Introduction

Understanding and disclosing the public’s preferences in order to shape health policy is an essential task that has been shown to improve the efficiency and quality of care [1]. Reproducibly measuring these preferences is a complex job, but there are methods for attributing value to goods or services for which there is no real market, as is the case with healthcare provided in the public setting. Contingent valuation (CV) is a method well grounded in economic theory which assumes that individual preferences can be interpreted in the form of a utility function, where two states (initial and final) can be compared in terms of the changes in the level of utility. This method has advantages in the case of the valuation of healthcare as it makes it possible to include use and non-use values, i.e. the value derived from viewing the product as a consumer good and the values relating to the very existence of the service [2].

The value attributed to a good or service with the CV method can be studied from the perspective of willingness to pay (WTP) or to accept (WTA). WTP is the maximum price a person would be prepared to pay to receive a service, while WTA is the minimum compensation the same person would require to abandon the service. In principle, the two approaches should yield similar values as they are valuations of the same good. However, numerous experiments have shown that the values obtained by WTA are consistently higher than those expressed by WTP when valuing the same good [3–8] irrespective of the method used to evaluate them [9].

Large positive differences between estimated WTA and WTP measures of the value of a gain, or the value of a loss, tend to accentuate the ambiguity that can arise under the rationale of the decision in health policy making. The Kaldor—Hicks compensation test classifies the policy intervention as socially beneficial if the total WTP of those who gain exceeds the total WTA of those who are harmed. A large difference between WTA and WTP will make this criterion more difficult to satisfy, hence favouring the status quo over the new policy intervention [8]. This may have implications for the design of policies and public funding or disinvesting services. Considering threshold values as determined by WTP for health gains or WTA for health losses relative to an origin on the cost-effectiveness plane, representing the endowment of health services for patients with current practice, they consequently suggest higher values on the cost-effectiveness plane, with losses in health relative to gains in health and a threshold line kinked at the origin, with greater slope (higher values), with negative incremental effects and costs, than with positive incremental cost and effects [10].

Economic theory attributes disparities in valuing a good or service using WTA and WTP to the fact that less substitutable goods are more difficult to replace and therefore greater compensation is required for their loss [11,12]. The scarcity of economic resources has a similar effect inasmuch as a person can reach the limit of their ability to pay before they are ready to accept compensation. Others argue that the disparity between WTA and WTP for the same good may well be related to hypothetical bias. The less information there is about the good valued (ex-ante studies) and the higher the costs of information, the greater the bias and the greater the WTA/WTP gap [13]. However, perhaps the idea that has been most explored when trying to understand these differences is risk aversion and, more recently, the endowment effect in which people who are given a good ask for a considerably higher price to relinquish it than similar people who do not have the good would be prepared to pay to buy it. New theoretical contributions from the field of psychology, such as the one provided by “prospect theory” (which is in the origin of the so called “behavioural economics”) emphasise aversion to loss as an explanation for the disparity between WTP and WTA, which means that the utility of the loss is greater than the utility of an equivalent gain [14,15].

In the above framework, valuation of certain health services based on WTA and WTP may be useful for distributing resources and more fully understanding service users’ representation of value, together with the perceived lack of substitutability or loss aversion with respect to the service. The purpose of this paper is to examine differences in health system users’ valuation of a service they receive based on the WTP and WTA perspectives by analysing the impact of the users’ risk attitude and other personal variables and service characteristics on the disparities between these two perspectives.

Methods

Results

A total of 662 subjects were included. Subjects who refused to participate (n = 95) did not differ in age and gender from those included. There were no differences in refusals in centres of high or low income areas.

Eighty one of the subjects included (12.2%) expressed a WTP of €0. When asked about their WTA, 14 people (2.1%) did not answer and 13 subjects (2.0%) said 0. Just 2 of the subjects who expressed a WTP greater than 0 did not answer the question about WTA and another 2 gave an answer of 0. The median value expressed for WTP was €10 (interquartile range: €5-€20), and for WTA €20 (interquartile range: €10- €30). We obtained the WTA/WTP ratio for 570 subjects with a mean of 1.66 (CI 95%: 1.53–1.79). The median distribution of the WTA/WTP ratio was 1 (interquartile range: 1–2), and 193 of the 570 subjects had a WTA/WTP ratio above 1 (33.9%, CI 95%: 30.0–37.8%).

