Health insurance in the Netherlands: The effect of mandatory deductibles on the use of health care

(1)

Health insurance in the Netherlands

The effect of mandatory deductibles on the use of health care

June 11th, 2018

Linda Halling

S1599852

Supervisor: Max van Lent

Second reader: Pierre Koning

Master Thesis Public Administration

Faculty of Governance and Global Affairs

Leiden University

(2)

- 1 -

Abstract

This study aims to find an answer to the question ‘’To what extent does an increase in the mandatory deductible in health insurance influence the use of health care in the Netherlands?’’. The prediction is that an increase in the mandatory deductible will lead to a decrease in the use of health care in the Netherlands, whereby the effect would be stronger for low-income people in comparison with high-income people. This study used data from a panel which made it possible to do a regression analysis with fixed-effects. The analysis showed that the mandatory deductible affects the use of health care but not in the direction we expected. With respect to whether or not people make use of health care the effects we found were so small, that we have to ask ourselves whether this actually leads to a difference in practice. The effects on the number of times show that people do not make less use of health care services and even tend to make more use of health care services. The distinctions between low-income and high-income showed that high-income people are more likely to react stronger and will not make less, but rather more use of health care services. The regression with health as a dependent variable showed that health condition gets worse when the deductible increases. Thus, the answer to the research question is that an increase in the mandatory deductible leads to an increase in the use of health care services which is not in line with the expectations and requires further research.

(3)

- 2 -

1

Introduction

‘’A fundamental principle … in health financing is the concept of universal coverage. This requires access for all people to appropriate promotive, preventive, curative and

rehabilitative health care at an affordable cost. ‘’

~ World Health organization, 2005

In modern society, it goes without saying that everyone who needs medical care should also be able to receive it. According to the quote above from the World Health Organization, this is called the concept of universal coverage and means that all people should have equal access to health care at affordable costs. Although the objective of universal coverage is widely accepted, the realization is dependent on the design of a system where financial contributions are collected from different sources and pooled, so that the risk for health care costs is shared among all people rather than borne by one individual and where the contribution can be used to provide effective health care. Around the world there are many different systems with different arrangements that can be observed (World Health Organization, 2005).

Risk sharing is especially important to keep health care affordable for everyone as the costs for health care are rising. The increase in costs for health care can be explained by changes in health care policy, the aging population, a productivity gap and a higher welfare. Policy changes can be influential as it determines the organization of health care and therefore the costs that come along with it. A policy might be focused on short waitlists which brings higher costs or might accept longer waitlists with lower costs for example. Secondly, the aging population mainly leads to an increase in health care costs, as there are more older people and older people use more health care in general than younger people. The productivity gap is especially relevant in health care because people cannot be replaced by computers and other technological innovations. This means that the individual productivity cannot rise with the same trend as the wages and leads therefore to higher costs. Lastly, an increase in welfare affects the costs of health care as people have higher standards for health

(5)

- 4 -

care and more money available for new methods. Innovation will lead to more expensive treatments for more diseases (CBS, 2015: 4-6).

Looking specifically at the Netherlands, there are indeed rising costs, but it is striking that since 2010 they have risen less strongly than before. According to Statistics Netherlands (from now on: CBS) this can be explained by different factors. The government made health insurers pay more for the risk of transcending costs for example. Also, while some new medicines came on the market for high prices, other medicines became cheaper because of expiring patents. In addition, there were negotiations between producers of medicines and government that led to substantial cutbacks in prices. Lastly, CBS described that the increase in the mandatory deductible is a reason for a reduction in the demand for care. In 2006, the Netherlands made a big change in their system of health care and insurance. They transited from a shredded system with different health insurances to a system with a universal insurance for everyone. In 2008 a new component was introduced into this new system: the mandatory deductible. This meant that everyone must pay for the first part when using health care (some services, for example visiting the family physician, are excluded) (Hamilton, 2009). The mandatory deductible was introduced with an amount of 150 euro per insured per year and rose gradually to 385 euros per insured per year in 2017. This increase in the mandatory deductible is said to reduce the demand for health care according to the CBS, but they do not clearly show how they came to that conclusion (CBS, 2015: 8 & Hamilton, 2009). Nevertheless, this is an interesting claim that brings us to the research question in the next paragraph.

1.1 Research question

This master thesis deals with the last argument of the CBS and focusses on a (possible) reduction of the actual use of health care in the Netherlands as an effect of the increase of the mandatory deductible. The research question is as follows: To what extent does an increase in the mandatory deductible in health insurance influence the use of health care in the Netherlands?. The increase in the mandatory deductible in health insurance is the independent variable and the use of health care in the Netherlands is the dependent variable and this study aims to find a causal relationship between those two variables.

Based upon the insurance theory and some existing literature about the demand of health care we expect to see a decrease in the use of health care when the amount of the mandatory deductible rises (which is also as in line with the findings of the CBS). This study makes a distinction between low-income and high-income people. One reason for this is that

(6)

- 5 -

the general use of care varies between low and high-income people. CBS claims that people with lower incomes in general make more use of the basic health care in comparison with people with higher incomes (CBS, 2015: 15). This is in line with our dataset. We find that the number of times people make use of health care is higher for low-income people than for high-income people (see descriptive statistics in chapter 5 for evidence). The difference in health care costs for low-income and high-income people can (partly) be explained by lifestyle. The percentage of people that are physically active and eat enough fruit and vegetables is higher among high-income people than for people with a low income. Also, the percentage of smokers and overweight people is lower for high-income people (CBS, 2015: 15). Our dataset shows indeed that a higher share of low-income people smokes in comparison with high-income people (appendix A). Thus, income seems like an important moderator for the use of health care and might therefore also be an important moderator for the respond to an increase in mandatory deductible. The theoretical framework will elaborate this further. Different authors made a distinction between low-income and high-income people in the context of respond to price changes for health care services, which makes it possible to compare the findings of this thesis with the findings in literature.

This study has a large-N quantitative design, which is a comparative method that uses information from many cases to test a causal hypothesis (Toshkov, 2016: 236-237). This approach is in line with the goal of this research as we are testing a causal relationship between the increase of the mandatory deductible and the use of health care in the Netherlands. By using the LISS-panel data this research has a panel design, whereby ‘‘several units are observed at a number of points in time’’ (Toshkov, 2016: 232). This panel data makes it possible to observe the behaviour of people over time. The use of health care by individuals will be observed over time in order to look for a causal relationship.

