Who is Buying Voluntary Private Health Insurance in Portugal: A Comparative Analysis for 2014 and 2019

Abstract

The Portuguese health system is defined as a National Health Service with universal health coverage of the population and almost free access to health care at any point of delivery. Despite this, the percentage of people who report having voluntary private health insurance has increased from 16% to 20.5% between 2014 and 2019. This paper aims to estimate the main determinants for having voluntary private health insurance in 2014 and 2019. We use data collected by the National Health Survey of 2014 and 2019 to compare results. A logistic model is estimated to explain the decision to hold an insurance policy. The results show that despite the increase in the number of people with private health insurance, the determinants are similar. Except for the role of being male, having had flu vaccination and being unemployed, which became significant in 2019. The most relevant results are (i) people who benefit from health subsystems, (ii) people who report long waiting times for medical care, (iii) people who have been vaccinated against the flu, and (iv) people who report unmet health needs are less likely to have private health insurance. The results of this paper indicate some potential inequalities in access to health care.

1. Introduction

Voluntary and private health insurance (VPHI) plays a dual role in countries such as Portugal [1] or the UK, complementing and supplementing the National Health Services (NHS).

The acquisition of health insurance is of great interest to individuals because of its ability to provide expeditious access to health services, the freedom to choose preferred healthcare providers, enhance the overall experience of hospitalisation, and grant access to medical services outside the scope of the NHS, such as dental care.

Portuguese NHS is characterised by its universal coverage of the population and a wide package of health services at nearly no cost at the point of delivery [1]. The NHS suffer from several challenges and weaknesses which have been identified in policy reports and other literature such as long waiting times for appointments, lack of coverage for some health care and limited freedom of choice [2, 3]. VPHI is an instrument people can use to overcome some of these difficulties [4, 5].

The percentage of people in Portugal who reported having a VPHI increased from around 16–20% between 2014 and 2019. However, this increase does not translate into an increase in the percentage of the current health expenditure financed by this source. The share of current health expenditure supported by VPHI has been stable over time: it was 7.98% in 2014 and 7.64% in 2019 [6]. However, the average premium paid for a health insurance policy has increased over time [7].

These observations raise the simple question of what drives people to buy VPHI in Portugal. Therefore, the aim of this paper is to estimate the main drivers associated with the demand for VPHI and to compare the results between 2014 and 2019. To achieve this purpose, we use data from the National Health Survey of 2014 and 2019 to estimate a logistic regression for each wave, using the same independent variables to ensure comparability.

2. Overview of the health insurance market and on VPHI in Portugal

The health insurance market is characterised by the presence of asymmetric information [8, 9, 10, 11] which manifests itself in two principal paths: moral hazard and adverse selection. Moral hazard emerges after the insurance contract has been signed, whereby the insured individual tends to exhibit a propensity to use healthcare services to a greater extent than necessary, as the financial consequences are borne by the insurer. Adverse selection, on the other hand, occurs before the insurance contract is signed when the insurer lacks the comprehensive means to assess the individual’s risk profile. There are, therefore, two possible outcomes: the individual may belong to a low-health risk category, resulting in predominantly good health and minimal healthcare expenses, or the individual may belong to a high-risk category, requiring substantial healthcare expenditure. The existence of adverse selection poses a significant challenge to insurance companies, as they may inadvertently attract a larger proportion of high-risk individuals, with potentially adverse financial consequences. However, a favourable situation of advantageous or propitious selection [12] can be observed which confers benefits for the insurance companies. This phenomenon is often linked to the individual’s risk aversion, with those in good health showing a tendency towards risk-averse behaviour, leading them to choose VPHI coverage more often. Such a decision is likely to be motivated by their desire to secure additional protection against potential health-related expenses and uncertainties.

The demand for Voluntary Private Health Insurance (VPHI) has been extensively studied in academic literature, with two recent comprehensive reviews shedding light on this subject. While Outreville’s work [13] focuses on the general demand for insurance, Kiil’s [14] research focuses specifically on the demand for VPHI. Both contributions thoroughly explore the socio-demographic determinants that influence individuals’ decision to purchase VPHI policy. Such determinants encompass a range of factors, including gender, age, education, income, marital status, labour status, and other relevant characteristics. In general, there is a discernible correlation between the likelihood of individuals purchasing VPHI and certain factors such as higher income levels, greater educational attainment, age, employment status, urban residency, and immigrant status. However, the results pertaining to gender, family composition, and pensioner status are more heterogeneous and less unequivocal, with varying outcomes evident across different research studies and country-specific contexts [15, 16].

