The impact of income and health on the amount of a
voluntary deductible in the case of health insurance in the
Netherlands
Aletta Verberg 10193162 Supervisor: Dr. J.C.M. van Ophem June 24th, 2016
2
Statement of Originality
This document is written by Student Aletta Verberg, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of contents
1. Introduction ... 4
2. Literature review ... 5
2.1. Deductibles and the health care premium ... 5
2.2. Asymmetric information ... 6
2.3. Previous research ... 7
2.4. The default option ... 9
3. Methodology ... 11
3.1. Explanation of the model ... 11
3.2. Variables in the model ... 13
4. Empirical results ... 17
4.1. Changes in the amount of voluntary deductibles in the period 2009 – 2015 ... 17
4.2. Variable descriptives ... 17
4.3. The probit regression ... 19
4.4. The ordered probit regression ... 22
5. Conclusion ... 26
Discussion ... 27
References ... 27
4 1. Introduction
Since the first of January of 2008 there is a new health insurance policy in the Netherlands. The main purpose of the new health insurance policy is to shift the health care expenses from public expenses to private expenses (ECORYS, 2011, p. 15), which is favourable for the government expenses. The policy is also used to create awareness among the consumer about health costs and the use of medical care. This shift of the expenses occurs due to the new implemented mandatory deductible. Besides a mandatory deductible for all insured, a voluntary deductible is an option. Van der Maat and de Jong (2010, p. 8) state that 96% of the insured know about the mandatory deductible. Only 7% of the questioned people choose a voluntary deductible according to data of 2009. Of this 7% more than half of the insured chose a voluntary deductible of €200 or less. What factors influence the decision to take a voluntary deductible and why is the voluntary deductible chosen usually €200 or less?
According to the research of van der Maat and de Jong regarding the effects of a voluntary deductible, the most important reasons to choose a voluntary deductible are the discount on the health insurance costs and the fact that no or very limited medical procedures are expected. The most important reason not to choose a voluntary deductible is that people don’t want to have to worry about their health costs and to be on the safe side. But does the height of income change this consideration? Does income influence the amount of a voluntary deductible in the case of health costs in the Netherlands? And what other factors might influence this choice? This thesis will try to answer these questions.
The structure of this thesis is as follows. The second chapter of this thesis contains a literature review of past research and gives insight about the determinants of variables influencing the choice of taking a voluntary deductible. Additionally, the healthcare system in the Netherlands is being explained. In the third part the model of the multiple regressions is being explained and in the fourth chapter results from the regressions are shown. Lastly, a conclusion is given.
5 2. Literature review
2.1. Deductibles and the health care premium
As mentioned in the first chapter, the deductible for health costs consist of a mandatory part and a voluntary part. This is to increase awareness of health costs among the insured. The amount of the mandatory deductible has increased since the implementation of the new healthcare system (ECORYS, 2011, p. 9). These amounts are listed in Table 1.
Table 1: Mandatory deductible values (HomeFinance, 2016)
2008 2009 2010 2011 2012 2013 2014 2015 2016
Mandatory
deductible € 150 € 155 € 165 € 170 € 220 € 350 € 360 € 375 € 385
On top of the mandatory deductible a voluntary deductible can be chosen. This voluntary deductible can be either € 100, € 200, € 300, € 400 or € 500. In 2009 van der Maat and de Jong (2010, p. 31) found that only 7% of their sample had a voluntary deductible. Striking is the distribution of the amount of the voluntary deductibles taken (
Table 2
), 47.8% has a voluntary deductible of 100 euros, whereas the percentages are decreasing when the amounts increase. Only at 500 euros there is a peak of 18.4%.
Table 2: Insured with or without a voluntary deductible in 2009 (van der Maat & de Jong, 2010, p. 31)
12 observations Number of
Percentage of observations with a
deductible
Percentage of total observations No voluntary deductible 974 - 93.0% Voluntary deductible 74 100.0% 7.0% € 100 35 47.8% 3.3% € 200 15 19.7% 1.4% € 300 9 12.1% 0.9% € 400 1 2.0% 0.1% € 500 14 18.4% 1.3%
The average discount on the health insurance costs is between € 4 and € 25 per month (see Table 8 in appendix), depending on the height of the voluntary deductible (Consumentenbond, 2016). The average premium of a standard health insurance is
6 € 103.77 in 2016. This number is based on various health insurances, all listed in
7 Table
9
, which can be found in the appendix. Having a voluntary deductible of € 500 could mean a discount of approximately 25% on the premium.Since it is mandatory for every citizen in the Netherlands to have a health insurance in the Netherlands (Rijksoverheid, 2016), the government has established a healthcare allowance for citizens with low income to keep healthcare affordable for everyone. This allowance can be received by applying for the healthcare allowance at the tax authorities (Ministerie van Volksgezondheid, 2016, p.10). An insured is only eligible for the healthcare allowance when meeting certain conditions. The insured has to be 18 years or older, has to be a Dutch citizan, has to have an income which is below € 27.012 (in 2016) and his total assets, excluding a house (Belastingdienst, 2016), should be below € 106.641 when his household consists of only one person, otherwise the combined assets should be below € 131.378 (Rijksoverheid, 2016). Between 2008 and 2016 this upper limit for income has fluctuated between € 25.000 and € 35.000 for individuals, while the combined upper limit for income for households was between € 30.000 and € 52.000 (HomeFinance, 2016).
2.2. Asymmetric information
An aspect of the decision making process of taking a voluntary deductible is the provided information about the compulsory and voluntary deductible and the insured health procedures. Van der Maat and de Jong (2010, p. 25) found that 96% of the questioned people in their research in 2009 where informed about the mandatory deductible. But when these people were asked which health services the consumer has to pay of the mandatory deductible, only 74% knew that the costs of going to the general practitioner are not part of the mandatory deductible, but are always paid by their health insurance. As a consequence, consumers might base their decision on wrong information.