1.2 Justification

The quote at the beginning of this chapter illustrates that health care should be at affordable costs if we want universal coverage. The rising costs of health care are a threat to the affordability and different interventions are done by governments to keep the increase in costs at hand. The Dutch government encourage the insured to make more balanced use of health care services by introducing a mandatory deductible in 2008. If the insured do not make too quick use of health care services, the cost increase in health care decreases and there are ultimately lower average costs for the insured. Furthermore, personal payments make an appeal on the own responsibility of insured parties (Kamerstukken II, 2006/07, 31094, nr. 3).

(7)

- 6 -

The introduction did not lead to much discussion, however the annual increase led to more discussion. The mandatory deductible is especially in times of elections one of the topics that keeps coming back. Main discussion point is whether people that need health care will not make use of health care because of the amount of the mandatory deductible (Volkskrant, 2017, HP De Tijd, 2017, Trouw, 2016 & Kamerstukken II, 2011/12, 32788, nr. B). This thesis tries to look whether people react to the increase in mandatory deductible and can therefore add something to this social discussion.

This master thesis is also of value for scientific community. Many researches are conducted about the demand of health care (Manning et al., 1987; Keeler & Rolph,1998; Duarte, 2012; Aron-Dine et al., 2013 & Dunn, 2016, and this is only a small selection of the total studies). The studies use different methods and focus on different things but the commonality between the studies is that they compare different insurance plans to estimate price elasticities. This master thesis does not use data with different insurance plans but looks at only one part of an insurance plan: the mandatory deductible. The mandatory deductible is also interesting as people cannot choose the amount by themselves which also distinguishes this master thesis from the existing literature. Therefore, this master thesis can add something to the existing literature as it gives insights about a specific part of health insurance and its effects on the use of health care services. Some of the previously mentioned authors made a distinction between low-income and high-income people. This master thesis also made this distinction and it’s interesting to look if the outcomes are comparable to those from existing literature as those investigated health insurances where people had freedom of choice that does not apply to the mandatory deductible we investigate.

1.3 Structure of the thesis

The design of the Dutch health care system is important in this study. Hence, a brief history and summary about health care insurance in the Netherlands will be given in the second chapter. This chapter makes clear what the main reasons for the new system were and how this is elaborated by highlighting the main components of the health care system. The third chapter describes why people take out insurance and summarizes the moral hazard problem in health insurance. It also gives an overview of relevant existing literature about the demand for health care with focus on the ground-breaking RAND-study. This chapter ends with describing the hypotheses that will be tested in this thesis. The research design will be discussed in the fourth chapter with the methods, data and operationalization of concepts. The fifth chapter consists of two parts. The first part gives more information about the data by

(8)

- 7 -

showing descriptive statistics. The second part discusses the outcomes of the regressions. The final chapter summarizes the study and its findings and gives recommendations for policymakers. The chapter closes with some notes on the study and suggestions for future research. Two appendixes are added at the end.

(9)

- 8 -

2

Health Insurance in the Netherlands

The system for health insurance in the Netherlands changed drastically about a decade ago. The implementation of the ‘Zorgverzekeringswet’ (from now on: ZVW) in 2006 was a new starting point with a universal health insurance for everyone as main component. The first paragraph gives a summary of the old system and how it moved towards the new system. The second paragraph gives an overview of the main goals and motives for change. The new system is described in the third paragraph and the fourth paragraph is a comprehensive elaboration of the mandatory and voluntary deductible.

2.1 From the old system towards the new system

The old system of health insurance was divided into different kinds of insurances. There was a mandatory health insurance for a big share of the Dutch population that was called ‘ziekenfonds’. This mandatory health insurance was a social institution targeted at the people with a low income and ensured that those low-income people had access to health care for a relatively low premium. Doctors and other health care institutions accepted those low prices, under the condition that people with higher income, who were able to pay the private prices, could not be insured by the ‘ziekenfonds’ (Hamilton, 2009: 11, 15). This social health insurance valued solidarity high and therefore no medical selection did take place: premiums were equal for everyone. The insurance was a policy with contracted care (instead of reimbursement of costs) (Hamilton, 2009: 16).

In the sixties this changed slightly. The insurance became an obligation case for people that worked for a wage less than the wage limit and the premium became income-dependent. People that lived below the welfare limit and were not obliged could also opt for a voluntary insurance at this social health insurance. Due to the increasing costs of health care, the wealthier people also wanted to insure themselves. As they were not allowed to insure themselves at the social health insurance, private health insurance companies came up. These private insurers used reimbursement of costs (instead of a policy with contracted care). In addition to these two types of health insurance, there was also a separate insurance for civil servants (Hamilton, 2009: 16-17).

(10)

- 9 -

For a long time, this old system continued to be as it was. Even though it was not imposed by a law, the private insures also had a social acceptation and premium policy. There was no risk selection and premium differentiation between young and old for example. In the seventies and eighties this changed. Private insurance became more attractive for younger people, because they offered them insurances with lower premiums than the premiums at the social health insurance. At the same time, private insurance became more expensive for older people. This led to problems for social health insurance, as the good risks (the younger people) switched to the private insurance and the bad risks (the older people) switched from private to social health insurance. The government intervened by stopping the voluntary social health insurance, which made a bigger share of the population designated on private insurance. To make sure that everything would go well, the government introduced a law that obliged private insurers to offer a standard insurance that must comply some legal requirements. Private insurers were obliged to accept certain groups and the premium for the basic insurance was set by the government (Hamilton, 2009: 15 – 19).

From that moment, the resistance to the central steering role of the government grew. According to the insurers, health care had to be modernized and there had to be more room for the negotiating parties. This should go together with the establishment of one single insurance system whereby broad and equal access to good quality care would be the main goal. Already in the years 1987 to 1993 there was a plan for a big system change. The plan was that health insurance should apply to all Dutch residents and had to include characteristics of both the social health insurance and the private health insurance. This new health insurance policy should have been executed by a new type of insurer. Unfortunately, these plans did not reach the finish line at that time. However, this plan provided new insights for health insurance and the eventual change in 2006 is not very different from what was described in this plan (Hamilton, 2009: 19-25).

