regressors which has some restrictions and the second set of regressors, respectively 1628 a speci® c variable is simply a positive constant times the relevant coe cient Thus, the sign and relative magnitude of the coe cient are informative.1 The determinants of the insurance decision are grouped into three as outlined above The ® rst group capture the in¯ uence of expected medical need The inclusion of the age variables is based on the hypothesis that medical need increases with age Van de Ven and van Praag (1981) note, however, that age is both an indicator of perceived medical need and the stock of wealth Young individuals or families are generally relatively less well-o€ but healthier So too are young individuals or couples The probability of insurance associated with the four age variables beyond the age of 35 con® rms the hypothesis that older people are less healthy and therefore, more likely to purchase private health insurance than younger people The doctor consultation and hospitalization variables are proxies for expected medical consumption The four doctor visit variables reported represent di€ erent time periods since the last doctor consultation, ranging in time from less than two weeks to 12 months The results in Table show that the more recent the last doctor visit, the higher the probability of purchase of private health insurance Similarly, if the individual had been admitted to hospital in the last 12 months, the likelihood of the individual being insured increases signi® cantly.1 Importantly, since the coverage of private health insurance under the present regulation does not include doctor consultations outside of hospitals, the relationship between doctor visits and private health insurance status is not subject to moral hazard It is reasonable, therefore, to treat doctor visits as exogenously determined The inclusion of a smoking variable may be seen as a proxy for expected health consumption Alternatively, it may be viewed as a proxy, in the absence of better indicators, of risk aversion in the insurance decision The results indicate that the probability of the purchase of private insurance is lower for smokers, ceteris paribus Our result is similar to that of Propper (1987) who interprets the negative sign on the coe cient as evidence that less risk averse individuals are less likely to purchase health insurance Gender also plays an important role in the insurance decision through its in¯ uence on expected medical consumption Sindelar (1982), for example notes that most of the higher demand for medical services by women may be explained by increased need during the reproductive years The results presented in Table indicate that, ceteris paribus, the probability of insurance is signi® cantly higher for women Family characteristics have an impact on expected medical need, but also a€ ect the insurance decision by changing 14 S Hopkins and M P Kidd expected utility gain directly For example, the presence of a spouse and/or dependent children may increase the risk aversity of the decision-makers in the family unit Ngui et al (1990) suggest that the composition of the family unit is important in the demand for private health insurance decision due to the impact that illness of one family member has on the utility of other family members The marital status variable con® rms the importance of the family characteristics in the insurance decision The dependent child variable, however, is negative and signi® cant in the ® rst regression only, indicating that dependent children decrease the probability of insurance There are two factors, in addition to the interdependent utility mentioned above, that may be important in this result First, young children are, in most cases, members of young and healthy families and these factors reduce the probability of family insurance Furthermore, Propper (1989) notes that it is probable that public sector treatment for children is viewed as no better or worse than private sector treatment Public hospitals in Australia, for example, have excellent specialized facilities and accommodation for paediatric services The second group of regressors are the material wellbeing variables of education and income Education is likely to have both a direct and indirect e€ ect on the private health insurance decision The direct e€ ect is related to the production function attributes of education in terms of the accumulation of health-related information and the appropriate combination of health inputs This view of the role of education in health decision-making has been well documented by Grossman (1972) and Muurinen (1982) The implication is that not only is a better educated person likely to be healthier which would lower the probability of insurance, but also he/she is likely to be better informed about both the services available in the public hospital system and the bene® ts of joining a private health insurance fund The indirect e€ ect of education is its impact on income Education and income are generally positively correlated (van der Ven and Van Praag, 1981) Higher income generally decreases the opportunity cost associated with the purchase of private health insurance Overall, increases in both income and education would be expected to lead to an increase in the probability of insurance These results are borne out in the results where the probability of insurance increases signi® cantly where an individual has either a university