Additionally, another form of asymmetric information plays a part. Abbring et al. (2003, p. 513) name averse selection as an important factor when making a choice for an insurance contract. The ‘high risk’ consumers are more likely to choose a high coverage in their insurance and are more likely to use health care. These consumers will not choose a voluntary deductible. Moral hazard can also occur. With moral hazard the behaviour of the consumer changes due to the insurance. In the research of ECORYS (2011, p. 42) on Dutch data a significant negative effect has
8 been found between a compulsory deductible and the demand for health. Van der Maat and de Jong (2010, p. 44) found that 5% of the insured lowered their demand for health due to the compulsory deductible. Moral hazard and adverse selection are hard to distinguish, since both cause a raise in health costs, but it is hard to measure whether this comes from the moral hazard or adverse selection. The consumer can first choose a different contract and then change his behaviour, or it can change his behaviour and therefore choose a different contract. Abbring et al. (2003) try to investigate different methods to separate these effects.
2.3. Previous research
In Australia two different health insurance policies were studied. One of the policies did not cover a lot of health care, while the other was an extended coverage of health care. Cameron & Trivedi (1991) studied the different factor that play a role in the choice between the two policies. Income was the most important factor, people with a higher income chose the extended coverage more often. The indicators for the state of health of the people played only a minor role in this decision. This gives an indication that income is more important in decisions about health care policies.
An investigation to factors influencing the choice on taking a deductible is done by Schellhorn (2001) on Swiss data. The Swiss health system also uses a mandatory and voluntary deductible, similar to the system of the Netherlands. Schellhorn (2001, p. 449) expects that men and women behave different in the choice. Schellhorn therefore assumes gender to be an explanatory factor and corrects for the variable gender by estimating his model for men and women separately. Additionally, age, education, health status and behavioural variables are important factors explaining the choice for a voluntary deductible. Age has a negative effect on the height of the deductible. A possible explanation is that older people tend to have more health problems. Schellhorn (2001, p. 449) gives risk adversity and increased health problems as a possible explanation for this relation. Also being overweight has a negative effect on the probability of choosing a deductible. As most important behavioural variables, smoking and drug use are named. Heavy smokers and stopped smokers are less likely to take high deductibles. In contradiction, female ex-users of hard drugs are more likely to take high deductibles, while users of soft drugs take lower deductibles. Lastly, income has a positive effect on the amount of deductible a person takes.
9 However, the main goal of Schellhorn was to determine whether the choice for a higher deductible influences the behaviour towards demand for health care. He did not find significant effects and therefore there is not enough evidence to state that the behaviour changes are due to a higher voluntary deductible. Van Kleef, van de Ven and van Vliet (2009) did find that having a mandatory and compulsory deductible is a useful tool to influence the demand of health care. The trade-off between financial risk and the expected medical care is more efficient due to the deductibles and therefore moral hazard has decreased.
Berkhout and van Ophem (2012) analysed the effect of moral hazard and adverse selection on the choice of taking a voluntary deductible in the Netherlands. The adverse selection arises since people who take more insurance are the ones who expect need for health care or are already unhealthy. Moral hazard occurs when people change their behaviour due to not having a deductible. People could be less careful since they will not pay extra costs in case of more demand for health care, or could be more careful due to a high voluntary deductible. Berkhout and van Ophem also took the risk adversity of an individual into account in their analysis. Some people tend to be more risk adverse and are more likely to not choose a voluntary deducitble. Berkhout and van Ophem eventually found that only moral hazard, gender, living with a partner, income and risk attitude influence the choice of taking a voluntary deductible. Moral hazard has a negative effect, which means that consumers that can adjust their health care demand downwards, will do so to take the voluntary deductible. Women and people living with a partner are more likely not to choose a deductible, while a higher income and more risk taking people are more likely to choose a voluntary deductible.
Concluding, previous research found that age, education, health status, behavioural factors such as smoking and drug use, income, gender, living with a partner, moral hazard and risk attitude are significant variables in the choice for a voluntary deductible. There is a negative relation between having a deductible and being female, moral hazard, living with a partner, age and being an ex-smoker and a positive relation between having a deductible and income, health status and education. Because of the complexity of distinguishing moral hazard and adverse selection, these variables will not explicitly be taken into account in this thesis, but these variables may play a role in the choice for a voluntary deductible.
10 2.4. The default option
When looking at the decision making process of taking a voluntary deductible, the consumer faces a decision with a default option. When a consumer chooses an insurance for the first time at a new insurance company, the default for the voluntary deductible is € 0. The consumer needs to change the amount to get a higher voluntary deductible. At the end of the year, a consumer has the opportunity to change its health insurance (and insurance company). If a consumer does nothing, the insurance is automatically extended and therefore the amount of voluntary deductible remains the same. If the consumer does change his insurance at the same insurance company, he can change his amount of voluntary deductible, but the default option is the same as the year before. At a new insurance company, the default is again € 0.
Bellman, Johnson and Lohse (2001) investigated the effect of a default option. They found that having a default option influences the behavior of the consumer in their decision-making process. Consumers tend to choose an option more often when it is the default option than in the case with the same options but when there is no default option. A default option in health insurance has as result that there are two moments where the consumer gets a nudge to keep a voluntary deductible of € 0. The first moment is when a new insurance policy is chosen and the second moment is when the insurance is automatically extended. The opt-in option may influence the choice of taking a voluntary deductible. On the other hand, when a consumer has chosen a voluntary deductible, this amount becomes the default until another amount is chosen.