2.2 Main goals and motives for change

The previous paragraph mentioned some criticism that have been important for the realization of the new health care system and brought certain goals with it. The new health care system had to end the fragmented situation and make health care accessible for everyone. Introducing one universal health insurance for everyone could do this and making insurance compulsory would make good care accessible for everyone. The criticism on the central steering role was also considered by designing a new system. The new system must bring a shift from central to a more decentralized level where government remains responsible for the accessibility,

(11)

- 10 -

affordability and quality of healthcare but gives insurers and health care providers more freedom and responsibilities in the implementation. Another important goal is the affordability of care. The new system should lead to more regulated competition. More competition is an incentive for efficiency and therefore the new system would perform better than the old system. In addition, the new system should give individuals both more financial responsibility and more freedom of choice when choosing health insurance. This implies that insurers will try harder to align their insurance with the needs of citizens and to achieve a good price / quality ratio. The three main goals of the new system can be summarized as good accessibility, less centralized control and more efficiency (Kamerstukken II, 2003/04, 29763, nr 3: 2).

In addition to these main goals, there were some critics of the old system that pushed for a shift to the new system. First, the old system was not transparent. The segmentation of the health insurance market made it impossible to take advantage of the benefits of a homogenous market. The segmentation also leaded to divergent premiums for people with similar income situations. Another criticism of the old system was that there was little room to switch between insurers as some people were obliged to be insured at the social health care for example. Because there were two different types of insurers (social health insurer and private health insurer), with both a different orientation, there was no level playing field. The social health insurers had an acceptation obligation and were focused on contracting health care for their clients, while the private insurers were focused on risk selection. Those differences and other market imperfections hampered the mobility of insured people. As a result, there was little incentive for insurers to get the maximum out of negotiations at the expense of cost control in the healthcare sector (Hamilton, 2009: 28).

2.3 The new health care system

The new health care system became effective on January 1st, 2006 when the ZVW was implemented. According to Dutch government, the new health care system is more transparent and creates a level playing-field which is important for a sustainable and affordable health care system in the future. The new system obliged all Dutch residents older than 18 years to take out a health insurance (children are insured by means of their parents). Everyone needs to take an insurance for (at least) a legally defined health care package by one

(12)

- 11 -

of the health insurers.1 The obliged health insurance is a standard package of essential care that is legally defined by the government. It contains necessary and ‘focused-on-healing’ care that has been tested against the requirements of demonstrable effect, cost-effectiveness and necessity of collective financing. This package is in line with the package of the social health care from the past. The package will be periodically evaluated and if necessary be adapted to new circumstances or insights to keep the package affordable over time (Hamilton, 2009: 33 & Rijksoverheid, n.d. A).

The insurance is executed by (private) health insurers who must meet the conditions of the ZVW. The private health insurer from the past did not have to change much, while the social health insurers from the past could transform themselves into private health insurers with the possibility to make profit. Health insurers need to operate nationwide. For some small insurers there can be an exception and they may choose a smaller area to operate their services. This should make the entry barrier for new insurers smaller and it gives insurers the possibility to execute growth strategies when contracting health care providers, which is both good for the competition between insurers. The competition is most evident in the purchase of health care. Insurers need to make good deals with the health care providers to maximize the quality/price ratio for their customers (Hamilton, 2009: 33-34). They also can compete by offering different packages. Next to the standard package, they may offer an additional insurance (for the dentist for example) and they can offer packages with different amounts of deductibles.2 A health insurer can also decide whether to offer insurance in the form of contracted care (‘natura-polis’) or reimbursement of costs (‘restitutie-polis’). Because people have freedom of choice, they can choose every insurer they want and have the possibility to switch from insurer at the end of each year. The insurance companies have an acceptation obligation, which means that they cannot refuse anyone (in his working area) who wants a health insurance. In addition, the insurer has a duty of care: an insured has right on care from a contracted institution (in the case of a ‘natura-polis’) or right on reimbursement of costs (in the case of a ‘restitutie-polis’) (Hamilton, 2009: 34-35).

The total premium burden is for a large part borne by the individuals by means of a nominal premium. Every insurer can set the nominal premium for each package that it offers, but only under the condition that the nominal premiums do not differ by age, health conditions or other social circumstances. However, an insurer may offer discounts for

1 _{Two groups are not obliged: soldiers in actual service and conscientious objectors (people with}

principal objection against insurance).

2

(13)

- 12 -

collectivities. Because the insurers have freedom to determine the level of the nominal premium, this gives an incentive for competition. A nominal premium is also good for the cost consciousness of the insured. Next to the nominal premium, there is a wage-related contribution. Employers contribute to this by making an obligatory payment to their employees of the income-related contribution they have paid. The revenue of these ends up in the health insurance fund. The government itself also pays a contribution to this fund. People younger than 18 years do not have to pay a nominal premium and the government pays for this group. Since 2008 people older than 18 years face a mandatory deductible next to the nominal premium they are obliged to pay and therefore contribute to the system.3 The mandatory deductible is discussed in detail in the next paragraph and will therefore not be discussed here.

In order to keep health care financial accessible for everyone, there is a health care allowance. The allowance is available for people where the premium burden is higher than acceptable and is thus based upon income. To encourage people to assess the various insurers on price (nominal premium), the average nominal premium instead of the actual nominal premium (for that person) is used as a yardstick for determining the level of the health care allowance. To make sure that insurers with more bad risks than others will not fall, there is a risk equalization. This mechanism compensates insurers with, on average, many bad risks. Without this risk equalization it would be impossible to obligate insurer to accept everyone (Hamilton, 2009: 35-37).

2.4 Mandatory and voluntary deductible

The introduction of the new health system in 2006 did not include a mandatory deductible. The system had a no-claim regulation to promote individuals’ own responsibility. There has been much criticism on this no-claim regulation, because people with bad health must pay for the system but will never get the benefit of the no-claim scheme due to their health condition. This criticism made the government decide to abolish the no-claim regulation and introduce the mandatory deductible in 2008 (Hamilton, 2009: 123-124).

The mandatory deductible implies that everyone is responsible to pay the first ‘x’ amount when they use health care. The amount of the mandatory deductible is determined by the government. With the introduction in 2008 the amount was set on 150 euros per year. What this means can best be explained by using an example. When you would break your leg

3_{The first two years (2006 and 2007) this was not the case. In those two years there was a no-claim regulation,}

(14)

- 13 -

and the treatment costs 1000 euros, you are obliged to pay 150 euros and the rest will be paid by the insurance company. If you need to use health care later the same year you do not have to pay as you already paid your whole mandatory deductible. There might be an exception when you would still have to pay. This would be in the case when a treatment requires a contribution. The contribution can be a fixed amount or a percentage of the costs of a treatment. Another example will make clear how this works in combination with the deductible. When a treatment is 100 euros and the contribution for this is 30 euros, you must pay 30 euros contribution and 70 euros if your mandatory deductible is not used. When you already paid your mandatory deductible, the last 70 euros will be paid by the insurance company (Hamilton, 2009: 124 & Rijksoverheid, z.d. B). Some services are excluded from the mandatory deductible: family physician, care related to pregnancy, childbirth and maternity, specific care in case of some chronical diseases, district nursing and follow-up visits and travel costs for organ donations. Over the years, the amount of the mandatory deductible has risen. The table below gives an overview of the amounts from 2008 till 2017 (Hamilton, 2009: 36 & Rijksoverheid, z.d. C).