degree and diploma relative to the default of no postsecondary education The probability of insurance rises with both individual and spousal income The income regressor reported in Table is a continuous variable Trials with income in discrete groups revealed that the relationship between Marginal e€ ects are not reported since almost all variables are qualitative (refer to Greene, 1990) It is possible that whether the individual has been hospitalized or not may be a function of the private insurance decision The problems associated with possible endogeneity are ignored here 15 1629 Demand for private insurance under Medicare income and the probability of insurance was monotonic This is in contrast to the view expressed by Feldstein (1973) who notes that higher income tends to make families more willing to assume risk, which reduces the demand for insurance Importantly, our results indicate that the view that high income families not purchase private insurance at, at least, the same rate as lower income families is not supported The ® nal group of variables relate to the local mix of public and private services The inclusion of the geographic location by state or territory and the capital city of each of the states recognizes that the mix of public and private facilities di€ ers between both states/territories and metropolitan and non-metropolitan areas and the di€ erence is re¯ ected in interstate variation in the demand for private insurance The results reported in Table show that individuals or family groups who live in the Australian Capital Territory (ACT), in the Victorian metropolitan area1 and in Queensland metropolitan or non-metropolitan areas are statistically less likely to have private health insurance than those residing in the default region of metropolitan New South Wales This result may re¯ ect interstate variation in the provision of services The ACT has the lowest private bed/population ratio of all states and territories The private sector provides only 11% of total hospital services in the ACT compared to 19% in NSW and 29% in Victoria (National Health Strategy, Issues Paper no 2, 1991) In Queensland, 31.6% of the population has private health insurance which represents the smallest state percentage The Victorian result is more di cult to explain in terms of interstate comparisons as Victoria has both the highest percentage of privately insured at 50.4% and the highest number of private bed-days per 1000 population in Australia The role of geographical location as a determinant of the demand for private health insurance di€ ers between the UK and Australia Propper (1989) ® nds that the demand for private health insurance is greater in the south-west of England and that the regional di€ erences can be largely explained by income, and as more private facilities are located in this region, by travel costs The ® rst of these explanations does not hold in the Australian case The ACT has the highest per capita income in Australia but has the lowest population of private insurance membership The overall results are largely consistent with those of Propper (1989), and Ngui et al (1990) and Cameron and Trivedi (1991) Propper ® nds that the probability of insurance is higher with income, employment and higher socioeconomic groupings but lower with dependent children Ngui et al who used 1983 Australia data ® nd that the probability of insurance is higher with income, employment, children, marital status and health status Cameron and Trivedi report that income, age and gender are important in 16 explaining the private insurance choice decision but other health risk factors of the number of doctor consultations and hospitalization are not Unlike our results, they ® nd that the relationship between income and insurance purchase is concave They report that the number of dependent children has a small positive but frequently insigni® cant e€ ect on insurance purchase V S I MU L A T I O N R E S U L T S : T H E I M PO R T A N C E O F I L L- H E A L T H , I N C O M E AND GEOGRA PHICAL LOCATION IN THE INSURANC E DECISION Tables to present results of simulations of the impact that changes in the important determinants of private health insurance have on the probability of purchase For the purpose of these simulations, a representative married man is de® ned as being in the age group 25Ð 34, with neither children nor post-secondary education, on an average income of $27 508 and living in South Australia His wife is on the average spousal income of $7361 and neither of them smoke, have seen a doctor nor been hospitalized in the last 12 months The simulations presented in Table focus on the issue of adverse selection in private health insurance markets Asymmetry of information is a health care market feature which makes discrimination between risk groups di cult and expensive In Australia, regulation of private health insurance markets institutionalizes adverse selection by prohibiting statistical discrimination between risk groups The results in Table provide evidence of the outcome of that regulatory structure Both age and frequency of hospitalization and doctor visits increase the probability of insurance purchase The probability of purchase for the representative man is 65.1% If the man’s age increase to 70 plus, ceteris paribus, the probability of insurance increases to 92.9% Similarly, if the health status of the representative man alters so that he has seen a