There are few studies on the demand for VPHI in Portugal. The oldest study was published in 2003 [17], the author found that some factors associated with buying VPHI such as being older, self-employed, living in urban areas, and receiving high income. Following this study, Tavares [18, 19, 20] has published some research work on the drivers related to holding VPHI policy in Portugal, focused on the relationship between health insurance demand and lifestyle decisions [18] and on the seniors’ segment [19, 20]. In this set of empirical work, demographic, socioeconomic, health status, and healthcare utilisation factors were used to explain paying for VPHI. There is a common conclusion across this set of studies which is the potential inequality of access to health care that may be created by holding VPHI in a health care system defined by NHS.

The contribution of the present analysis is two-fold. Firstly, it provides a study for a representative study of the Portuguese population, and secondly, it provides a comparative analysis between two years, 2014 and 2019. This comparison not only allows to draw conclusions on the evolution of the demand drivers for VPHI but also provides well-grounded information for policymakers aiming to reduce health inequalities.

3. Research design

3.1 Data and sample

We use data collected by the two waves of the National Health Survey: 2014 and 2019. These surveys are standardised and regulated at the European level (European Parliament and European Council Regulation no 1338/2008; European Commission Regulation no 2018/255). The wave in 2014 included 18,204 individuals while the wave in 2019 included 14,617 individuals.

3.2 Variables

3.2.1 Dependent variable

The dependent variable is obtained from the question about holding voluntary private health insurance (VPHI). This is a binary variable which takes value 1 if the respondent has VPHI and 0 otherwise.

3.2.2 Independent variables

Independent variables are described in Table 1 and may be grouped into socio-economic, insurance status, health status, and health care utilisation.

Group of variables	Independent variables	Description
Demographic	Male	Dummy variable. Takes value 1 is male, 0 otherwise
	Age	Median point of the correspondent age class.
Socio-economic	Education	The number of years of education completed.
	Income (Q1–Q5)	Set of dummy variables expressing the quantile of net monthly income per equivalent adult. The lowest income quantile corresponds to the first quantile, which is Q1. The reference category is Q5 corresponding to the highest income quantile.
	Urban	Dummy variable. Takes value 1 if the residence area is densely inhabited, and 0 otherwise.
	Rural	Dummy variable. Takes value 1 if residence area is sparsely inhabited, 0 otherwise.
	Moderate urban	Reference category.
	Marital status
	Single	Dummy variable. Takes value 1 if single; 0 otherwise.
	Married	Dummy variable. Takes value 1 if married; 0 otherwise.
	Divorced	Dummy variable. Takes value 1 if divorced; 0 otherwise.
	Widow	Reference category.
	Employee
	Self-employed
Insurance status (health sub-system)	ADSE	Dummy variable. Takes value 1 if covered by public servant health sub-system; 0 otherwise.
	SMAS	Reference category. Takes value 1 covered by private bank employee’s health sub-system; 0 otherwise.
Health status	SAH	Self-assessed health ranges from 1 to 5 levels, where 1 represents ‘very bad’ and 5 ‘very good’ health. The variable is taken as an approximation to a continuous variable.
	Chronic_disease	Dummy variable. Takes value 1 if suffers from at least one chronic disease; 0 otherwise
Health care use	Unmet_needs	Dummy variable. Takes value 1 if reported unmet health care needs due to financial constraints; 0 otherwise.
	Waiting_care	Dummy variable. Takes value 1 if waiting for a medical appointment or treatment beyond reasonable time; 0 otherwise.
	Flu_Vaccine	Dummy variable. Takes value 1 if vaccinated in the last 12 months; 0 otherwise.

Table 1.

Description of independent variables.

4. Quantitative analysis

Firstly, we perform a descriptive statistic. Secondly, we estimate a logistic regression to explain the holding voluntary private health insurance (VPHI) because (i) the dependent variable is binary and logistic regression is widely used in this case, (ii) it provides the estimation of odd ratios related to each independent variable which are easy to interpret, (iii) it allows for easy assessment on model fit, and finally (iv) we aim to have the set of independent variables jointly explaining VPHI, without excluding or selecting some of the independent variables.

Several diagnostic tests are performed, we begin by estimating variance inflation factors (VIF) to test multicollinearity. Then goodness-of-fit tests are undertaken using Pearson test, estimation of the area under ROC curve (receiver operating characteristic curve), and percentage of correctly classified cases. Robust standard errors are computed to correct for heteroscedasticity. The results are obtained using STATA 15 econometric software.

5. Results

5.1 Descriptive statistics

Table 2 presents the descriptive statistics for the independent variables used in the logistic regression. The majority of the individuals are women and the mean age increased between 2014 and 2019. A large percentage of people afford a small income and the level of formal education increased by about 4 years. There is a significant share of people benefiting from ADSE insurance coverage and the majority of people report not-so-good levels of health. Finally, there is a large percentage of people reporting unmet health needs, either from waiting or from financial barriers, and a substantial increase in the percentage of people reporting being shot against influenza.