Table 1: Overview mandatory deductible in euros per year

Year 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Mandatory Deductible

150 155 169 170 220 350 360 375 385 385

People can opt for a voluntary deductible in addition to the mandatory deductible. This voluntary deductible works in the same way as the voluntary deductible but makes the total deductible higher. To reassure transparency (comparability of the packages), there are different tranches that the insurance companies may offer. Those tranches are: 100, 200, 300, 400 and 500 euros per year. The voluntary deductible adds up to the mandatory deductible. So, in 2008 the highest possible deductible is (150 + 500) 650 euros. Every insurer needs to offer at least one package without a voluntary deductible but is not obliged to offer all the tranches of the voluntary deductible. When people take a voluntary deductible, they get offered a discount on the nominal premium. This discount is set by the insurance company and must be equal for everyone with the same voluntary deductible (Hamilton, 2009: 126-127).

(15)

- 14 -

3

Theoretical Framework

Insurance is a device to share risk but why are people willing to take out insurance and what conditions must be met for insurance to work? The first paragraph makes clear why people take insurances by describing a risk-averse person. The second paragraph describes the conditions that must be met to make insurance work, which is important for understanding which problems may occur in health insurance. The third paragraph describes the major problems in health insurance. It will turn out that the demand for health care is important for whether there will be an over-consumption problem. The fourth paragraph will give an overview of the most important studies about the demand for health care. Main study in this part is the ground-breaking RAND-study that came up with an estimate for the price elasticity. The last paragraph of this chapter lists the hypotheses based upon the framework.

3.1 Insurance theory: demand side

The demand side is important if we want to know why people are willing to take insurance. In general, insurance companies have a positive expected value which means that an individual contributes more than it will benefit in the long run. So, when people know this why do they still want to insure themselves? The answer to this question is that people are risk-averse and will look for certainty to minimize risk. Insurance might be beneficial for such people as it reduces the uncertainty that causes disutility (Barr, 2012: 83).

There are two underlying conditions for risk-averse people that help understand why a risk-averse person would buy insurance. The first underlying condition is that someone needs to face uncertainty about utility outcomes otherwise the person has no risk that he or she needs to fear. The specific uncertainty in health care is that people do not know when or how much health care they will demand (Barr, 2012: 83 & Arrow, 1963: 855). Insurance is a solution to reduce uncertainty because the law of large numbers. The law of large numbers means that there will be less uncertainty for a bigger group which Barr illustrates with an example: ‘‘I do not know whether I will die this year, but the death rate for men aged 50 to 65 is known and stable’’(Barr, 2012: 85). The certainty about the aggregate probability makes it possible for individuals to gain from risk-pooling. How bigger the number of people participating, how smaller the variance around the average income which is equal to the risk

(16)

- 15 -

of the individual. Thus, people can acquire certainty because of pooling all risks together (Barr, 2012: 86).

Second underlying condition is that a risk-averse person has a decreasing marginal utility of income. This is important as the concavity of the utility curve determines the difference between the utility of expected income and the expected utility of income which is decisive for the willingness to pay for insurance. To make clear, people that are risk-neutral will not have a willingness to pay as there is no difference between the utility of expected income and the expected utility of income. Concluding, only for a risk-averse person it is beneficial to buy insurance (even if the expected pay-out is lower than his premiums payment) as ‘’ uncertainty per se causes disutility; hence certainty is a commodity yielding positive marginal utility, for which he will pay a positive price’’ (Barr, 2012:83-84).

3.2 Insurance theory: supply side

For private insurance to work, there are four conditions that must be met. The first condition is that the probability for something to happen must be independent. If the probability is dependent it will be a common shock against which cannot be insured as there are no winners and losers: everyone is a loser and pooling will therefore not lead to less risk for an individual. The second condition is that the probability should be less than one. If the probability is equal to one, you know for sure that it will happen, and insurance would not be beneficial because the premium would exceed the actuarial loss. The third condition is that the probability must be known. If the probability is unknown, it is impossible for the insurer to calculate a premium and offer insurance. The last condition is that there should not be asymmetric information. Asymmetric information may lead to adverse selection and moral hazard as the insurer has less information than the person buying insurance (Barr, 2012: 88-91).

3.3 Problems in health insurance

The major problems in health insurance have to do with this last condition that are known as adverse selection and moral hazard. Adverse selection might occur when an individual knows more than the insurer. According to Barr (2012: 92) ‘’adverse selection causes inefficiencies because there is an incentive for the worst risks to sign up for insurance and for the best risks to self-insure; where the problem is serious, the market may fail entirely’’. Solution for this problem is to make insurance compulsory, so that the good risks will not opt out of a pooling equilibrium (Barr, 2012: 92, 240 & 243). As everyone in the Netherlands is obliged to take

(17)

- 16 -

out a standard package of health insurance and because the insurers are obliged to accept everyone (in combination with risk equalization) there will not be any adverse selection problems (Hamilton, 2009).

The problem of moral hazard, a result of asymmetric information, is more relevant for this thesis. Moral hazard occurs in two ways in health insurance. Patients might influence the probability of requiring medical treatment or may influence the costs of it. The probability of requiring medical treatment increases when people take fewer health precautions when they are insured compared to when they have no insurance and would have to bear the costs themselves. When coverage is more complete, individuals must bear the consequences and costs of their behaviour less and will thus take less health precautions. The third-party-payment problem also feeds the incentive to consume more health care than needed because it influences the costs of health care. The third-party-payment problem exists because the insurance company does not get involved in decisions of doctors. Because the doctor gets paid by a fee, he or she is not constrained by the patients’ ability to pay and when there is complete coverage the patient does not face the costs of treatment either (except for the premium of the insurance he already paid). Thus, in case of complete coverage, both relevant actors do not face private costs which gives them an incentive to consume all health care that gives private benefits (Barr, 2012: 241).