Table 1 shows the characteristics studied for subjects where we were able to estimate the WTA/WTP ratio. There were no significant differences in any of the variables studied with respect to all the subjects included in the study.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the people who expressed WTA and WTP.

https://doi.org/10.1371/journal.pone.0188969.t001

Table 2 shows the best equation to explain the WTA/WTP ratio when risk attitude was measured by means of the self-reported score. Expressed risk avoiding attitude was related to increases in the WTA/WTP ratio of 16% to 18% (models 1 and 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Factors explaining the WTA/WTP ratio, measuring self-reported risk score.

https://doi.org/10.1371/journal.pone.0188969.t002

Table 3 shows the model adjusted by the behaviour under risk (lottery games). Risk avoiding behaviour, measured by lotteries, was not associated with the ratio between WTP and WTA expressed for lotteries (models 3 and 4).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Factors explaining the WTA/WTP ratio, measuring risk aversion by lotteries.

https://doi.org/10.1371/journal.pone.0188969.t003

For lower levels of education (illiterate and no schooling), the WTA/WTP ratio increased by about 39–40% compared to primary education, while for people with higher education this ratio decreased by similar range (32–33%, models 1 and 3). When adjusted for social class, the most disadvantaged classes increased their expressed WTA/WTP ratio up to 90% compared to the better-off social groups (models 2 and 4). The nationality of the subject was associated with the WTA/WTP ratio, which was about 74% lower for Spaniards than for foreigners in models 1 and 3. This association lost statistical significance after adjusting for social group. Satisfaction with the service evaluated was consistently associated with the WTA/WTP ratio. The difference between the ends of the satisfaction scale (4 points) expressed mean differences in the WTA/WTP ratio of between 96% and 112% (models 1, 2, 3 and 4). HRQoL expressed in utilities was also significantly associated with the differences between expressed WTP and WTA. Patients who perceived better HRQoL expressed lower WTA/WTP ratios, with about 30% difference between the ends of the HRQoL measurement scale (models 2 and 4). Presenting chronic diseases, having been admitted to hospital and having taken out other insurance did not contribute explanatory power to the disparities in the WTA/WTP ratio. Family income showed collinearity with social group and educational level, so it was not included in the models.

Health centre characteristics were not related to the WTA/WTP ratio.

Models were repeated by attributing to interviewees who had offered zero WTP, the same WTP values of people of the same age and gender living in neighbourhoods with similar income and belonging to the same social group. Neither the explanatory factors nor the strength of the associations changed significantly.

Discussion

The perception of a health service such as the one evaluated here is different from the WTP and WTA perspective, even when this valuation is conducted with the same subjects and at the same time. Expressed WTA was greater than WTP by a mean 66% for the same service, although over half of the interviewees offered the same value for both questions. In addition to factors related to disparities between WTP and WTA which had been studied quite widely, such as socioeconomic status, age and satisfaction with the product, it seems that the attitude to risk may be associated with differences in the expression of these values.

The disparities found between WTA and WTP for the service studied are consistent with the literature. The reason for this disparities are usually grouped in economics, into three types of categories: those related to the elicitation technique [29], those related to the nature of the good studied [30] and those intrinsic to the subject who makes the valuation, where we would highlight risk avoiding [14]. Although some explanations of the WTP-WTA divergence operates within the welfare economics theory, others challenge this framework [31].

As for the type of technique, it seemed that WTA/WTP disparities were smaller with iterative deal questions than with payment card designs and also when the question was put to the same subjects [4] as in this case.