Independent variables		2014	2019
Demographic	Age (years)	53.1	56.7
	Male (%)	43.6	43.4
Socioeconomic	Income Q1 (%)	22.1	18.3
	Income Q2 (%)	20.6	24.7
	Income Q3 (%)	19.8	20.7
	Income Q4 (%)	18.9	18.1
	Income Q5 (%)	18.6	18.2
	Education (years)	8.0	12.2
	Urban (%)	30.1	28.9
	Rural (%)	37.2	32.7
	Single (%)	24.6	23.9
	Married (%)	51.7	50.0
	Divorced (%)	9.4	10.0
	Self-employed (%)	7.0	6.8
	Employee (%)	29.8	31.7
	Unemployed (%)	11.4	6.9
Insurance status	ADSE (%)	13.3	13.7
	SAMS (%)	1.5	1.6
Health status	Chronic_disease (%)	61.2	57.8
	SAH [levels 1, 2 and 3]	55.5	56.9
Health care use	Waiting_care (%)	21.4	24.8
	Unmet_needs (%)	10.6	10.0
	Flu_vaccine (%)	18.6	44.3

Table 2.

Descriptive statistics.

5.2 Estimated results

The results obtained with the estimation of the logistic regression for having voluntary private health insurance (VPHI) are presented in Table 3, for survey wave 2014 and wave 2019, where bold font in the table points to differences between the survey waves.

	2014		2019
	OR	P > z	OR	P > z
Age	0.990	0.000	0.978	0.000
Male	1.028	0.549	0.810	0.000
Income Q1	0.188	0.000	0.252	0.000
Income Q2	0.285	0.000	0.251	0.000
Income Q3	0.415	0.000	0.373	0.000
Income Q4	0.553	0.000	0.566	0.000
Education	1.117	0.000	1.018	0.002
Urban	1.254	0.000	1.119	0.038
Rural	1.106	0.074	0.994	0.916
Single	1.265	0.050	1.186	0.117
Married	1.548	0.000	1.327	0.003
Divorced	1.660	0.000	1.449	0.001
Self-employed	2.525	0.000	1.652	0.000
Employee	1.716	0.000	1.657	0.000
Unemployed	0.852	0.082	0.799	0.039
ADSE	0.367	0.000	0.468	0.000
SAMS	0.651	0.005	0.740	0.091
Chronic_dis	1.098	0.075	0.980	0.706
SAH	1.173	0.000	1.251	0.000
Waiting_care	0.932	0.245	1.129	0.028
Unmet_needs	0.784	0.007	0.771	0.004
Flu_vaccine	1.007	0.928	1.147	0.006
_cons	0.075	0.000	0.494	0.005
N	18,153		13,973
Wald chi2(22)	2394.14		1632.34
Prob > chi2	0.000		0.000
Pseudo R2	0.180		0.131
Pearson chi2	14,441.99		11,850.15
Prob > chi2	0.985		0.258
VIF	1.75		1.64
Correctly classified (%)	84.34		80.30
Area under ROC curve	0.795		0.750
%people with VPHI	16.34		20.66

Table 3.

Logistic regression results.

Firstly, the preliminary test for multicollinearity shows that VIF values are compatible with the absence of multicollinearity in both logistic estimations as they are under the value 10. The post-estimation testing shows that there is a general good fit of the model in both regressions as shown by no statistical significance of Pearson value, by the large area under the ROC curve (about 0.79 and 0.75 for the first and second regression), and by the large percentage of cases correctly classified (about 84% and 80% to wave 2014 and wave 2019, correspondently).

Secondly, concerning the factors associated with holding a private health insurance policy. In general, socioeconomic factors influence this asset of a person; younger people, more educated, employed (both self-employed or employed), with higher incomes, and with better health status tend to drive people to buy a VPHI. There is no evidence that single people are more, or less, interested in buying VPHI, but both married and divorced people are likely to buy such insurance. People benefiting from the health sub-system of protection are less likely to have a VPHI and identically happens to bank employees in 2014, but it loses statistical significance in 2019. People reporting unmet health care needs, as expected, are less likely to have a VPHI and people reporting flu vaccine in the previous 12 months are more likely to report having a VPHI in 2019.

6. Discussion

The Portuguese health system is defined as a National Health Service, providing universal coverage for the population and for a very wide range of health services. Nevertheless, it faces some challenges which may lead people to demand VPHI. The aim of this study was to estimate the main drivers for buying VPHI in Portugal.

6.1 Key findings

The key findings of this study are threefold. First, the drivers for buying VPHI did not change significantly between 2014 and 2019. Secondly, people who benefit from ADSE and SAMS health insurance sub-systems are less likely to buy VPHI, despite the lack of confirmation in 2019 for people integrated into SAMS. Thirdly, and more importantly, people with unmet health care needs, due to financial barriers, are less likely to benefit from VPHI; people who report waiting too long for a medical appointment or treatment, and people who adopt a preventive behaviour for influenza illness by taking a vaccine are more likely to have a VPHI policy.