The elasticity of the demand determines whether this leads to over-consumption. Figure 1, on the next page, shows this. The red line shows the demand for health care and the grey line at P* shows the marginal costs of insurance. In the left graphic, there is inelastic demand for health care. The equilibrium without insurance is at P* and Q1. When the individual is insured and P is zero (as neither the patient as the doctor face private costs) the quantity does not change and stays Q1. In the right graphic with elastic demand for health care there is a different outcome. Without insurance the equilibrium is (again) at P* and Q1 but the quantity changes to Q2 when the individual is insured and P is zero. The difference between Q2 and Q1 is the result of over-consumption and leads to higher social health care costs which is undesired as it reduces social welfare (Barr, 2012: 241). Because elasticity is decisive for whether moral hazard occurs and leads to overconsumption it is defined by Arrow (1963: 961) as the moral hazard effect of health insurance. The observation that medical insurance increases the demand for medical care is the underlying argument. The literature review will elaborate this last claim further.

(18)

- 17 -

Figure 1: Moral hazard leads to over-consumption when demand is inelastic

The problem of moral hazard, which can lead to over-consumption, can be reduced by introducing an incentive mechanism that shares the costs between individual and insurer. Such a mechanism let the individual face some of the costs with the objective to give him or her more awareness of the costs for health care. A first option is to higher premiums for frequent claimants. Second option is introducing a deductible that makes the insured pays the first ‘X’ when they use health care (this master thesis looks at this solution). A third option is coinsurance, whereby the insured pays a certain amount or a ‘X’ percentage of any claim. It must be said, that those solutions cannot solve the whole problem as they only make people aware of a small share of the health care costs (Barr, 2012: 93-94 & 242).

3.3 Literature review about the demand for health care

The ground-breaking study from Manning et al. (1987) is published three decades ago but remains one of the most important studies in the field of health insurance and demand for health care. The article of Manning et al. (1987) reports the results of the Rand Health Insurance Experiment (HIE) in the United States of America. The HIE was a randomized experiment that started in 1974 and was initiated by the federal government. The study was, among others, aimed at narrowing uncertainty about how demand responds to insurance-induced changes in price. The HIE enrolled families in six sites between 1974 and 1977. The families were assigned to different insurance plans that differed in the level of cost-sharing: free plan, 25% coinsurance rate plan, 50% coinsurance rate plan, 95% coinsurance rate plan and individual deductible plan. People in the free plan have no out of pocket costs, while people in the three coinsurance plans must pay the corresponding percentage for medical

(19)

- 18 -

services. The individual deductible plan is a plan with a 95% coinsurance rate for outpatient services subject to a limit of $150 per person or $450 per family per year and has free inpatient care (Manning, et al., 1987: 255).4

The sample means showed that the mean for face-to-face visits is highest for the free plan group and decreases for every increase in coinsurance rate (25, 50 and 95%). Same yields for the outpatient expenses, which is highest for the free plan group and decreases for every increase in coinsurance rate. No such trend is visible for the inpatient dollars. The individual deductible plan shows a different pattern and looking at outpatient care the plan looks like a combination of the 50% or 95% coinsurance rate plan while for inpatient care it looks more like the free or 25% coinsurance rate plan. The authors also looked at the probability of any use of health care services and found that the likelihood of any use decreases when the coinsurance rate rises. The individual deductible plan has estimates that lie between the estimates for the 50% and 95% coinsurance rates (Manning et al., 1987: 258-260). Based upon the ANOVA model, the authors came to the general conclusion that ‘’the use of medical services responds to changes in the amount paid out-of-pocket’’ (Manning et al., 1987: 258). The authors divided these results by subgroups and looked among others for differences across income groups. They found that the probability of any use of medical services increases with income and for every income groups decreases the probability of any use of medical services when the coinsurance rate goes up. The individual deductible plan has probabilities that lie between the estimates for the 50% and 95% coinsurance rates. They also looked at the likelihood of one or more admissions across income groups and find for the coinsurance plans that the lowest income group is most likely to have one or more admission (followed by middle-income group and high-income group is least likely to have one or more admissions). The individual deductible shows reverse effects with highest likelihood for the high-income groups. For all income groups yields that the likelihood decreases as the coinsurance rate goes up. The individual deductible plan shows values that lie between the free plan and 25% coinsurance rate. Note that they looked at differences in responses rather than a (partial) effect of income (Manning et al, 1987: 260-261-262). All results together ‘’leave little doubt that demand elasticities for medical care are nonzero and … that the response to cost sharing is nontrivial’’ (Manning et al, 1987: 267). The authors came up with three methods to estimate this price elasticity. Those three methods suggest a price elasticity between -0.1 and -0.2. A price elasticity describes how strongly the consumer demand

4_{Inpatient care is the care when you are admitted to hospital and need to stay overnight. Outpatient care is all the}

(20)

- 19 -

responds to a change in price. A negative price elasticity means that an increase in the independent variable leads to a decrease in the dependent variable. The found price elasticity between -0.1 and -0.2 means therefore that an increase in costs for health care (out of pocket expenses) will lead to a decrease between 0.1 and 0.2 in the use of health care (Manning et al, 1987: 268).

The core findings of this famous RAND-study are more recently re-examined by Aron-Dine et al. (2013). They first show the estimates of the treatment effects. The use of health care overall shows a ‘’consistent pattern of lower spending in higher cost-sharing plans’’(Aron-Dine et al, 2013: 203). They split the estimates of the treatment effects into inpatient and outpatient care. For outpatient care there is a pattern visible that how completer the coverage how more use of outpatient care. For inpatient such a pattern cannot be distinguished but they note that the effect of cost sharing for inpatient spending is consistently small and generally insignificant which suggest that this type of health care may be less price sensitive (Aron-Dine et al., 2013: 203-204). The authors continue with investigating possible threats to validity of interpreting the outcomes as causal estimates and investigate the robustness of the treatment effects but the conclusion that the RAND-data gives evidence that there is a response to cost sharing remains. Must be said that the sensitivity analysis revealed uncertainty about the magnitude of this response (Aron-Dine et al, 2013: 205-211). The authors also looked critical at the suggested price elasticity of -0.2. They showed some relatively simple and transparent ways to transform treatment effects into elasticity estimates and obtain pairwise elasticities that mostly have a range within -0.1 and -0.5. They continue with describing different methods and problems that may occur when one wants to obtain elasticity estimates. They conclude that, even with the development and use of new methods, it remains difficult to transform treatment effects into elasticity estimates (Aron-Dine et al., 2013: 212-220).