doctor in the last two weeks and been hospitalized in the last 12 months, the probability of purchase rises to 78.9% The results provide substantial evidence that the privately insured are sicker and that the composition of individuals in the private health insurance pool is adverse for the insurer The results in Table also con® rm the earlier results that the presence of dependent children lowers rather than raises the probability of insurance This result holds across all three representative individuals The second set of simulations presented in Table considers the impact of a change in income on the probability of insurance The relationship between income and private health insurance has been a key feature in the policy debate on the relationship between the private and public sectors in The result for the Victorian metropolitan area is statistically signi® cant in the ® rst regression only 1630 S Hopkins and M P Kidd Table The impact of ill-health on the probability of insurance Table The impact of a change in income on the probability of insurance Probability of purchase (%) Representative man Representative man with depkid Rep man seen Dr in last weeks and hospitalized in last 12 months Representative man is now aged 70 + Single representative man Single representative man with depkid Single rep man seen Dr in last weeks and hospitalized in last 12 months Single representative man now aged 70 + Single representative woman with all other characteristic as for rep man Single representative woman with depkid Single rep woman seen Dr in last weeks & hospitalized in last 12 months Single representative woman is now 70 + 65.1 63.4 78.9 92.9 46.8 45.0 63.8 86.1 59.7 57.9 74.8 91.3 Mean income for the head of household and spouse is 27 508 and 7361 with a standard deviation of 14 852 and 11 359 respectively Representative man is married, aged 25Ð 34, has no dependent children, lives in SA, no post-secondary education, self and spouse earn average income, not seen Dr or hospitalized for at least 12 months, non-smoker health The issue here is essentially that of the appropriate market segment of the two sectors, in spite of the fact that Medicare is compulsory The Medicare levy is set at a ¯ at rate of taxable income and is therefore proportional This assumed proportionality of the Medicare levy when the income taxation system is progressive provides the justi® cation for encouraging those on high incomes to purchase private insurance and to use the private sector facilities.1 In this way, the capacity constrained public health sector is available for those who cannot a€ ord private insurance and private sector facilities In the results presented in Table 3, the probability of insurance increases with an increase in income for all representative individuals Notably, an increase in spousal income is not as important in the probability of purchase decision Spousal income, however, is considerably lower on 17 Probability of purchase (%) Representative man Representative man’s income increases by s.d Representative man’s spouse’s income increases by s.d Single representative man Single representative man’s income increases by s.d Single representative woman Single representative woman’s income increases by s.d 65.1 73.3 69.3 46.8 56.5 59.7 68.6 Mean income for the head of household and spouse is 27 508 and 7361 with a standard deviation of 14 852 and 11 359 respectively Representative man is married, aged 25Ð 34, has no dependent children, lives in SA, no post-secondary education, self and spouse earn average income, not seen Dr or hospitalized for at least 12 months, non-smoker Table The impact of a change in geographical location on the probability of insurance Geographic location Representative Representative Representative married man single man single woman South Australia S.A metro NSW Vic metro Victoria Qld metro Queensland W.A metro W.A Tas metro Tasmania Nth Territory ACT 65.1 70.9 58.1 51.5 59.3 36.5 40.2 57.9 65.4 72.0 59.2 60.3 45.6 46.8 53.5 39.6 33.3 41.0 21.3 24.1 39.3 47.2 54.8 40.7 41.7 28.4 59.7 65.9 52.4 45.7 53.9 31.3 34.8 52.2 60.0 67.1 53.6 54.7 40.0 average than head of household income The results provide no support for the view that the wealthy are uninsured The ® nal set of simulations presented in Table consider the impact of a change in geographical location on the The Federal Coalition Parties’ Fightback policy package, for example, suggested the use of a tax incentives to encourage those on high incomes to purchase private health insurance 18 Whether the ¯ at rate Medicare levy (presently set at 1.25%) is actually proportional or not is complicated by two important considerations First, the Medicate levy does not fully ® nance health care expenditure The di€ erence between the Medicare levy and the Commonwealth government’s contribution to total health care expenditure is ® nanced from general taxation revenue (McClelland, 1991) The second consideration is the relationship between gross income and taxable income on the grounds of both horizontal and vertical equity Individuals with the same gross income may have di€ erent taxable income due to unequal access to tax deductions which reduce taxable income And individuals with di€ erent levels of gross income may also have unequal access to tax deductions 1631 Demand for private insurance under Medicare probability of purchase Across the three representative individuals, the Hobart (Tasmanian