It had been hypothesised that the disparities between WTA and WTP were higher in services that could not be found in ordinary markets or, in the case of public goods, than when the experiences related to goods that were easy to find in the market [30]. Health care is a private good, in the economic sense, but it owns some "public good" characteristics, as its potential positive externalities (or negative if the need is not attended). In the field of health, the WTA/WTP ratio varies depending on the nature of the evaluated service, reaching a value of 4 when surgical techniques to improve hearing for children were assessed [5], and approaching 1 where the provision of informal care to the chronically ill was evaluated [32]. Also related to the type of good studied are the ability to find substitute goods and the cost of obtaining information about these exchanges, both of which are characteristics that might increase the WTA/WTP ratio [13]. In our case we did not find that subjects who used substitute goods (other insurance) expressed a lower WTA/WTP ratio, although it did happen with Spanish subjects compared to foreigners. Better understanding of the environment’s healthcare services can be attributed to subjects born in Spain as a group due to more prolonged contact with them.

There are two other features related to the good evaluated and which may influence the WTA/WTP ratio: the need for this good and satisfaction with its use. Those subjects with worse health expressed greater differences in the WTA/WTP ratio, as expected However, the factor with the most individual impact on this ratio was satisfaction. This may be related to a perceived lack of substitutability of the service or uncertainty about substitute goods, factors that we found widen the differences in the valuations from the different perspectives [12,33]

A third category of explanatory reasons for the disparities between WTA and WTP concern the subject who values the good, playing the role of demand. The literature suggested there was a greater gap between WTA and WTP in older people and in women [6,15,34]. However, although age and gender were introduced as an adjustment factor to enhance the explanatory power of the model, they were not statistically significant. There are other individual characteristics which welfare theory argues play a fundamental role in explaining the differences found in the valuation of a good or service by WTA and WTP. Perhaps the most researched is the income effect, by which an individual may not have sufficient resources to acquire a utility but does demand compensation for losing it. In our case, family income was not included in the best model, even though it was an explanatory variable for disparities between WTA and WTP, as there was clear collinearity with social group and educational level which meant not all of them could be in the model. Both belonging to a “high” social group and having a higher educational level point to the most advantaged social sectors, and both characteristics were associated with a lower WTA/WTP ratio. Some authors suggest that the differences between WTA and WTP are often too large for the income effect to be the only variable that explains them [35]; in this case we found that the characteristics indicating an individual’s socioeconomic status were significant in explaining the different valuations but were not fundamental.

The subject characteristics explaining the difference in WTP and WTA that have little to do with classical economic theory, and are closer to recent theoretical developments in psychology, are those that focus on resistance to loss or risk aversion and may vary for each good evaluated. Risk aversion has been found to be a variable that explains behaviour both in risk environments and in decisions which in principle do not involve risk [15]. This paper makes two new contributions in this respect. Firstly, it shows that the more “risk avoiding” profiles are those with a higher WTA/WTP ratio, i.e. the different valuations for gains and losses may be associated with the valuation of the good assessed, and it includes this fact in the specific area of a health-related service. Secondly, it points out that the way in which risk attitude is disclosed is not neutral, since results and conclusions may vary depending on the disclosure method selected. Risk aversion as an attitude was related to the WTP/WTA gap, but behaviour under risk was not. This could be due to a methodological issues but it is known that, attitudes and behaviors are correlated but not always aligned [36,37].

This paper may have some limitations. We cannot rule out the possibility of strategic bias, which could lead to much higher figures for WTA compared to WTP because the subject believes that in this way they can express a reluctance to change the service. This is particularly important given that when the study was conducted there was an open debate about whether or not to change health service copayment schemes.

Other type or limitation is related to the difficulty in assessing the significance of the rate of zero answers to the WTP question. In this case it was not higher than that found in other studies evaluating health services [19,38], and it does not appear to have any influence on the explanatory variables for the WTA/WTP ratio since the model with values “attributed” to the zeros had similar characteristics to the one presented. What may be a surprise is the high percentage of responses in which WTA is identical to WTP. We cannot be sure about the effect the design may have had on the results presented. Open-ended payment cards were used to learn WTA and WTP values, we referred to out-of-pocket payments and the same subjects were asked about the two values for the same service. The former tends to overestimate the WTA/WTP ratio, while the latter two do the reverse [4].

Finally, behaviours under risk were disclosed using hypothetical lotteries and not real lotteries with possible monetary gains and losses for the participants.

