Abstract
Health economists have developed questionnaires to measure how much people care about health inequality that may be considered unfair (“equity”) relative to overall health (“efficiency”) and methods to analyze the data building on the social welfare function (SWF).1–3 The resulting estimates of health inequality aversion can be used in SWF-based frameworks such as distributional cost-effectiveness analyses to help decision makers evaluate tradeoffs between improving total health and reducing health inequality that may be considered unfair. 4 These questionnaires find that members of the public are highly averse to health inequality, implying that health gains to disadvantaged groups are worth many times more than gains to better off groups. One study of the members of the public in England, for example, estimated that the median respondent valued gains in life expectancy to the lowest social class 6.8 times more than gains to the highest social class. 3 In fact, a large proportion of respondents in previous studies—sometimes more than half—expressed extreme aversion to health inequality to the extent that they violate monotonicity.1,2
However, such findings are likely to be influenced by framing effects and other cognitive biases that are well known to psychologists.5–7 For instance, a number of studies have evaluated how preferences are influenced by the presentation of outcomes, such as relative versus absolute levels8,9 gains (such as cases detected) versus losses (such as cases missed), 10 and probability of life versus probability of death. 11 Moreover, recent research has found that both numerosity (i.e., units used in the choice experiments, such as days or years) and unitosity (i.e., respondents’ association of small units with small changes and large units with large changes) can influence responses. 12 However, such cognitive effects have not been thoroughly tested in the context of elicitation of health inequality aversion.
Results that are skewed by cognitive biases induced by the study design may not be sufficiently reliable to inform public policy making, as they may not generalize from the study setting to real-world policy settings. If the findings of academic work on health inequality aversion are to influence public policy decisions, then the direction and magnitude of potential cognitive biases need to be more thoroughly examined. Of the many possible cognitive biases we could have examined, we have targeted those that appear to be the most policy relevant in terms of assessing the generalizability of findings from study settings to real-world policy settings.
Aim
This study aimed to assess how far a standard questionnaire instrument for eliciting aversion to social inequality in health is vulnerable to large cognitive effects that make a substantial difference to the estimated degree of health inequality aversion. The four potential effects we examined were between 1) realistic small health inequality reductions compared with unrealistically large health inequality reductions, 2) population-level compared with individual-level descriptions of health inequality reductions, 3) concrete compared with abstract intervention scenarios, and 4) online compared with face-to-face mode of administration. Only the second of these is a framing effect in the classic sense of using different ways of describing exactly the same decision problem. However, all four can be thought of as cognitive effects, in the broad sense that they relate to issues of cognitive psychology and information processing, as explained below.
The 4 Hypothesized Cognitive Effects
Small versus unrealistically large health inequality reductions: Empirical studies typically use hypothetical scenarios that involve unrealistically large changes in health, such as hypothetical government programs that will extend average life expectancy by several years and reduce inequality by a few years.1–3 However, general population average health gains of this size are unrealistic in the short run and unlikely to be achievable even with a massive and sustained “once in a generation” program of cross-government social, political, and economic reform. In practice, the public policy alternatives actually considered by social decision makers deliver much smaller average health benefits to the population. For example, a case study of different ways of promoting uptake of the National Health Service Bowel Cancer Screening Programme among disadvantaged groups estimated incremental gains in general population average life expectancy of only a few hours. 4 This is because most people do not have bowel cancer and will gain nothing from the screening program, but a few people will gain many years of life. It is not known how far findings from studies of unrealistically large gains are applicable to more realistic settings involving small health gains. We hypothesize that using small health gains may substantially reduce the degree of inequality aversion. For instance, Olsen 13 argued that a “minimum threshold quantity” of health gains may exist beyond which individuals’ equity preferences take effect. Below this threshold, individuals may concentrate on the gains for some few. This was empirically demonstrated by Rodríguez-Míguez and Pinto-Prades, 14 who found that individuals prefer to concentrate on total health gains for smaller individual gains and express inequality aversion for larger individual gains. This may be because health inequality is a less familiar concept than total health and is perceived at a less fine-grained level of detail. So respondents may see small reductions in health inequality as worthless while still seeing small total health gains as worthwhile.