metropolitan) probabilities are the highest, followed very closely by the Adelaide (South Australian metropolitan) probabilities, and the Queensland probabilities of purchase are the lowest The signi® cance of the results in Table is that for a given representative individual, the di€ erent probabilities of purchase are a result of the di€ erent location only, not the di€ erent socio-economic composition of population in that location This immediately raises the issue of what factors explain such a wide discrepancy in propensities to privately insure between States For example, a representative married man has a probability of purchase of 72% if he lives in Hobart, but of 36.5% if he lives in Brisbane There are three possible explanations: interstate di€ erences in the price of private health insurance policies, interstate di€ erences in the copayments in the private sector and ® nally, interstate differences in access to the public sector which, ceteris paribus, should be re¯ ected in public hospital waiting lists The costs of insurance policies in respective States does accord with the di€ erent propensities to purchase For example, in October 1990, the average cost of basic hospital insurance was $10.02 per week for a family in Tasmania and $15.30 per week for a family in Queensland These ® gures represent the minimum and maximum values for all Australian states (Willcox, 1991) Copayments for using the private sector also vary between states The ® gures on the average patient copayment for ward accommodation in private hospitals indicate that residents of New South Wales face almost twice the contribution of residents of Victoria, Queensland and South Australia (National Health Strategy, 1991) Figures are not available publicly for the other States A further complicating factor is that a low level of private health insurance does not imply a small private sector The Queensland population has the lowest level of private health insurance coverage of 31.6% in 1990, yet Queensland has the second highest private bed to population ratio (1989/90 ® gures) (National Health Strategy, 1991) Overall interpretation of the results, however becomes confused in the peculiarities of the mix of the role of the private and public sector in Australia Not only are the private bed-days in private hospitals important, but account must also be taken of the private bed-days in public hospitals as well as the overall hospitalization rate for the State In Queensland, for example, there is limited use of public hospitals by private patients, but Queensland has a large private sector, as mentioned earlier The interstate di€ erences in supply of public hospital beds and interstate variations in the demand for those beds should, ceteris paribus be re¯ ected in public hospital waiting lists Waiting list data in Australia are notoriously unreliable and even where they are available, comparisons between States are di cult because of the lack of consistency in the collection and presentation of the data A pattern in the relationship between the propensity to purchase private insurance and the cost of private health insurance in terms of both the premia and copayments and size of the public and private sectors is di cult to discern The negative relationship between the propensity to purchase private insurance and the insurance policy costs is indicative of a downward sloping demand curve An indication of the true relationship between the cost of insurance policies, the supply of health services and the demand for private health insurance, however, requires an analysis of time series data and is beyond the scope of this present study Time series data on the percentages of the population with private health insurance, however, indicate that the Queensland population has traditionally had the lowest national percentage It was 51.1% in 1977 (Hart, 1990) This was followed closely, however, by the Tasmanian percentage of 54.9 Data on premia over the same time span are unavailable The time series data, however, con® rm the common perception of why Queensland has a lower propensity to purchase private health insurance It is due to an historical tendency towards public sector care rather than private sector care For example, before Medicare, the Queensland population had access to free shared-ward accommodation with treatment by hospital doctors V I C O N CL U S I O N Analysis of the determinants of demand for private health care in Australia indicates that three sets of variables are important These are health status, material wellbeing and the relative importance of the private and the public health insurance coverage Our results con® rm that privately insured individuals are on the whole sicker But we refute the view that there is a tendency for high income families not to privately insure Geographical location is an important determinant of the probability of insurance The explanation of why it is so important can in part be explained by reference to the interstate di€ erences in the relative supply of private versus public facilities and in the price of private insurance The complete explanation, however, relates to both the supply and price factors, and to interstate di€ erences in the historical tendencies to demand private insurance A C K N O WL ED G EM E N T S This research was assisted by ® nancial support from the Curtin University Research Grants Scheme We thank Dr Darrel Doessel, and Dr Thorsten Stromback