Espectroscopía infrarroja por transformada de Fourier

This section reports descriptive statistics for the key variables measuring VPHI coverage andthe use of

health care services. Additional descriptive statistics are reported in the empirical chapters where this is relevant. Moreover, the marginal response distributions for the remaining variables included in the dataset can be found in Appendix A.

Table 5.10 shows descriptive statistics for the variables measuring VPHI coverage along the dimensions of VPHI supplied by commercial insurers and membership of ‘denmark’, thus allowing for an assessment

of double coverage. The individuals who do not know their exact insurance status are dropped.48

Table 5.10 Types of voluntary private health insurance schemes held Member of ‘denmark’ VPHI supplied by

commercial insurer _{Yes No Total}

Through own employer

- Employer pays all 9.93% (n = 492) 8.03% (n = 398) 17.95% (n = 890)

- Employee contributes 4.38% (n = 217) 3.03% (n = 150) 7.40% (n = 367) Through partner’s employer 2.02% (n = 100) 1.51% (n = 75) 3.53% (n = 175) Individually purchased 2.28% (n = 113) 1.27% (n = 63) 3.55% (n = 176) No 35.26% (n = 1,748) 32.30% (n = 1,601) 67.56% (n = 3,349) Total 53.86% (n = 2,670) 46.14% (n = 2,287) 100.00% (n = 4,957)

It is seen from Table 5.10 that while 32 percent of the sample do not hold VPHI, the individuals in the remaining part of the sample all hold some type of VPHI coverage. More than half of the respondents are members of ‘denmark’. Among the members of ‘denmark’, a considerable share also holds employment- based VPHI. While the far majority of the individuals with employment-based VPHI are insured through their own employer, some individuals have VPHI through their partner’s employer. The employers are seen to pay the entire premium for the majority of the individuals who are insured through their own employer. However, a notable share contributes to the premium out of the pre-tax income. Finally, it is

48 _{The dropped individuals are distributed as follows: 51 did not know whether they were members of ‘denmark’,}

221 did not know whether they were insured through their own employer, 60 were insured through their employer but did not know whether the premium was fully paid by the employer; 100 did not know whether they were insured through their partner’s employer; 58 did not know whether they had purchased VPHI from a commercial insurance company on an individual basis.

88

seen from Table 5.10 that some of the members of ‘denmark’ have taken out VPHI from a commercial insurance company on an individual basis. While this is perfectly possible, it cannot be ruled out that some of these individuals have confused VPHI with other types of insurance sold by commercial insurers, such as insurance that pays out a fixed amount of money in the event of a critical illness.

Table 5.11 shows descriptive statistics for the variables measuring the use of the types of health care services analysed in the empirical chapters for the full sample and broken down by insurance status. Health care use is measured by self-reported number of visits within the previous 12 months, as discussed in section 5.2.1.4. It is seen from Table 5.11 that the distribution in the use of health care services within the previous 12 months is right-skewed with a high concentration of zeros for all services except for contacts to GPs and dentists and the use of prescription medication, where more than half of the sample reports a positive use. Comparing the health care use of the uninsured to the sample average, it is seen that the percentage with a positive use is lower among the uninsured for contacts to GPs, physiotherapists, chiropractors, psychologists, specialists, dentists, and hospitalisations and higher for ambulatory contacts and regular use of prescription medication. Considering average use, the pattern differs somewhat in that the uninsured have less contacts to physiotherapists, chiropractors, and dentists than the sample average but more contacts to GPs, psychologists, ambulatory providers, and hospitalisations. Hence, the descriptive evidence on differences in use between the individuals with and without VPHI, respectively, does not reveal any clear patterns.

Within the group of privately insured, the average number of contacts to GPs, physiotherapists, specialists, and dentists as well as ambulatory contacts and hospitalisations during the 12 month period is above the average of the full sample for members of ‘denmark’ and below the average for individuals with employment-based VPHI. Except for physiotherapist contacts, this trend is confirmed by considering the distribution of visits, where the percentage of individuals with positive use is above the average of the full sample for members of ‘denmark’, and below the average for individuals with employment-based VPHI. Likewise, the percentage with a regular use of prescription medication is above the sample average for members of ‘denmark’ and below the sample average for individuals with employment-based VPHI. These differences support the strategy outlined in section 1 of analysing membership of ‘denmark’ and employment-based VPHI separately. The descriptive statistics provided in Table 5.11 do not reveal any clear-cut patters regarding how the use of physiotherapy, chiropractic care, and psychological counselling differs between insurance groups. The use for the individuals with combinations of ‘denmark’, employment-based VPHI, and VPHI purchased through a commercial insurer on an individual basis generally lies in the interval between members of ‘denmark’ and employment-based VPHI, although with some deviations. Finally, the group of individuals who are only covered by VPHI purchased on an individual basis through a commercial insurer is rather small and thus not considered further here.