Population-level versus individual-level descriptions of health inequality reductions: Studies typically present health benefits in terms of the average change to individuals (e.g., 2 y per person). However, health benefits to a population can also be expressed in terms of total gains to the group (e.g., 2 000 000 person-years across a million people). We hypothesize that when average benefits are small (e.g., a few hours per person), framing the same health benefits in terms of population totals may lead to larger inequality aversion than using average health benefits per person. A number of factors may influence preferences in this context. For instance, gains expressed in larger units (i.e., years) compared with smaller units (i.e., hours) 12 may incline respondents to prefer population-level scenarios. Respondents may not consider a small health gain in hours as worthwhile compared with a gain in years. Also, the perspective of health benefits (i.e., individual or population level) may result in different value judgments: this was demonstrated by Gyrd-Hansen and Kristiansen, 15 who found that the spread of health gains was more pronounced with the societal (or population-level) perspective. Based on this, we hypothesize that using population totals may make the reductions in health inequality seem larger and more worthwhile.
Concrete versus abstract intervention scenarios: Abstract scenarios are typically used, first, so that respondents do not bring their own unobserved cognitive “baggage” to the interpretation of the scenarios and, second, so that findings can be applied to multiple policy contexts. Arguably, however, respondents will always fill in the missing gaps in the scenario, and abstract scenarios may be more susceptible to this problem than more tightly described concrete scenarios. Furthermore, it is not known how far people’s abstract views about health justice are transferable to their context-specific views. When more concrete scenarios are given, and people are encouraged to think more realistically rather than abstractly, this may have an effect (in either direction) on the level of inequality aversion they support.
Online versus face-to-face mode of administration: Typical surveys of inequality aversion have been conducted face to face, which is time-consuming and costly. Increasingly, there is an interest in using online surveys to elicit social values of health,16–21 partly because of the high speed, convenience, and low cost of conducting the surveys although also because face-to-face administration may suffer from socially desirability bias. 22 There is a growing literature investigating differences between face-to-face and online survey responses, both in surveys of population opinion and those involving tradeoff exercises. For instance, in a randomized study involving person tradeoff value judgments of health states, Damschroder and others 23 found that tradeoff values did not differ between computer-based and face-to-face elicitations. Similar findings were observed by Mulhern and others 24 comparing computer-assisted personal interview versus face-to-face interviews. However, Norman and others 25 found that value judgments of EQ-5D health states differed by modes of administration, with the computer-based group choosing more extreme responses and having larger standard deviations compared with the face-to-face group. Some of these differences in valuation of health states may be explained by difference in sample characteristics, level of effort or commitment to providing a considered response, the level of support available to comprehend the task, and by social desirability bias. 26 Despite this growing literature, there is a dearth of studies investigating the influence of mode of administration on value judgments of health inequality aversion. We cannot determine a priori which mode of administration gives more appropriate results for use in policy making. On one hand, an online private environment is closer to the ballot box where citizens cast their secret vote without having to justify their choice. On the other hand, online surveys may be less likely to reflect considered opinions, since questions on social values require careful deliberation and are not topics that people are familiar with. Therefore, understanding the direction and magnitude of potential differences in findings between the 2 modes of administration is of interest. We hypothesize that using online surveys may lead to smaller inequality aversion relative to face-to-face administration.
Methods
Questionnaire
The basic questionnaire instrument used in this study (see Supplemental Appendix 2) is adapted from Abasolo and Tsuchiya 1 and Dolan and Tsuchiya. 3 It starts by presenting background information about the current level of inequality in health between “the richest fifth” of people in England, who on average live 74 y in full health (i.e., quality-adjusted life expectancy [QALE]) and “the poorest fifth,” who on average live 62 y in full health. 28 Both groups are made up of about 10 million individuals.