Additionally, Johnson and Goldstein (2003) state that having default options influence the choice of consumers in three different ways. They state that having an default implies an recommended action by the policy-makers for the consumers, which therefore is used by the consumers. In the case of deductibles, the consumer experiences the no voluntary deductible category as recommended action. Not choosing the default also requires effort, in this case of being informed about the voluntary deductible and it’s advantages and disadvantages, while accepting the default of € 0 is easy and the psychological effort, since people don’t want to worry about their health costs (van der Maat & de Jong, 2010, p. 8). Lastly, Johnson and Goldstein state that a change often involves a trade-off. People tend to give more value to loosing an amount than gaining the same amount, which means that people
11 need to be compensated for more than what they loose. This is also called loss aversion. Johnson and Goldstein make the conclusion that changing the default option would change the choices the consumers would make and that policy-makers should choose a default which requires no action.
Because of the default option, there may be an alternative decision model for the choice to take a voluntary deductible. Instead of choosing between € 0, € 100, € 200, € 300, € 400 or € 500, the choice could be between taking the default option (and choose no voluntary deductible) or not taking the default option, followed by the choice between the amount when a voluntary deductible is chosen. There are now two moments of decision making. These choices could be correlated, but for this thesis no correlation is assumed for simplicity reasons.
To find factors influencing the decision making process, the default option of the voluntary deductible on € 0 is most likely an explanatory aspect and could cause the low group of consumers that choose a voluntary deductible. This could lead to look at the decision process with an alternative decision model where the choice between the amount and taking a voluntary deductible are separated.
12 3. Methodology
In this paper LISS (Longitudinal Internet Studies for the Social sciences) panel data is used. The data is administered by CentERdata (Tilburg University, The Netherlands). The LISS panel data is a sample of Dutch individuals who participate in monthly Internet surveys. The panel is based on a true probability sample of households drawn from the population register. A survey is taken in the panel every year, covering a large variety of domains including work, education, income, housing, and personality (CentERdata, 2016). This particular data set is a combination of the background variables and a survey especially focused on health issues. In the different sections different waves are used. These waves correspond to the year in which the surveys were taken. The main focus of this thesis is on the data from 2015, health wave 8 and the background variables of July 2015. In total there are 2592 observations used for 2015.
3.1. Explanation of the model
Two different models are used in this thesis. In the first model, the dependent variable is a binary variable with only two values, an insured either has a voluntary deductible (D=1) or it has no voluntary deductible (D=0). Since the outcome is either zero or one, a nonlinear model a model holding the outcome between zero and one is needed. Therefore, the probit model is used. The outcome of the probit model gives the probability of choosing a voluntary deductible when the variables have certain values. The model is as follows (Stock & Watson, 2012, p. 432):
𝑃 𝐷 = 1 𝑋&, … , 𝑋)) = Φ(𝛽. + 𝛽&𝑋&+ ⋯ + 𝛽)𝑋)
In this model 𝑋) are the different explanatory variables, the 𝛽) are the matching parameters and Φ is the cumulative standard normal distribution function. The outcomes of the model are probabilities. The marginal effect of a unit change in X1 is the coefficient 𝛽& multiplied by the density, holding the other variables in the model, 𝑋2, … , 𝑋), constant (Stock & Watson, 2012, p. 432).
Secondly, the different values of the voluntary deductible are taken into account. The voluntary deductible is either € 0, € 100, € 200, € 300, € 400 or € 500. There are now six categories representing the amount of the deductible. The
13 dependent variable has become an ordinal variable. Therefore, the ordered probit regression described by Long and Freese (2006, p. 184) is used. The structural model includes a latent variable between −∞ and +∞ and is as follows:
𝑦6∗= 𝛽
. + 𝛽&𝑋&+ ⋯ + 𝛽)𝑋) + 𝜀6
The measurement model divides 𝑦∗ into 𝐽 catgories, which is 6. The values of 𝑦 6 depend on the cutpoints 𝜆;. The cutpoints 𝜆. and 𝜆; are defined as −∞ and +∞. The error term 𝜀6 is assumed to be normally distributed and 𝜎2 = 1 is imposed for identification reasons.
𝑦
6=
0 𝑖𝑓 𝑦
6∗≤ 𝜆
.1 𝑖𝑓 𝜆
.< 𝑦
6∗≤ 𝜆
&2 𝑖𝑓 𝜆
&< 𝑦
6∗≤ 𝜆
23 𝑖𝑓 𝜆
2< 𝑦
6∗≤ 𝜆
D4 𝑖𝑓 𝜆
D< 𝑦
6∗≤ 𝜆
F5 𝑖𝑓 𝜆
F< 𝑦
6∗𝑃 𝑦
6∗= 𝑗 = 𝑃(𝜆
IJ&< 𝑦
6∗≤ 𝜆
I)
𝑗 = 0, 1, … , 5
As mentioned in the previous chapter, there may be a decision model with two decision moments. This is illustrated in Figure 1
: Alternative decision model
. In the alternative decision model a consumer first chooses between having a voluntary deductible or not, and if he chooses a voluntary deductible, he chooses the amount. An ordered probit regression determines the estimators of second decision moment, with 5 possible options.