There are also several strengths in the design that could be underlined. Given that the description of the scenario may have implications for the expressed values [39], we chose as realistic a description as possible, making reference to a good immediately obtained. The payment card format shows advantages and inconveniences compared with other methods of eliciting values for WTP and WTA [40], but it is a commonly accepted tool. To minimize any possible bias that the range presented in the card might introduce, we asked a question in two phases, the first with a wide range and the second attempting to determine the real value stated. As recommended in the principal guides, the survey was a face to face interview, managed by a trained personnel [41], outside the health care process, immediately after the visit.

Although the present results cannot be extrapolated to other services received for other people in other contexts (i.e., other health services organizations or other cultural environments), it would appear that the disparities between WTA and WTP when evaluating a health-related service are consistent, and some of the assumptions drawn from outside classical economic theory need to be used in order to understand them. It is well known and documented that subjects behave differently with respect to gains compared to losses [42]. However, this point is not only relevant to learning what the value attributed to a particular service is, but in itself may also be an expression of the rejection of its loss. A debate exists about the opportunity to include loss aversion in health policy planning, since it affects resource allocation not only through the cost-effectiveness ratio but also through societal values concerning the distribution of health care resources [43]. It may even be that the preference for a certain good (or service) depends on whether it is going to be introduced or withdrawn, on when this will take place and on the quantity of wellbeing that it engenders [44]. This factor should be borne in mind when disinvestment approaches concerning clinical services and public health interventions should being planned [45].

Individual preferences should be aggregated in any way to shape social preferences. One traditional debate in the economic field is whether a Social Welfare Function (SWF), could be obtained. This function can be understood as a set of individual preferences on possible social states and associates a social preference to each possible configuration of individual preferences, and can then be used to rank economically feasible allocations of resources in terms of the social welfare they entail. The SWF should fulfil a list of logical norms in the process of aggregation of preferences (completeness, reflexivity, transitivity, universality or unrestricted domain condition, unanimity, Pareto efficiency and independence of irrelevant alternatives). As early as 1950, Kenneth J. Arrow demonstrated that this may be incompatible with democratic norms [46], hence some of the above conditions must be relaxed (i.e. universality or unrestricted domain status, unanimity or Pareto criterion). The discussion about how to aggregate individual preferences in a SWF, which will be the main tool for decision-making, continues today [2].

Cost-benefit analysis has been proposed as a valid method for analyzing changes in a SWF. While the SWF methodology relies on an interpersonally comparable utility function, CBA measures impacts on individuals in terms of money (individual willingness-to-pay or accept). A given outcome (a possible consequence of policy choice) is converted into a “vector,” of utility numbers, one for each person in the population. These vectors are then ranked by some rule—most simply a utilitarian function. Utilitarian function focuses on the maximisation of benefits, adding together the benefits to each individual regardless of their characteristics. If income and other determinants of individual utility are symmetrically distributed among the winners and losers of a given policy choice, CBA is a reasonable framework for a utilitarian SWF [47]. However, the WTA-WTP differences for the same service accentuate an existing problem. When a (health) policy is going to be developed, the fundamental problem is the need to compare benefits provided to some people with harms imposed on others [8]. Maybe it is more useful for policy makers to address the fundamental question, whether the benefits to those who gain from the policy justify the harms to those who lose, but keeping in mind that perceived utility is different under gain or losses perspective and that there is a profile of people who express greater differences between two perspectives.

Conclusion

The valuation of primary care health services varies from the gain or loss perspective assessed by WTP and WTA respectively. Both the need for and the difficulty in obtaining substitute goods, or satisfaction with the service, increase the gap between the two valuations as had been suggested previously. The income effect may also play a role in the disparity between WTA and WTP. Expressed risk avoiding could be associated with an increase in the WTA/WTP ratio, which is consistent with psychological theories that propose loss aversion as a key element in explaining this fact. Joint analysis of the factors that form part of the subject’s personality, along with other socioeconomic characteristics and factors that define the quality of the good or service, need to be built into analysis of the valuation of goods and services in healthcare.

So a profile of loss aversion to health care services could be identified and this would mean that these characteristics should be taken into account both in the design and implementation of new healthcare interventions services, as in the making decision for disinvestment.

Supporting information

Acknowledgments

Publication charges have been supported by of ISCIII and FEDER Funds, Grant Number RD12/0001/0012.