Keeler & Rolph (1998) also used data from the RAND-study but focus more on the number of episodes. They argue that people that become sick must decide whether they will use medical care and when the costs of treatment seem higher than the benefit people will decide to postpone the use of health care and will look if home remedies work. If the costs of treatment do not exceed the benefit and the person will decide to use health care, the doctor helps to decide the number of follow-ups (if needed) and thus how much to spend on care (Keeler & Rolph, 1998: 341). ‘’Thus, an episode of treatment contains two decisions of interest: The first, made primarily by the patient, is whether to seek care at all; the second, made jointly with the doctor, is to decide on the level of treatment’’ (Keeler & Rolph, 1998:

(21)

- 20 -

341). This distinction made it possible to test the main hypothesis that patients are more affected by cost-sharing than providers. They use four types of episodes in their analysis: accurate, chronic, well and dentist. For those four types the authors estimate occurrence ratios for outpatient episodes and hospital episodes. It turned out that for every type of episode the occurrence estimate for outpatient care decreases as the coinsurance rate goes up. The estimates for the individual deductible plan have comparable values as the 50% coinsurance rate plan for acute and chronic type of episodes and have comparable values as the 95% coinsurance rate plan for well and dentist type of episodes. The occurrence estimates for hospital episodes do not show such a pattern and the authors describe that this has to do with some assumptions that are made. The authors also looked shortly at the price elasticities and find elasticities around -0.2 for the low coinsurance range (0-25 percent) as well for the upper coinsurance range (50-95 percent). (Keeler & Rolph, 1998: 357-362).

Dunn (2016) takes a different approach to estimating the demand of health care than Manning et al. (1987) and Aron-Dine et al. (2003). He uses an instrumental variable strategy and uses information from MarketScan commercial claims database for 2006 and 2007 to study the demand response to overall utilization which is broken into the number of episodes and the utilization per episode (Dunn, 2016: 75-77). Dunn argues that the negotiated price between insurers and medical providers is a so-called cost shifter instrument: ‘’the negotiated prices are set prior to insurers making offers to consumers. Thus, the average negotiated service price in an MSA will shape the incentives of insurers when offering plans, but the average negotiated prices should not have a direct effect on the insurance selection by consumers’’ (Dunn, 2016: 78). The results of this paper show estimates for the overall utilization response to the out-of-pocket-price that vary between -0.620 and -0.161. The most preferred IV-strategy used two instruments that ‘’do not rely directly on price information from the enrollee’s plan, both of the instruments contribute significantly to explaining the variation in out-of-pocket prices, and they pass basic tests of validity’’ (Dunn, 2016: 82). The elasticity for overall utilization according to this IV-strategy is -0.199. Dunn specified this effect into the effect on weighted number of episodes and the utilization per episode. The effect on the weighted number of episodes with the preferred IV-strategy shows an elasticity of -0.197. Dunn found less significant and smaller elasticities for the utilization per episode as the preferred IV-strategy estimated an elasticity of -0.0503. The overall conclusion is that consumers response on out-of-pocket-prices is negative and that the estimates match de findings in the RAND-study (Dunn, 2016: 82-87).

(22)

- 21 -

The last article that will be discussed is the article of Duarte (2012). The article of Duarte distinguishes itself by estimating price elasticities across health services whereby data from Chile is used. Duarte looked if there is empirical evidence that the respond to prices varies across health care services and individual characteristics. A distinction is made between elective care where he uses home visits, psychologist and physical therapy evaluations and acute care with appendectomy, cholecystectomy (gall bladder removal) and arm cast setting as health care services. For the elective care the elasticity estimates associated with the coinsurance plans vary between -2.08 and -0.32. Home visits and psychologist have relatively high estimates (-1.89 and -2.08) and according to Duarte this can be interpreted as moral hazard as people react to out-of-pocket prices. He elaborated this further by estimating the health care costs when people would have a higher coinsurance rate and concluded that people respond to the generosity of plans. This suggest that increasing the coinsurance rates for elective care could lead to a reduction in health care costs, but further research focused on possible welfare effects is needed. The elasticity estimates for acute care vary between -0.02 and -0.07, which is in line with the expectation that people who face urgent care are less responsive to prices (Duarte, 2012: 832-833). Duarte extended his analysis by looking for differences among income groups and age. He finds that the price elasticity for the number of visits for low-income individuals explains a bigger share of the price elasticity of total expenditure compared to high-income individuals. ‘’This evidence shows that low-income individuals forgo medical care more often than do high-income individuals’’ (Duarte, 2012: 834). The analysis for difference in age groups shows that older consumers may be less responsive to price than younger consumers. This might be the case because they find it harder to compute expected expenditures or they have less information on health alternatives than younger consumers (Duarte, 2012: 843-835).

3.4 Hypotheses

This thesis looks for an answer to the research question: To what extent does an increase in

the mandatory deductible in health insurance influence the use of health care in the Netherlands?. Based upon the theoretical framework, the expectation is that the use of health

care will decrease as the mandatory deductibles rises. Some specific hypotheses are drawn up and will be tested to answer the research question. The first two hypotheses are about whether people make use of health care services.

Manning et al. (1987) described that the probability of any use for outpatient care decreases as the coinsurance rate goes up. This implies that if someone faces more costs the

(23)

- 22 -

likelihood for any use of health care becomes smaller. Therefore, the first hypothesis is: If the

mandatory deductible rises, the choice whether or not people make use of health care will be affected in a negative way.

As Manning et al. (1987) found that the probability of any use of health care services for all different coinsurance rates increases with income. Therefore, the second hypothesis is:

The likelihood for making any use of health care services is less likely to change for high-income people than for low-high-income people when the mandatory deductible increases.

The third and fourth hypotheses are about the number of times that people make use of health care. Based upon the sample means, occurrence estimates and significant negative price elasticities from Manning et al. (1987), Aron-Dine et al. (2013), Keeler & Rolph (1998), Dunn (2016) the expectation is that people will use health care services less often when the mandatory deductible rises. The hypothesis is formulated as: When the mandatory deductible

rises, the number of times that people make use of health care will decrease.

The article from Manning et al. (1987) and Duarte (2012) are relevant for the distinction between low-income and high-income. Manning et al (1987) showed that when there is more cost-sharing the likelihood for one or more admission becomes smaller and the likelihood with the individual deductible plan is higher for high-income people than for low-income people. Duarte found that low-low-income people forgo medical care more often than high-income people. Based upon this, the last hypothesis is: An increase in mandatory

deductible will lead to less often use of health care services that is stronger for low-income people than for high-income people.