and other workshop participants from the School of Economics and Finance, Curtin University 1632 RE F ER E N C E S Access Economics (1991) Federal Opposition Coalition Fightback Package: Health Arrow, K J (1963) Uncertainty and the welfare economics of medical care, American Economic Review, 53, 941Ð 73 Australian Bureau of Statistics (1990) Health Insurance Survey, Catalogue no 4335.0, June Ben-Akiva, M and Lerman, S T (1985) Discrete Choice Analysis: Theory and Application to Travel Demand (Cambridge MA, MIT Press) Cameron, A C and Trivedi, P K (1991) The role of income and health risk in the choice of health insurance: evidence from Australia, Journal of Public Economics, 45, 1Ð 28 Doessel, D P (1992) An economic analysis of hospital insurance under Medicare In Selby Smith, C (ed.) Economics and Health: 1991 Proceedings of the Thirteenth Australian Conference of Health Economists, Public Sector Management Institute, Monash University, Melbourne Feldstein, M S (1973) The welfare loss of excess health insurance, Journal of Political Economy, 81, 251Ð 80 Greene, W H (1990) Econometric Analysis (New York, Macmillan) Grossman, M (1972) On the concept of health capital and the demand for health, Journal of Political Economy, 80, 223Ð 55 Hart, R F G (1990) In-patient waiting lists: the need for adequate data in Australia Paper presented at the 19th Conference of Economists, University of New South Wales, September McClelland, A (1991) Spending on Health: The Distribution of Direct Payments for Health and Medical Services, National Health Strategy, Background Paper no 7, July Muurinen, J M (1982) Demand for health: a generalised Grossman model, Journal of Health Economics, 1, 5Ð 28 National Health Strategy (1991) Hospital Services in Australia: Access and Financing, Issues Paper no 2, August Ngui, M., Burrows, C and Brown, K (1990) Health insurance choice: an econometric analysis of ABS health and health demand insurance surveys In Selby Smith, C (ed.) Economics and Health: 1989 Proceedings of the Eleventh Australian Conference of Health Economists, Public Sector Management Institute, Monash University, Melbourne Propper, C (1987) An econometric estimation of the demand for private health insurance in the UK Centre for Health Economics, University of York, Discussion Paper no 24 Propper, C (1989) An econometric analysis of the demand for private health insurance in England and Wales, Applied Economics, A21, 777Ð 92 Rees, R (1989) Uncertainty, information and insurance In Hey, J Current Issues in Microeconomics (Macmillan, London) Ch Rice, R (1966) An analysis of the hospital as an economic organisation, The Modern Hospital, 106, 87Ð 91 Sindelar, J L (1982) Di€ erential use of medical care by sex, Journal of Political Economy, 90, 1003Ð 19 van de Ven, Wynand, P M M and Bernard, M S van Praag (1981) The demand for deductibles in private insurance: a probit model with sample selection, Journal of Econometrics, 17, 229Ð 52 S Hopkins and M P Kidd Willcox, S (1991), A healthy risk? Use of private insurance National Health Strategy Background Paper no 4, March APPENDIX All variables are taken from the Australian Bureau of Statistics National Health Survey data set A superscript s in Table indicates that the variable refers to the spouse characteristics Age variables : default variable is age 18Ð 24 Smoker: dummy variable = if individual smokes, = otherwise Drvisit: four variables representing period since last doctor consultation Drvisit1 = less than weeks since last doctor consultation, drvisit2 = weeks to less than months ago, drvisit3 = months to less than months ago and drvisit4 = months to less than 12 months ago The default variable for the last doctor consultation was more than 12 months ago Hosp: dummy variable = if the individual has been to hospital in the last 12 months Gender: dummy variable = if the individual is a male Marital status: dummy variable = if the individual is presently married Depkid: dummy variable = if the individual has a dependent child or children Education variable s are represented by a series of dummy variables where no post-secondary education is the default Degree = bachelor degree or higher, trade = trade certi® cate or apprenticeship, diploma = post-secondary certi® cate or diploma, other = other post-secondary quali® cations Income: income is a continuous variable and represents the aggregation of eleven discrete income groups Geographic location: dummy variable = nswmet (Sydney) nsw = the remainder of New South Wales, vicmet = Melbourne, vic = the remainder of Victoria, qldmet = Brisbane, qld = the remainder of Queensland, samet = Adelaide, sa = the remainder of South Australia, wamet = Perth, wa = the remainder of Western Australia, tasmet = Hobart, tas = the remainder of Tasmania, nt = Northern Territory and act = The Australia Capital Territory  ... Australia The ® rst contribution of this paper is an examination in an econometric framework of the relationship between the demand for private health insurance and the determinants of the demand The. .. to drop out of the market as they perceive that the premium is in excess of their probability of loss The third contribution is a consideration of the regional di€ erences in the demand for private... households In the empirical analysis reported below, the income unit is the unit of observation and the head of the household is the representative decision maker The con® guration of the dependent