89

Table 5.11 Health care use broken down by insurance status Insured Visits to: ’denmark’ (n = 1,748) Employment- based (n = 623) Commercial (n = 63) Combi- nations (n = 922) Uninsured (n = 1,601) Total (n = 4,957) GPs 0 16.30% 20.39% 14.29% 17.68% 18.11% 17.63% 1 16.99% 23.43% 26.98% 21.15% 18.49% 19.18% 2 or more 66.70% 56.18% 58.73% 61.17% 63.40% 63.18% Mean (std. err.) 3.75 (0.12) 2.87 (0.16) 4.14 (1.45) 2.90 (0.11) 4.15 (0.16) 3.62 (0.08) Physiotherapists 0 81.86% 80.10% 82.54% 77.22% 83.95% 81.46% 1 2.80% 5.46% 4.76% 3.58% 3.25% 3.45% 2 or more 15.33% 14.45% 12.70% 19.20% 12.80% 15.09% Mean (std. err.) 2.13 (0.20) 1.61 (0.23) 2.41 (1.59) 1.94 (0.19) 1.94 (0.22) 1.97 (0.11) Chiropractors 0 88.04% 88.28% 84.13% 83.30% 91.76% 88.34% 1 2.40% 1.77% 4.76% 4.34% 1.31% 2.36% 2 or more 9.55% 9.95% 11.11% 12.36% 6.93% 9.30% Mean (std. err.) 0.60 (0.06) 0.61 (0.09) 0.40 (0.13) 0.84 (0.09) 0.45 (0.05) 0.59 (0.03) Psychologists 0 93.48% 94.86% 93.65% 93.49% 94.25% 93.91% 1 1.03% 0.64% 0.00% 1.30% 0.69% 0.91% 2 or more 5.49% 4.49% 6.35% 5.21% 5.06% 5.18% Mean (std. err.) 0.44 (0.06) 0.37 (0.08) 0.78 (0.50) 0.44 (0.08) 0.45 (0.07) 0.44 (0.04) Specialists 0 62.64% 74.64% 69.84% 66.92% 67.65% 66.65% 1 19.97% 13.64% 19.05% 18.44% 16.43% 17.73% 2 or more 17.39% 11.72% 11.11% 14.64% 15.93% 15.61% Mean (std. err.) 0.85 (0.05) 0.59 (0.07) 0.52 (0.13) 0.67 (0.05) 0.75 (0.05) 0.75 (0.03) Dentists 0 13.10% 18.78% 26.98% 11.82% 26.11% 17.95% 1 25.46% 35.79% 31.75% 29.61% 26.86% 28.06% 2 or more 61.44% 45.43% 41.27% 58.57% 47.03% 53.98% Mean (std. err.) 1.91 (0.04) 1.48 (0.05) 1.33 (0.15) 1.76 (0.04) 1.56 (0.04) 1.71 (0.02) Ambulatory 0 69.57% 77.21% 74.60% 76.03% 69.96% 71.92% 1 13.90% 10.27% 4.76% 11.06% 10.99% 11.86% 2 or more 16.53% 12.52% 20.63% 12.91% 19.05% 16.22% Mean (std. err.) 1.00 (0.08) 0.87 (0.14) 0.73 (0.18) 0.63 (0.06) 1.09 (0.08) 0.94 (0.04) Hospitalisations 0 86.84% 89.25% 88.89% 91.21% 88.69% 88.58% 1 8.92% 7.54% 6.35% 6.72% 7.62% 7.89% 2 or more 4.23% 3.21% 4.76% 2.06% 3.69% 3.53% Mean (std. err.) 0.24 (0.03) 0.16 (0.02) 0.22 (0.10) 0.12 (0.02) 0.21 (0.02) 0.20 (0.01) Medicine use Yes 51.77% 33.23% 38.10% 36.98% 52.28% 46.68%

90

No 48.23% 66.77% 61.90% 63.02% 47.72% 53.32%

6

A reader’s guide to the thesis

The main part of the thesis is made up by one review paper (chapter 2) and three empirical papers with original research (chapters 3-5). Chapter 6 discusses and concludes. Finally, a Danish summary is included at the end of the thesis.

The empirical chapters are all based on data from the cross-sectional sample of the Danish population described in detail in section 5. Given the intention that each chapter can be read independently and in an arbitrary order, there will be some repetition of general issues. Moreover, the chapters are written with an eye to publication in different academic journal. Hence, the style of writing and reference differs somewhat between the chapters. The reader is asked to bear with these inconveniences.