The respondent is then presented with the choice experiment, which consists of 4 questions (Q1–Q4), each with a different scenario or presentation (see below). Under each question (Q1–Q4), there are 5 pairs of hypothetical health programs, A and B. Hence, by the end of the choice experiment, each respondent completes 20 paired choices (i.e., 4 questions [Q1–Q4] multiplied by 5 pairs of health programs in each question).
At each pair, respondents are asked to choose between program A and B or indicate indifference. These programs show the health gains received by the richest and the poorest fifth of the population and the total health gain for the 2 groups. It is assumed that the remaining three-fifths of the population are not affected by any of the hypothetical health programs. In the first pair in each question, programs A and B produce the same amount of total health benefit across the groups but benefit the 2 groups differently: program B offers a reduction in health inequality compared with program A. In subsequent pairs, program A remains the same, but the health gain to the worse-off group in program B becomes smaller and smaller, and so both the reduction in health inequality and the gain in total health in program B become smaller and smaller. Hence, program B always reduces inequality but offers less total health, except for the first pair, for which both programs produce the same total health. The degree of inequality aversion is captured by observing the point at which a respondent is indifferent between the 2 programs. The format of paired choices remains the same across the 4 questions (Q1–Q4); however, the scenario or presentation of the question changes (see below).
Participants had to respond to all choice pairs to complete each task, irrespective of whether they were pro-rich or strict egalitarians; that is, there was no quicker route to complete the questionnaire by taking one or the other position, and there was no exit option available.
The first question (Q1) is the “large-average” presentation, which corresponds to the format used in previous studies. In this question, the first pair presents program A, giving a 7-y gain in life in full health to the richest fifth and a 3-y gain to the poorest fifth, and program B, giving a 3-y gain to the richest fifth and a 7-y gain to the poorest fifth. In the subsequent 4 pairs, the health gains in program A and the health gain to the richest in program B are fixed, while the health gain to the poorest in program B decreases gradually from 6 y to 3 y. Supplemental Appendix 2 reproduces all questions in full.
All subsequent questions use the same background inequality across the richest fifth and the poorest fifth and the same ratios of health gains (but scaled down proportionately). In the second question, the health gains are small and measured in hours per head. In the first pair, program A gives a 7-h gain in life in full health to the richest fifth and a 3-h gain to the poorest fifth, while program B gives a 3-h gain to the richest fifth and a 7-h gain to the poorest fifth. The gain for the poorest group under program B at the fifth pair is 3 h. Thus, this represents the “small-average” presentation.
The third question represents the small-population presentation. In the first pair, program A gives a 7000 person-year gain in life in full health to the richest fifth as a group (consisting of 10 million people) and a 3000 person-year gain to the poorest fifth as a group, while program B gives a 3000 person-year gain to the richest fifth and a 7000 person-year gain to the poorest fifth. The gain for the poorest group under program B at the fifth pair is 3000 person-years. Note that, since each quintile is assumed to consist of 10 million individuals, 7000 person-years amounts to 0.0007 y per head, which is equivalent to 6.132 h per head. We use 7 h in the second question instead of 6.132 h to maintain the 7:3 ratio of benefits across the richest and the poorest.
The fourth question (used only in face-to-face mode of administration) introduces a more concrete presentation to the health inequality and the health program, using a topic taken from Asaria and others. 4 A presentation is given on the different take-up rates of bowel cancer screening by income groups, followed by a description of 2 health programs on reminders to participate in screening: one that sends impersonal reminders to all eligible individuals (benefitting the richest fifth more) and another that sends personalized and general practitioner–endorsed reminder letters to individuals in deprived areas who have a lower take-up rate (benefitting the poorest fifth more). Besides the concrete context, this question is exactly the same as the third question.
Before completing these 4 questions, respondents completed a set of questions on attitudes to the welfare state and income redistribution. 28
Data Collection
There were 2 samples. One was a “discussion” sample, in which members of the public were invited to participate in a citizens’ panel event involving presentations by facilitators to introduce the questionnaire, facilitated discussions in groups of 5 or 6, individual completion of the questionnaire, sharing the responses within the group, and opportunities to change the questionnaire responses. Two citizens’ panel events were held with 2 different subsamples of participants: a 5-h event including lunch, on Saturday, 21 September 2013 (n = 29), in the City of York and a 3-h event with the same format and question order but excluding lunch and postlunch tasks not reported in this article, on Saturday, 22 February 2014 (n = 23), at the University of York. Payments of £70 and £30 were offered per participant in the first and second events, respectively.