Figure 1: Alternative decision model
Voluntary deductible? Yes € 100 € 200 € 300 € 400 € 500 No
14 The first decision is determined by the first model:
𝑃 𝐷 = 1 𝑋&, … , 𝑋)) = Φ(𝛽. + 𝛽&𝑋&+ ⋯ + 𝛽)𝑋)
The model of the second decision is now:
𝑦
6=
1 𝑖𝑓 𝑦
6∗≤ 𝜆
&2 𝑖𝑓 𝜆
&< 𝑦
6∗≤ 𝜆
23 𝑖𝑓 𝜆
2< 𝑦
6∗≤ 𝜆
D4 𝑖𝑓 𝜆
D< 𝑦
6∗≤ 𝜆
F5 𝑖𝑓 𝜆
F< 𝑦
6∗𝑃 𝑦
6∗= 𝑚 = 𝑃(𝜆
LJ&< 𝑦
6∗≤ 𝜆
L)
𝑚 = 1, … , 5
3.2. Variables in the modelAs mentioned in section 3.1., the dependent variable in the probit regression will be a dummy variable. The individual either has a deductible (D=0) or does not (D=1). When using the ordered probit regression, the dependent variable DD has multiple values, either € 0, € 100, € 200, € 300, € 400 or € 500. In the survey of the data an option ‘I don’t know the amount of deductible’ included. This observations with this outcome are dropped in the regressions, since the main purpose of this thesis is to determine why a deductible is chosen. Now people that are not informed enough about their voluntary deductible are taken out of the sample. The latent variable DD is defined by:
𝐷𝐷 = 𝐷0 ∗ 0 + 𝐷1 ∗ 100 + 𝐷2 ∗ 200 + 𝐷3 ∗ 300 + 𝐷4 ∗ 400 + 𝐷5 ∗ 500
For the dependent variable the following variables are used:
D Dummy variable with a value of 1 if the individual has a voluntary deductible, 0 otherwise.
D0 Dummy variable with a value of 1 if the individual has a deductible of € 0, 0 otherwise.
15
D1 Dummy variable with a value of 1 if the individual has a deductible of € 100, 0 otherwise.
D2 Dummy variable with a value of 1 if the individual has a deductible of € 200, 0 otherwise.
D3 Dummy variable with a value of 1 if the individual has a
deductible of € 300, 0 otherwise.
D4 Dummy variable with a value of 1 if the individual has a deductible of € 400, 0 otherwise.
D5 Dummy variable with a value of 1 if the individual has a deductible of € 500, 0 otherwise.
DD Latent variable of amount of deductible consisting of D0, D1, D2, D3, D4 and D5.
Since the effects of moral hazard and adverse selection are difficult to separate and measuring the opt-in effect is too complex for the purpose of this thesis, these variables have not been taken into account in this thesis. Based on previous research the following explanatory variables are part of the analysis:
Female Dummy variable with a value of 1 if the individual is female, 0
otherwise.
Age Age of individual with a minimum age of 18 years.
Age >65 Dummy variable with a value of 1 if the individual is older than
65 years old, 0 otherwise.
Income Gross monthly income in euros of an individual.
Income 0 Dummy variable with a value of 1 if the individual has no gross
income, 0 otherwise. This variable is not in the model, but used as base category.
Income 1 - 500 Dummy variable with a value of 1 if the individual has a gross
income between € 1 and € 500 per month or less, 0 otherwise.
Income 501 – 1000 Dummy variable with a value of 1 if the individual has a gross
income between € 501 and € 1000 per month, 0 otherwise.
Income 1001 – 1500 Dummy variable with a value of 1 if the individual has a gross
16
Income 1501 – 2000 Dummy variable with a value of 1 if the individual has a gross
income between € 1501 and € 2000 per month, 0 otherwise.
Income 2001 – 2500 Dummy variable with a value of 1 if the individual has a gross
income between € 2001 and € 2500 per month, 0 otherwise.
Income 2501 – 3000 Dummy variable with a value of 1 if the individual has a gross
income between € 2501 and € 3000 per month, 0 otherwise.
Income 3001 – 4000 Dummy variable with a value of 1 if the individual has a gross
income between € 3001 and € 4000 per month, 0 otherwise.
Income >4000 Dummy variable with a value of 1 if the individual has a gross
income above € 4000 per month, 0 otherwise.
Partner Dummy variable with a value of 1 if the individual has a
relationship and lives together with this partner, 0 otherwise.
Primary school Dummy variable with a value of 1 if the education level of the
individual is primary school, 0 otherwise. This variable is not in the model but used as base category.
VMBO Dummy variable with a value of 1 if the education level of the
individual is VMBO (intermediate secondary education), 0 otherwise.
HAVO/VWO Dummy variable with a value of 1 if the education level of the
individual is HAVO or VWO (higher secondary education), 0 otherwise.
MBO Dummy variable with a value of 1 if the education level of the
individual is MBO (intermediate vocational education), 0 otherwise.
HBO Dummy variable with a value of 1 if the education level of the
individual is HBO (higher vocational education) 0 otherwise.
WO Dummy variable with a value of 1 if the education level of the individual is WO (university), 0 otherwise.
Child Dummy variable with a value of 1 if the individual has children
under 18 years old living at home, 0 otherwise.
Job Dummy variable with a value of 1 if the individual has a paid
17
Average health Dummy variable with a value of 1 if the individual describes
health as poor or moderate, 0 otherwise. This variable is not in the model, but used as base category.
Good health Dummy variable with a value of 1 if the individual describes
health as good, 0 otherwise.
Very good health Dummy variable with a value of 1 if the individual describes
health as very good, 0 otherwise.
Excellent health Dummy variable with a value of 1 if the individual describes
health as excellent, 0 otherwise.
BMI Body Mass Index. Weight in kilograms divided by square of
length in metres.
Chronically ill Dummy variable with a value of 1 if the individual suffers from
a long-standing disease, affliction, handicap or consequences of an accident, 0 otherwise.
Smoker Dummy variable with a value of 1 if the individual currently
smokes, 0 otherwise.
Ex-smoker Dummy variable with a value of 1 if the individual used to
smoke in the past, 0 otherwise.