Chapter 2 reviews the empirical literature on what characterises the privately insured in universal health care systems and assesses how well the empirical evidence corresponds with the theoretical predictions. This information is useful in itself, as well as in order to guide the selection of covariates in subsequent empirical chapters. The review is restricted to consider individually purchased policies, given that the theoretical frameworks for analysing individually purchased and employment-based VPHI differ markedly. Empirical studies were identified by performing searches in electronic databases and examining weekly reports on new health economics research. The literature search identified a total of 24 articles and 15 working papers, the majority of which were published within the recent decade. Socioeconomic characteristics, including income, are generally found to be important determinants of having private health insurance. Likewise, the empirical evidence generally supports the theoretical prediction of individuals selecting themselves into duplicate VPHI based on the quality of care available within the universal health care system, just like the demand for VPHI is consistently fund to be negatively affected by the effective insurance premium. On the contrary, the empirical evidence on the importance of risk preferences is sparse and points in different directions. Finally, with few exceptions, the privately insured are generally found to be in better health, thus rejecting the standard theory of adverse selection. The literature provides several possible explanations for the absence of adverse selection.

While the determinants of individually purchased VPHI have been studied extensively in the literature as evident from chapter 2, empirical evidence on what characterises the group of individuals with employment-based VPHI in universal health care systems is restricted to a few studies.

Chapter 3 estimates the determinants of employment-based VPHI ownership within the Danish workforce and explores whether these differ for employees who receive the insurance free of charge and those who pay the premium out of their pre-tax income. It was found that the probability of having employment-based VPHI is positively affected by private sector employment, size of the workplace,

91

whether the workplace has a health scheme, income, being employed as a white-collar worker, and age until the age of 49, while the presence of subordinates, gender, education level, membership of 'denmark' and living in the capital region are not significantly associated with insurance coverage. As expected, the characteristics related to the workplace are by far the quantitatively most important determinants. The association between employment-based VPHI and self-assessed health is found to be quadratic such that individuals in good self-assessed health are more likely to be insured than those in excellent and fair, poor or very poor self-assessed health, respectively. Finally, the probability of having employment-based VPHI is found to be negatively related to the level of satisfaction with the tax-financed health care system. The results are not affected notably by applying a bivariate probit model with sample selection in order to distinguish empirically between employees who receive the insurance free of charge and those who pay the premium out of their pre-tax income. Hence, these two groups may reasonably be combined in future analyses of employment-based VPHI in Denmark, even though the underlying decision processes leading to insurance coverage differ somewhat.

Another key issue in the economic literature on private health insurance is one of identification; more precisely how to separate the causal effect of VPHI on the use of health care services from differences in use that are attributable unobserved factors affecting both the probability of having VPHI and the use of health care services. This issue is the focal point of chapters 4 and 5.

Chapter 4 estimates the effect of employment-based VPHI on the use of covered health care services using the method of propensity score matching. This method is based on an assumption of selection on observables, which is argued to be plausible given the institutional setting of employment-based VPHI in Denmark and the wide set of relevant covariates available in the data. The chapter seeks to comply with the common critique of matching estimators that they require the researcher to make a large number of choices in the estimation process by assessing the sensitivity of the results with respect to several possible specifications of the propensity score and matching algorithms. For the total sample of occupationally active, the estimates of how employment-based VPHI affects the probability of having had one or more hospitalisations, physiotherapist, chiropractor, psychologist, specialist, and ambulatory contacts within the previous 12 months are positive for all health care services except for psychologist visits, but do not differ significantly from zero. Restricting the sample to private sector employees, it is found that employment- based VPHI increases the probability of having had any ambulatory contacts (such as examinations, scans, same-day surgery, and control visits) by 6-7 percentage points in addition to the baseline probability of 22.4 percent.

Chapter 5 investigates how the estimated effects of individually purchased VPHI varies with different untestable assumptions by discussing and comparing the results obtained by four fundamentally different identification strategies: 1) Joint parametric modelling relying on functional form and an instrumental variable, 2) propensity score matching relying on selection on observables, 3) a standard univariate

92

parametric estimator relying on functional form and selection on observables and finally 4) non- parametric bounds using weaker assumptions. The results show evidence of a positive and significant effect of VPHI on the use of dental care, physiotherapy, and chiropractic care, irrespective of the method applied. The effect of VPHI on the use of ambulatory care is insignificant, while the results differ across methods for general practice and prescription drug use. The joint parametric model allowing for selection on unobservables generally produces higher estimates than the identification strategies relying on selection on observables. It is shown by means of bounding that the exclusion restriction does not have much identifying power on its own, which implies that the results from the joint parametric model mainly rely on functional form. Moreover, it is clear from the various bounds that while strong assumptions of selection do not rule out incentive effects, only one set of bounds identify a positive sign of the effect of VPHI for all outcomes.

Chapter 6 summarises and discusses the findings, policy implications, and limitations of the empirical chapters and concludes.