(24)

- 23 -

4

Research Design

This study will investigate to what extent the mandatory deductibles affect the use of health care in the Netherlands and aims to find a causal relationship there. This chapter describes what methods are used, which data is used and how concepts are operationalized with this data.

4.1 Method of Analysis

This study is an explanatory research that tries to unfold a causal relationship. We will look for a causal relationship between the amount of the mandatory deductible and the use of health care. Thus, in this study we investigate to what extent the amount of the mandatory deductible affects the use of health care. Use of health care is defined in two ways: whether or not people make use of health care services and the number of times that people make use of health care services. In order to be able to answer the research question even better and more extensively, a distinction is made between low-income people and high-income people and it has been examined whether there has been a change in the health condition of the participants. How this is further operationalized will become clear in the next paragraph.

This research has a large N-design, because data is used from more than 100 cases. Large N-design is a comparative method that uses information from many cases to test a causal hypothesis and is therefore in line with the goal of this study (Toshkov, 2016: 236-237). The used dataset in this study is important for the research design. This study uses panel-data, which makes it possible to observe several units at a number of points in time (Toshkov, 2016: 232).

The method of analysis in this study is a regression with fixed effects. Especially the fixed effects are important as this makes it possible to follow people over time (which is possible because we have panel-data). A linear probability model is used in the regressions because it suits our dependent variables as most of them are not binary.

In the analysis we use individual over years as the primary unit of analysis. To denote an individual, we use i. The specification of the regression is as follows

(25)

- 24 -

where outcome Yi represents the dependent variable, for example whether or not people make use of health care service, and βQi is the effect of the independent explanatory variable which is the amount of the mandatory deductible. The effect of the control variables, net income and health condition, are denoted in the equation as γAi and γBi. The intercept is denoted as α and the standard error is εi.

The regressions are executed for different groups. The first group consists of all participants. The outcomes of the regressions for this group tells us something about the effect in general for the entire Dutch population. The second group is a group that only included people that do not have a voluntary deductible and thus only face the mandatory deductible. When people have voluntary deductible in addition to the mandatory deductible this might influence the actual effect of an increase in the mandatory deductible. Therefore, using this second group can take away some disturbance in the results that is caused by people that have a voluntary deductible in addition to the mandatory deductible. The third group goes a step further for the same reason as the second group and excluded the people that have an additional insurance as well. People with an additional insurance for dental care for example will have to pay less or nothing from their mandatory deductible when visiting the dentist. Thus, this second groups tries to take away the disturbance in the effect that is caused by the voluntary deductible and by additional insurance. For all groups yields that only people that are 18 or older are included in the regressions as people became obliged to take out health insurance and have a mandatory deductible from the age of 18 years.

4.2 Data

As described before, this study uses panel-data. The used panel is the LISS-panel, which stands for Long-term Internet Studies for the Social sciences. The LISS-panel contains exclusively households from the Netherlands and from all walks of life. Approximately 5000 households spread throughout the Netherlands participate in the panel. The households that participate are selected by the Dutch bureau for statistics (CBS) and CentERdata (CentERdata, n.d., A). This implies that households cannot register themselves for participation in the LISS-panel, which makes the panel more reliable. If households could register, this would probably lead to a poorer reflection of Dutch society because households that would like to participate voluntarily in such a panel often have similar characteristics.

Participants of the panel fill in questionnaires via the internet about relevant topics for scientific, social and policy research. The LISS-panel has different core-studies: health, religion and ethnicity, social integration and leisure, family and household, working and

(26)

- 25 -

schooling, personality, politics and values and economic situation (distinguished in assets, income and housing). This study only used the health core-study and the background information of the participants. The health core-study has ten different waves. The questionnaires from the first wave were administered in 2007, while those from the last wave were administered in 2017. Unfortunately, there is no data available for the year 2014 because no questionnaires were filled in that year (CentERdata, n.d., B). The questions in the health core-study are related to the health condition of people, peoples’ lifestyle, the use of health care and health insurance. One of the most important questions is ‘how often did you use the following health care service over the past twelve months?’ with different categories as family physician, physiotherapist and medical specialist for example. Other important questions are ‘how would you describe your health, generally speaking?’ and ‘In [current year*] you have an obliged own risk of 220 euro. Besides a voluntary own risk is possible. How much is your voluntary own risk in [current year*]?’ (CentERdata, n.d., C). The complete list of variables that were in the composed dataset is added as appendix B.

4.2 Operationalization

As already described before, the dependent variable is use of health care. Use of health care is defined in two ways that will be discussed separately.

The first way to define use of health care is by looking at whether or not people make any use of health care services per year (so: yes or no). This is operationalized in two ways that are similar to each other. The first variable is called visits2. In the questionnaire the participants have answered questions about how many times they used twelve different health care services for the past twelve months. Those twelve health care services are merged into one variable, which is the visits variable. For every value bigger than 0 (for this visits variable) the visits2 variable takes the value of 1 and value 0 of visits is equal to value 0 for the visits2 variable. Thus, visits2 is a binary variable where 0 stands for did not make use of any health care service and 1 stands for did make use of at least one health care service. The second operationalization is almost the same. As family physician does not fall in the mandatory deductible, we excluded this health service as people will not have to pay for using its service. The other eleven health care services were merged into the visits4 variable, where 0 stands for did not make use of any health care service and 1 stands for did make use of at least one health care service excluding family physician.

The second definition of use of health care is the number of times that people make use of health care services. This definition has also been operationalized in two ways. The

(27)

- 26 -

visits variable that was used to derive the visits2 variable from is the first way of operationalization. The visits variable merged the twelve health care services and is a numeric variable, which means that every value between 0 and ∞ (only whole numbers) could be filled in. The second operationalization excludes family physician for the same reasons as described by the earlier definition of whether or not people make use of health care services. The visits3 variable thus merged eleven health care services and is also a numeric variable.

The health condition of the participants is also used as a dependent variable in the regression analysis. The health condition is operationalized with the CH004 variable. The question that was asked is ‘how would you describe your health, generally speaking?’. This variable has values between 1 and 5 as there are five categories: poor, moderate, good, very good and excellence.

The independent variable, that is used in all the regressions with the above described dependent variables, is the amount of the mandatory deductible. The amount of the mandatory deductible is determined by the government. An overview of the amounts is given in table 1 in chapter 2 of this thesis. The variable verplrisico shows the amount of the mandatory deductible for the corresponding year.