Hospital Dummy variable with a value of 1 if the individual has spent
time in a hospital or clinic in the past 12 months, 0 otherwise.
Complementary Dummy variable with a value of 1 if the individual has a
complementary health insurance, 0 otherwise.
Allowance Dummy variable with a value of 1 if the individual received
health care allowance in the previous year, 0 otherwise.
Alcohol Dummy variable with a value of 1 if the individual drinks
18 4. Empirical results
4.1. Changes in the amount of voluntary deductibles in the period 2009 – 2015 In Table 3 the amount of voluntary deductibles chosen in the dataset used in this thesis are shown, starting from 2009. It is clear that the number of people that don’t know their amount of voluntary deductible is declining over the years, which means that people are better informed nowadays. There also is a decline in the number of people that choose a deductible of €100, and an increase in the number of people that choose €500. The total amount of people choosing a voluntary deductible has slightly increased from 2009 until 2015, but is still not higher than about 20%. The percentage of people with no voluntary deductible remains at about 70%.
Table 3: Overview percentage voluntary deductibles 2008-2015 from CentERdata (2016)
Voluntary deductible amount 2009 (%) 2010 (%) 2011 (%) 2012 (%) 2013 (%) 2015 (%) No deductible 69.38 70.90 68.26 70.11 72.65 70.89 Deductible 14.32 14.57 17.93 17.74 17.32 20.12 € 100 5.42 4.44 3.70 1.95 1.63 0.91 € 200 5.64 6.31 9.51 8.29 2.55 2.67 € 300 1.08 1.30 1.44 3.94 6.29 5.48 € 400 0.30 0.27 0.49 0.60 0.76 1.97 € 500 1.88 2.25 2.79 2.96 6.09 9.09 Don’t know 16.30 14.54 13.80 12.15 10.03 8.99
When comparing the numbers of this dataset to the numbers from van der Maat and de Jong (2010), this data has substantially more people with voluntary deductibles in the sample, since in the sample of van der Maat and de Jong only 7% chose for a voluntary deductible in 2009 (see Table 2), which were 74 observations. In this dataset it was almost 15% in 2009. This indicates that the sample used may not be fully representative, or that the sample of van der Maat and de Jong is not fully representative since the data sample used in this thesis is larger.
4.2. Variable descriptives
When looking at the data of 2015 that is used for the regressions on the deductibles, the observations where the individual did not know their amount of voluntary deductible are now excluded from the sample, it is clear that choosing no deductible
19 is most common (77.9%, see Table 4). What stands out about the frequencies of the people choosing a deductible is the frequency of the choice for a deductible of 500 euros. Out of the 573 people with a deductible, 259 choose for the highest amount possible, which is 45.2%.
Table 4: Overview deductibles in 2015
Amount of voluntary deductible Number of observations Percentage of observations with a
deductible
Percentage of total observations No voluntary deductible 2019 - 77.89% Deductible 573 100.00% 22.11% € 100 26 4.54% 1.00% € 200 76 13.26% 2.93% € 300 156 27.23% 6.02% € 400 56 9.77% 2.16% € 500 259 45.20% 9.99%
In Table 5 there is an overview of all the variables used in the probit and ordered probit regression. In total there are 2592 observations. The variables ‘Income 0’, ‘Average health’ and ‘Primary school’ are base categories and are not in the regressions.
Table 5: Variable descriptives
Variable Obs Mean Std. Dev. Min Max
D 2592 .2210648 .415044 0 1 DD 2592 83.52623 167.814 0 500 Female 2592 .5408951 .498421 0 1 Age 2592 55.87616 161.963 18 93 Age >65 2592 .3499228 .477037 0 1 Income 2592 2372.552 4495.814 0 214264 Income 0 2592 .0748457 .263193 0 1 Income 1 – 500 2592 .0289352 .1676567 0 1 Income 501 – 1000 2592 .1145833 .3185799 0 1 Income 1001 – 1500 2592 .1277006 .3338206 0 1 Income 1501 – 2000 2592 .1427469 .3498821 0 1 Income 2001 – 2500 2592 .1334877 .3401666 0 1 Income 2501 – 3000 2592 .1246142 .3303447 0 1 Income 3001 – 4000 2592 .1392747 .3462998 0 1 Income >4000 2592 .0578704 .2335432 0 1 Partner 2592 .6300154 .4828933 0 1
20 Primary school 2592 .0636574 .2441888 0 1 VMBO 2592 .2426698 .4287798 0 1 HAVO/VWO 2592 .0945216 .2926095 0 1 MBO 2592 .2364969 .4250127 0 1 HBO 2592 .255787 .4363869 0 1 WO 2592 .1068673 .3090041 0 1 Child 2592 .285108 .451553 0 1 Job 2592 .5852623 .4927718 0 1 Average health 2592 .2002315 .4002507 0 1 Good health 2592 .586034 .4926376 0 1
Very good health 2592 .1697531 .3754882 0 1
Excellent health 2592 .0439815 .2050935 0 1 BMI 2592 25.80166 4.455571 11.6921 62.1025 Chronically ill 2592 .3665123 .4819447 0 1 Smoker 2592 .1651235 .3713636 0 1 Ex-smoker 2592 .4347994 .4958263 0 1 Hospital 2592 .1157407 .3199755 0 1 Complementary 2592 .7719907 .4196296 0 1 Allowance 2592 .1983025 .3987981 0 1 Alcohol 2592 .2503858 .4333188 0 1
4.3. The probit regression
The results of the probit regression and the marginal effects of the probit regressions can be found in Table 6. The effects differ between the two regressions, but the direction of the effect and the significance of the variables are the same throughout the regressions.