6.2 Interpretation of findings

Firstly, one major finding points to the association between unmet healthcare needs, due to financial barriers, and a lower likelihood of benefiting from VPHI, which was also found before [20, 21, 22]. This evidence raises questions about the equity access to healthcare services by people with lower incomes. In fact, income quintiles confirm that income plays a role in buying VPHI and people with higher incomes afford a VPHI.

Another major finding concerns the role that waiting for a medical consultation and treatment plays as a possible driver for buying private health insurance which became evident in 2019. It is well-known the waiting lists in the Portuguese NHS, both for consultation and treatments. Despite the relationship found does not allow for establishing a causal effect, one may suggest that there is in fact a causal relationship such that people look for VPHI as a possible solution to the excessive waiting times. The other direction of causality is less likely under the context of VPHI as a complement to the NHS. It could be that people holding a VPHI consider that waiting time for consultations and appointments within NHS is longer based on their expectations. A similar result has been found in Spain [23] where a reduction of the waiting times reduces the probability of buying VPHI.

One additional major finding is the positive relationship between people assuming preventive behaviour against flu disease and holding a VPHI policy. The flu vaccine is not compulsory in Portugal. It is voluntary but it is highly recommended and nearly free to people requesting at the primary health care units. The evidence of this relationship is only available for 2019, despite the smaller sample size. This relationship is interesting as it brings into discussion the asymmetric feature of the insurance market. On the one hand, it may be that insurance companies motivate the insures to be vaccinated against flu to minimise moral hazard [24] or there is advantageous selection and so it may be that people are more risk averse and cautious so that they get flu vaccine to minimise the risk of getting ill [25]. Another explanation is that it could be a result of NHS campaign to vaccinate seniors and other vulnerable groups [26] which could be inferred from the substantial increase in the percentage of people vaccinated against influenza disease in 2019.

Secondly, people benefiting from ADSE and SAMS health sub-systems are less probable to buy VPHI, despite the lack of confirmation in 2019 for people integrated into SAMS sub-system. This result is expected as health sub-systems function as a second layer of health protection. People benefiting from this level of protection must pay a percentage of their salary, so it is absent the incentive to pay even more for a third layer of protection provided by VPHI [19, 20].

Thirdly, there is no major change in the drivers for buying VPHI between 2014 and 2019. Demographic and socioeconomic drivers are identical for the two waves of the National Health Survey. As expected, older people are less likely to buy VPHI as found [19, 20]. This happens not only as a cream-skimming market strategy by insurance companies but also because primary care contractual goals are favourably biased towards older people [27]. Males are less likely to hold a VPHI, despite the opposite relationship found previously [18]. It may be difficult to explain such a result, however since buying VPHI implies a decrease in the real wages and forces shopping trade-offs, it might be that men are becoming less willing to give up on purchasing power.

Finally, other associated factors are higher incomes, and higher education and people in urban areas are more likely to buy VPHI as found earlier [17]. While unemployed people may struggle to buy a VPHI, employed people, either self-employed or employees, are motivated to buy a VPHI. Finally, despite the lack of evidence concerning people suffering from chronic diseases, findings show that better health status is related to holding a VPHI. This may result from selection strategies by the insurance companies to deal with adverse selection [9, 27] or it may be associated with advantageous selection of some people who prefer to have VPHI despite their good health status [12, 28].

6.3 Strengths and limitations

One limitation arises from the survey question about holding VPHI. It may be that some people confuse VPHI with health clubs or provider networks, where prices are lower than those of private healthcare providers. Unfortunately, there is no instrument to identify this situation. Another limitation is the impossibility of analysing causality or dynamic effects. The methods used here provide evidence of the correlation between VPHI and factors associated with the decision to take out health insurance. The main strength of the work is the comparison performed between two waves of the National Health Survey, both including a representative sample of the population.

6.4 Policy implications

The findings reached in this work continue to provide evidence of the increasing health inequalities and care inequities. The unfairness of the resulting health outcomes requires special attention from policymakers. Ensuring policies that contribute to the mitigation of health inequities is a primary concern nowadays [29]. Given the high share of out-of-pocket payments in Portugal and the different access to VPHI, well-designed policies are needed to improve access to health care for people with low incomes.

Acknowledgments

The authors acknowledge CEISUC/CIBB is funded by national funds through FCT—Foundation for Science and Technology, I.P., under the Multiannual Financing of R&D Units 2020–2023.

Conflict of interest

The author declares no conflict of interest.

Author funding

The author received no financial support for the research, authorship, and/or publication of this article.

Declaration

The author declares this work does not require any human/animal subjects to acquire ethical approval.