The participants for the citizens’ panel were recruited through 1) advertisements in a monthly free local magazine (Your Local Link) distributed to all homes across York (July and August 2013) and 2) 810 leaflets distributed door to door in 10 of the most deprived streets in York. A quota was set so that each of 8 age/sex groups had a capacity for 3 to 4 participants, including 1 from a postcode with higher deprivation. Occupation information was also collected at the screening stage. Those with university academic/research jobs were excluded because they may have had previous training or exposure to handling similar tasks involving tradeoffs between competing social values and therefore may not have the same cognitive biases as the general public.
The second sample was an “online” sample that included 83 respondents. The first 3 questions above (large-average, small-average, small-population) were posted online (hosted by Smart-Survey). The survey was publicized on social media, the York Local Link magazine above, and the Web site of the Centre for Health Economics at the University of York. Respondents could complete the survey anonymously or leave their contact details. No remuneration was offered for taking part. To make the discussion and online groups comparable, we again excluded those with university academic/research jobs from the online sample using information on respondent occupation.
Research ethics approval for the study was obtained from the University of York Health Sciences Research Governance Committee.
Analysis
Prior to pooling across the 2 discussion group samples, we compared their results against each other and found no differences in the basic pattern of findings in terms of the level of inequality aversion.
Each of the main questions allowed us to distinguish 5 different types of value judgment, which correspond to 5 different principles of health justice (see Table 1). At each pair, respondents have 3 choices: program A (A), program B (B), or indifference (=). The pro-rich (AAAAA) always choose program A, while the health maximizers (=AAAA) are indifferent in the first pair but choose A subsequently. Collectively, we label these first 2 types as nonegalitarian.
Types of Value Judgment about Health Justice
Our third type of preference is the “trader” or “weighted prioritarian” (BXXXA), who chooses B in the first pair then switches to A at some point (indicated by the XXX in the middle, see below). The term weighted prioritarian means people who give priority to the worse off but not exclusively. Hence, they will not violate monotonicity (any increase in individual health will result in an increase in social welfare, other things being equal). Strictly speaking, a respondent who switches to program A only in the final pair (BBBBA) might be “leximin” rather than “weighted prioritarian”; that is, they give almost exclusive priority to the health of the worse off group but are willing to use a second principle such as health maximization as a tie-breaker.
The next category of respondents is maximin (BBBB=), who give fully exclusive priority to the health of the worst off by choosing program B in the first 4 pairs but become indifferent in the final pair. This preference can be represented by the limit of a standard SWF as the inequality aversion parameter tends to infinity. Finally, we label respondents who prefer program B in the final pair as strict egalitarians (BBBBB). This preference violates monotonicity and so cannot be represented using standard monotonicity-respecting SWFs. Collectively, we label these last 2 types as strong egalitarians.
Within the weighted prioritarian type (BXXXA), we can distinguish 7 distinct response categories by breaking up the XXX in the middle. The first of these (BAAAA) represents a tradeoff point of 6.5 (since the respondent switches at some point between 7 and 6 units of health benefit). The second subcategory (B=AAAA) represents a tradeoff point of 6 (since the respondent is indifferent at 6 units of health benefit) and so on down in half units to the seventh subcategory (BBBBA), which represents a tradeoff point of 3.5. Similarly, the pro-rich and health maximization categories represent tradeoff points of >7 and 7, respectively, and the maximin and strict egalitarian categories represent tradeoff points of 3 and <3, respectively. We thus obtain 11 separate response categories, which can be ranked in order from the least egalitarian (>7) to the most egalitarian (<3).