The estimation of the effect of being older than 65 years has a significant negative effect on having a voluntary deductible, possibly because of the health problems that come with age. Being female results in a lower probability of choosing a deductible, the same as was found in other literature. Even though other literature found that education had an influence on the choice of a deductible, this regression only accepts that having a university degree is positive for the probability at a 10% significance level.
When looking at the effects of the income categories, only the highest two categories show a negative significant effect at a 10% level. Van der Maat and de Jong (2010) already suggested that people don’t want to worry about their health costs and that one of the main reasons to choose a deductible is because of the
21 discount. Together this could explain that people with income above € 3000 per month care less about the discount than the (psychological) effort they would have to make for the deductible, than the people with a lower income. People with a lower income choose voluntary deductibles more often. These results are different from previous studies, but because of the 10% significance level this needs further research.
The valuation of the own health gives the highest significant effects. Valuating the own health as very good or excellent has a positive effect on choosing a deductible, compared to people who think they aren’t as healthy. People who consider themselves as healthy may not expect high health costs and may benefit from the discount. A higher BMI gives a lower probability to get a deductible. Being chronically ill also has a significant negative effect, which supports the reasoning that healthy people choose voluntary deductibles more. Additionally, the people that stopped smoking have a lower probability to choose for a voluntary deductible than people that did not smoke at all, or are still smoking. Usually (stopped) smokers care less about their health, which explains the effect. Lastly, having a complimentary health insurance has a negative effect on choosing a voluntary deductible. These people expect other costs, or are very risk adverse, causing less chosen voluntary deductibles. However, this variable is endogenous and therefore the effect is not the best estimate.
The marginal effects of the probit regression give being older than 65, being female, valuation of health, being chronically ill, BMI and being a stopped smoker as significant results. These are mostly in line with previous research and theories. The valuation of health gives the highest positive effect.
Table 6: Marginal effects of probit regression
Variable D D Marginal effects Female -0.314** -0.0860** (0.0656) (0.0181) Age >65 -0.326** -0.0846** (0.105) (0.0260) Partner 0.0131 0.0035 (0.0671) (0.0182) Income 1 – 501 0.210 0.0618 (0.178) (0.0563)
22 Income 501 – 1000 0.0147 0.0040 (0.126) (0.0344) Income 1001 – 1500 -0.0502 -0.01340 (0.124) (0.0326) Income 1501 – 2000 -0.217+ -0.0548+ (0.120) (0.0280) Income 2001 – 2500 -0.110 -0.0287 (0.116) (0.0292) Income 2501 – 3000 -0.160 -0.0412 (0.118) (0.0286) Income 3001 – 4000 -0.225+ -0.0567+ (0.115) (0.0267) Income >4000 -0.276+ -0.0667+ (0.145) (0.0309) VMBO -0.118 -0.0310 (0.147) (0.0377) HAVO / VWO 0.0878 0.0246 (0.165) (0.04754) MBO 0.0224 0.0061 (0.148) (0.0406) HBO 0.215 0.0609 (0.149) (0.0439) WO 0.312+ 0.0934+ (0.163) (0.0530) Child 0.0983 0.02716 (0.0716) (0.0201) Job 0.0555 0.0150 (0.0933) (0.0251) Good health 0.131 0.0353 (0.0973) (0.0258)
Very good health 0.389** 0.1170**
(0.116) (0.0379) Excellent health 0.581** 0.1903** (0.159) (0.0592) BMI -0.0159* -0.0430* (0.00735) (0.00199) Chronically ill -0.319** -0.0834** (0.0741) (0.0185) Smoker -0.0274 -0.00739 (0.0861) (0.0230) Ex-smoker -0.192** -0.0516** (0.0684) (0.0181) Hospital -0.103 -0.0269
23 (0.105) (0.0264) Complementary -0.509** -0.154** (0.0662) (0.0217) Allowance -0.0940 -0.0249 (0.0832) (0.0214) Alcohol 0.0200 0.00544 (0.0731) (0.0200) _cons 0.240 (0.289) N 2592 2592
Standard errors in parentheses + p < 0.1, * p < 0.05, ** p < 0.01
4.4. The ordered probit regression
To determine the factors influencing the amount of the voluntary deductible, an ordered probit regression has been done. In Table 7 the output of the ordered probit regression including and excluding a voluntary deductible of € 0 can be found. The results of the regression with six categories gives no significant cuts. There is not enough evidence to assume that the effects differ between the categories.
The output excluding the zero category gives two out of four significant cuts, the estimates for 𝜆I, at a level of 1%. The third and fourth cut are not significant. The sample has now decreased to 573 observations. A reason for the difference in significance in the estimates for 𝜆Imay be explained by the alternative decision model as described in the second chapter. The choice on the amount of voluntary deductible may be a choice after the decision is made to take a voluntary deductible. According to the regression without the zero category (the second column in Table 7: Output
ordered probit regression
being 65 years or older and living with a partner are significant estimates and have a negative relation with the amount of voluntary deductible, compared to not living with a partner. All the income categories are not significant, therefore income does not have an effect on the amount of voluntary deductible according to this regression. Education has the biggest effect. The probability that individuals with a university degree choose a higher amount of voluntary deductible is higher than the probability for individuals that did not go to high school, since the estimation of having a university degree is significant. Having
24 a complementary health insurance has again a negative effect on the amount of voluntary deductible chosen, but the variable still might be endogenous.