The used control variables are net income and health condition. Net income is operationalized as net income per month. The used variable is nettoink and is a numeric variable. Health condition as control variables is operationalized in the same way as it is when it is used as a dependent variable (as described above).

Lastly, a distinction is made between low-income people and high-income people. The effect for every dependent variable is distinguished for those two groups by adding a condition to the regression with ‘if’. What is considered as low-income and what is considered as high-income then? To operationalize this, we used the middle income as a limit. The middle-income for all the used years are added up and divided by the number of years to come with an average middle-income. The average middle-income per month for the used years is 1,732.5 It is decided to round this amount to 1,750 euros per month. Low-income people are those people that have a net income per month less than 1,750 euros per month and high-income people are those people that have a net income that is equal or more than 1,750 euros per month.

5_{(1604+1670+1670+1670+1695+1773+1826+1877+1903) / 9. Note that year 2009 is not used as there were no}

(28)

(29)

- 28 -

5

Results

First, some descriptive statistic will be presented and discussed to show what the dataset looks like. This paragraph shows some main characteristics of the people that participated in the LISS-panel and shows how the most important variables are distributed. Second, the outcomes of different regressions will be presented and discussed. The regressions differ in dependent variable and are executed with and without control variables. Every regression has an additional table whereby the distinction between low-income and high-income is made.

5.1 Descriptive statistics

The first part of this paragraph is about the characteristics of the participants of the LISS-panel. Table 2 combines demographic characteristics and health condition of the participants. The mean of gender is 1.507 which means that there are slightly more women among the participants than men (as 1 stands for male and 2 stands for female). The average age of the participants is 39 years, with the youngest participant being 0 and the oldest participant 103. As described in the previous chapter, the people younger than 18 will not be used in the regression as they are not obliged to take out health insurance and do not have a mandatory deductible for that matter. The net income of the participants is widely distributed with a minimum of 0 and maximum of 298535 euros per month. One might wonder whether this maximum is realistic. We acknowledge this, but as it is hard to distinguish whether people made mistakes while completing the questionnaire we decided to keep these maximum values. The mean of net income per month is 1176.014.

Looking at the health condition, generally speaking we find a mean of 3.116. The participants could choose between five categories, where 1 stands for poor, 2 stands for moderate, 3 stands for good, 4 stands for very good and 5 stands for excellent. Therefore, a mean of 3.116 means that the average health condition the participants described is between good and very good. The table makes a distinction between low-income people (with a net income per month below 1,750 euros) and high-income people (equal or more than 1,750 net income per month). In line with the claim of CBS in the introduction, we find that the health condition of low-income participants (3.078) is somewhat worse than the health condition of high-income participants (3.178).

(30)

- 29 -

Table 2: Demographic characteristics and health condition generally speaking

A. Demographic characteristics

Obs. Mean St. Dev. Min. Max.

Gender 102,913 1.507 0,500 1 2

Age 102,913 39.415 21.760 0 103

Net income (per month)

102,589 1176.772 4216.069 0 298535

B. Health condition, generally speaking

Participants in general

51,395 3.116 .771 1 5

Low-income 31,888 3.078 .781 1 5

High-income 19,507 3.178 .751 1 5

It is also interesting to know how the participants are insured. Table 3 shows whether or not people have a complementary insurance in addition to the obliged insurance. Turns out that 79% of the participants did take out a complementary insurance (for instance for dentistry, physiotherapy or alternative medicine). This shows that the participants are willing to pay extra contribution in order to acquire more certainty that they don’t face high costs when they are in need of dental care or physiotherapy for example. Figure 2 shows that a little more than half of the participants has no voluntary deductible, which means they only face the mandatory deductible. In 2008 this question was not asked, which is also visible in the figure. Also striking is that ten percent of the participants do not know how much their voluntary deductible is. The overall conclusion of the figure is that it seems that people do not prefer to have a higher deductible than obliged. Table 3 and figure 2 together show that the participants of the LISS panel seem to be risk-averse as described in the theoretical framework.

(31)

- 30 -

Table 3: Complementary insurance

Did you take out a complementary health insurance (for instance for dentistry, physiotherapy or alternative medicine)?

Frequent Percent Cumulative

Yes 38,297 78.94 78.94

No 10,219 21.06 100.00

Total 48,516 100.00

Figure 2: Voluntary deductible

So far, the characteristics of the participants. The most important variables in this study are visits2 and visits. As described in the previous chapter, visits 2 shows whether or not people use health care services and visits shows how often people use health care services. The table below shows these important variables and makes a distinction between low-income and high-income participants.

Visits2 has two categories, where 0 stands for used health care service and 1 stands for did not use health care service over the past year. Overall the mean is 0.919 which means that a great share of the people (almost everyone) did use a health care service over the past year. The standard deviation is .273, which looks relatively big but this is not surprising as there are only two absolute options (0 or 1). The means for low-income (0.920) and high-income (0.918) show that there is not much difference between those groups for this variable.

53% 2% 4% 3% 1% 4% 10% 23%

How much is your voluntary deductible?

0 100 200 300 400 500 Don't know Not asked in 2008

(32)

- 31 -

The visits variable has a mean of 10.663 for participants in general, which means that the number of times they used health care services per year is about 11. Again, a big standard deviation with 27.782. This can be explained by the wide range as the minimum value is 0 (did not use health care service at all) and a maximum value of 1404. This maximum value implies that someone used health care services 4 times per day every day of the year. This is not very credible, but it has been decided to include these values in the regressions anyway as it is difficult to make a distinction between incorrectly entered values and actual outliers. The distinction between low-income and high-income participants is relevant for this variable as there is quite a difference. Low-income participants use on average 12 times per year health care services, while high-income participants on average use 9 times per year health care services. This is in line with the claim from CBS that was mentioned in the introduction.

The summaries of visits3 and visits4 are added in appendix C. The visits4 variable showed somewhat smaller means for all three groups, interesting finding that mean for low-income for visits4 is slightly lower than for high-low-income people (thus contrary to what we find here with visits2). The summary of visits3 showed lower maximum values and lower means for all three groups. The ratio between the mean for low-income and high-income people is comparable with the visits variable.

Table 4: Summary of visits2 and visits

A. Visits2 (whether or not people use health care services)

51,119 .919 .273 0 1

Low-income 31,706 .920 .272 0 1

High-income 19,413 .918 .271 0 1

B. Visits (how often people use health care visits per year)

51,119 10.663 27.782 0 1404

Low-income 31,706 11.611 30.096 0 1404

High-income 19,413 9.114 23.437 0 1245