Despite the cuts that are not significant in the ordered probit regression on the six categories, some estimates were significant as well. Being older than 65 years decreases the probability of choosing a higher amount of voluntary deductible and having a higher education has a positive effect, which is the same as in the regression on the five categories. Additionally, being female now has a negative significant effect, just as BMI, being an ex-smoker and being chronically ill. Considering your health as very good or excellent increases the probability of taking a higher amount of voluntary deductible. Having an income above 4000 euros has a positive effect, but only at a 10% significance level. The variable living with a partner is not significant. These results were also found earlier, in the probit regression. According to the ordered probit regression for the second choice of the alternative decision model, being 65 years or older, living with a partner and having a university degree are the variables with significant estimators to explain the amount of voluntary deductible chosen. This deviates from the variables that were found to determine whether people choose a voluntary deductible or not, which is the first decision in the alternative decision model. The results of the ordered probit regression on the six categories give roughly the same results as the probit regression in section 4.3.
Table 7: Output ordered probit regression
Variable (1) DD Including 0 category (2) DD Excluding 0 category Female -0.319** -0.109 (0.0631) (0.108) Age >65 -0.344** -0.364* (0.102) (0.180) Partner -0.0175 -0.241* (0.0645) (0.111) Income 1 – 500 0.197 -0.0191 (0.169) (0.263) Income 501 – 1000 0.0333 -0.0464 (0.121) (0.205) Income 1001 – 1500 -0.0135 0.181 (0.119) (0.201) Income 1501 – 2000 -0.197+ -0.0742
25 (0.115) (0.191) Income 2001 – 2500 -0.0741 0.0844 (0.111) (0.181) Income 2501 – 3000 -0.155 -0.118 (0.112) (0.178) Income 3001 – 4000 -0.150 0.293 (0.109) (0.179) Income >4000 -0.235+ 0.0735 (0.137) (0.219) VMBO -0.0422 0.474+ (0.145) (0.268) HAVO / VWO 0.160 0.465 (0.161) (0.290) MBO 0.0337 0.0722 (0.146) (0.264) HBO 0.280+ 0.470+ (0.146) (0.263) WO 0.430** 0.780** (0.159) (0.283) Child 0.0832 -0.0927 (0.0686) (0.113) Job 0.0522 -0.0438 (0.0894) (0.144) Good health 0.159+ 0.131 (0.0954) (0.181)
Very good health 0.417** 0.239
(0.112) (0.201) Excellent health 0.594** 0.280 (0.150) (0.249) BMI -0.0192** -0.0217+ (0.00713) (0.0118) Chronically ill -0.322** -0.143 (0.0719) (0.131) Smoker -0.0700 -0.192 (0.0826) (0.135) Ex-smoker -0.183** -0.00950 (0.0659) (0.116) Hospital -0.128 -0.312+ (0.102) (0.190) Complementary -0.553** -0.467** (0.0628) (0.103) Allowance -0.0854 0.0151 (0.0802) (0.137)
26 Alcohol 0.0502 0.197 (0.0704) (0.124) Cut 1 -0.282 -2.448** (0.281) (0.499) Cut 2 -0.242 -1.617** (0.281) (0.491) Cut 3 -0.119 -0.740 (0.281) (0.489) Cut 4 0.185 -0.459 (0.281) (0.488) Cut 5 0.324 (0.281) N 2592 573
Standard errors in parentheses + p < 0.1. * p < 0.05. ** p < 0.01
27 5. Conclusion
Since the implementation of the new health care policy in the Netherlands, people can choose for a voluntary deductible. The number of people that choose a voluntary deductible is increasing, but has not been higher than about twenty percent of the population.
Determining the variables that play a role in the choice of a voluntary deductible or not, the probit regression gives being older than 65, being female, the valuation of health, being chronically ill, BMI and being a stopped smoker as significant variables. The valuation of health gives the highest positive effect. These results are also found in the ordered probit regression on six categories. According to the ordered probit regression, using the alternative decision model with two consecutive independent choices, being 65 years or older, living with a partner, and having a university degree are variables explaining the amount of voluntary deductible chosen. The results of the two ordered probit regressions differ from each other.
The main conclusion from this thesis is that the own valuation of people about their health is an important variable in choosing a voluntary deductible an. An effect of income can not be stated. Having a university degree and being older than 65 are important indicators for choosing a voluntary deductible and choosing the amount.
28 Discussion
In this thesis some problems occurred which could influence the results. There is only a small group of people choosing a voluntary deductible, which makes it difficult to do the ordered probit regression to determine the variables that influence the choice on the amount of voluntary deductible. There is an information problem about the voluntary deductible and the default option could influence the behaviour of the consumers.
The people that did not know whether they had a voluntary deductible or not have not been taken into account in this thesis. However, this group is most likely not informed enough and therefore probably have taken the default option of € 0. When this is the case, this could change the results of the regressions. There were some variables not taken into account even though other literature mentioned these variables to be important. Moral hazard, adverse selection and risk adversity should be investigated more.
The assumption for no correlation between the process of choosing a deductible, and the process of choosing the amount of the voluntary deductible may not be true. More research to this result is needed to state this assumption and the two processes in stead of one.
Further research to this topic could be done to the effect of the default option of no voluntary deductible. Dividing the decision process into two decision moments, first choosing a voluntary deductible or not, and after that the choice of the amount is also a topic to elaborate on, including the correlation between these two decisions.
29 References
Abbring, J. H., Chiappori, P. A., Heckman, J. J., & Pinquet, J. (2003). Adverse selection and moral hazard in insurance: can dynamic data help to
distinguish. Journal of the European Economic Association , 2/3 (1), 512-521.
Belastingdienst. (2016). Vermogen. Retrieved on May 5th, 2016, from
Belastingdienst:
http://www.belastingdienst.nl/wps/wcm/connect/bldcontentnl/belastingdienst/p rive/toeslagen/hoe_werken_toeslagen/kan_ik_toeslag_krijgen/vermogen/ Bellman, S., Johnson, E. J., & Lohse, G. L. (2001). To opt-in or opt-out?: it depends on the question. Communications of the ACM , 44 (2), 25-27.
Berkhout, P., & van Ophem, H. (2012). Adverse selection and moral hazard in the choice of a deductible in health costs insurance: an empirical analysis on Dutch data. Department of Quantitative Economics of the University of
Amsterdam.
Cameron, A. C., & Trivedi, P. K. (1991). The role of income and health risk in the choice of health insurance. Journal of Public Economics , 45, 1-28.
CentERdata. (2016). References. Retrieved on February 10th, 2016, from LISSdata: http://www.lissdata.nl/assets/uploaded/References_LISS.pdf
Consumentenbond. (2016). Korting bij verhogen eigen risico. Retrieved on February
14th, 2016, from Consumentenbond:
http://www.consumentenbond.nl/zorgverzekering/zorgverzekering/eigen- risico-zorgverzekering-overzicht/korting-bij-verhogen-eigen-risico/
Consumentenbond. (2016). Premieoverzicht zorgverzekering. Retrieved on February 28th, 2016, from Consumentenbond:
http://www.consumentenbond.nl/zorgverzekering/zorgverzekering/zorgpremie -overzicht/premieoverzicht-zorgverzekering/
ECORYS. (2011). Evaluatie naar het verpicht eigen risico. Rotterdam: Ministerie van Volksgezondheid, Welzijn en Sport.
HomeFinance. (2016). Informatie over zorgverzekering. Retrieved on May 5th, 2016, from HomeFinance:
http://www.homefinance.nl/zorgverzekering/informatie/informatie.asp HomeFinance. (2016). Zorgtoeslag. Retrieved on February 14th, 2016, from HomeFinance:
30 Johnson, E. J., & Goldstein, D. G. (2003). Do Defaults Save Lives? Science (302), 1338-1339.
Long, J., & Freese, J. (2006). Regression Models for Categorical Dependent Variables Using Stata. Stata Press.
Ministerie van Volksgezondheid, W. e. (2016). Het Nederlandse zorgstelsel. Den Haag: Rijksoverheid.
Rijksoverheid. (2016). Het zorgverzekeringsstelsel in Nederland. Retrieved on February 14th , 2016, from Rijksoverheid:
https://www.rijksoverheid.nl/onderwerpen/zorgverzekering/inhoud/zorgverzek eringsstelsel-in-nederland
Rijksoverheid. (2016). Wanneer heb ik recht op zorgtoeslag? Retrieved on February 14th, 2016, from Rijksoverheid:
https://www.rijksoverheid.nl/onderwerpen/zorgverzekering/vraag-en- antwoord/wanneer-heb-ik-recht-op-zorgtoeslag
Schellhorn, M. (2001). The effect of variable health insurance deductibles on the demand for physician visits. Health Economics (10), 441-456.
Stock, J., & Watson, M. (2012). Introduction to Econometrics. Pearson.
Van der Maat, M., & de Jong, J. (2010). Eigen risico in de zorgverzekering: het verzekerden perspectief. Utrecht: NIVEL.
Van Kleef, R., van de Ven, W., & van Vliet, R. (2009). Shifted deductibles for high risks: More effective in reducing moral hazard than traditional deductibles. Journal of Health Economics (28), 198-209.
31 Appendix
Table 8: Average discount voluntary deductible (Consumentenbond, 2016)
Insurance company €100 €200 €300 €400 €500 Anderzorg €48 €84 €132 €168 €288 Anderzorg €48 €84 €132 €168 €288 Avéro Achmea €44.40 €88.80 €150 €200.04 €249.96 Azivo €36 €72 €108 €144 €180 Bewuzt €36 €72 €108 €144 €180 CZ Zorgbewust nvt nvt nvt nvt €210 CZ Zorg-op- maat/Zorgkeuze €36 €72 €108 €144 €210 CZdirect nvt nvt nvt nvt €239.40 De Amersfoortse €50.04 €100.08 €150 €200.04 €300 De Friesland €51 €99 €144 €198 €264 Delta Lloyd €39.96 €80.04 €120 €159.96 €200.04 Ditzo €50.04 €85.08 €140.04 €200.04 €270 DSW €48 €96 €144 €192 €276 FBTO €50.04 €99.96 €150 €200.04 €249.96 Hema nvt nvt nvt nvt €288 Kiemer nvt nvt nvt nvt €264 Menzis Basis Voordelig nvt nvt nvt nvt €264 Menzis Basis / Basis Vrij €36 €72 €108 €144 €240 Ohra nvt nvt nvt nvt €200.04 ONVZ nvt nvt nvt nvt €300 OZF Achmea €44.40 €88.80 €133.20 €177.60 €222 PNOzorg €40.80 €78 €114 €147.60 €300 Pro Life €44.40 €88.80 €133.20 €177.60 €222 Salland €36 €72 €108 €144 €210 Salland ZorgDirect €36 €72 €108 €144 €210 Stad Holland €48 €96 €144 €12 €276 Univé €48 €72 €108 €144 €180 Zekur Gewoon nvt nvt nvt nvt €180
Zekur Gewoon Vrij €48 nvt nvt nvt €183.60
VGZ €36 €72 €108 €144 €180
Zilveren Kruis €44.40 88.80 €133.20 €177.60 €222
ZieZo (Zilveren
Kruis) nvt nvt nvt nvt € 264
Zorg & Zekerheid €72 €120 €165 €219 €300
AVERAGE PER
YEAR € 45 € 85 € 128 € 163 € 294 AVERAGE PER
MONTH € 4 € 7 € 11 € 14 € 25