The 5 value judgments discussed above can be represented as iso-welfare curves using the Atkinson Index (Figure 1). The x-axis and y-axis show the QALE in the richest and poorest fifths, respectively. The initial QALE before the program is 62 y and 74 y for the richest and poorest fifth, respectively, which would increase to 65 and 81 y, respectively, if program A was implemented. The health distributions resulting from different value judgments are represented as diamonds in Figure 1, while preference functions are represented as iso-curves.
Iso-welfare curves representing response categories.
These response categories can be converted into health inequality aversion parameter estimates by fitting an SWF through the 2 outcomes where the respondent is indifferent and then solving for the inequality aversion parameter. Our aim in this article is to perform methodological tests of reliability rather than to estimate a health inequality aversion parameter for policy purposes. However, we compute implied weights to give readers insight into the magnitude of the inequality aversion parameters and to help those who wish to compare our findings with those of other studies. Supplemental Appendix 1 includes a lookup table along with the calculations underpinning the conversion process.
To examine 1) small versus unrealistically large health inequality reductions, the first (large-individual) and second (small-individual) questions are used, within each of the 2 samples (citizens’ panel and online). For 2) population-level versus individual-level descriptions of health inequality reductions, the second (small-individual) and third (small-population) questions are used, within each sample (citizens’ panel and online). To explore 3) concrete versus abstract intervention scenarios, the third (small-population) and fourth (concrete) questions are used, within the citizens’ panel sample. Finally, to examine 4) online versus discussion group mode of administration, the first, second, and third questions from the citizens’ panel sample are compared with the corresponding questions in the online sample. In each case, the cumulative distribution across the 11 ordered response categories (from less to more egalitarian) are compared using the Wilcoxon rank-sum test. In addition, the proportions of nonegalitarian (pro-rich and health maximizing) and strong egalitarian (maximin and strict egalitarian) responses are compared using the chi-square test.
Results
The Sample
Table 2 shows that the first 2 samples had similar age and gender characteristics. Overall, the discussion sample was slightly more affluent than the other 2 samples. Respondents in both of the online groups were slightly more likely to have egalitarian social attitudes than respondents in the discussion group.
Descriptive Statistics of the Discussion Group and Online Survey Respondents
1 suggests most egalitarian and 5 suggests most nonegalitarian.
Distribution of Switching Point by Question and by Sample
Figure 2 shows the distribution of responses across the five principles of health justice, inferred from the switching points by question and by sample.
Inferred principles of health justice by question and sample design. Pro-rich (AAAAA); health maximiser (=AAAA); weighted prioritarian (BXXXA); Maximiner (BBBB=); Strict egalitarian (BBBBB). The vertical line indicates the location of the median respondent.
Each stacked bar indicates the proportion of responses ranging from pro-rich on the left end to strict egalitarian on the right end. As can be seen, in the first 3 questions, the median respondent is “strong egalitarian” (i.e., “maximin” or “strict egalitarian”) in the discussion sample but is always “weighted prioritarian” in the online sample. However, the median respondent in the discussion group sample shifts to “weighted prioritarian” in the concrete question (this question was not used on the online sample).
Table 3 presents the tradeoff points of the median respondent and the implied equity weight to the poorest fifth compared with the richest fifth. In general, the discussion group had lower point estimates of median tradeoff points; however, the confidence intervals (CIs) of the tradeoff points in the discussion and online groups always included the strong egalitarian tradeoff point (3.0 or <3), and therefore the CIs for the implied weights were undefined.
Median Response in Different Groups, with 95% Confidence Interval in Brackets
Table 4 shows the results of statistical tests for the hypotheses. We adopt a conservative approach throughout and always report 2-sided tests. We use labels D and O for discussion and online groups, respectively. Hence, D1 represents Q1 in the discussion group, and so forth.
Nonparametric Statistical Tests: All Respondents
Strong egalitarians = maximiner or strict egalitarian; Nonegalitarians = pro-rich or health maximizers. P values provided for test of equality of proportions. Negative sign indicates less egalitarian (or nonegalitarian in the third column) in the first group compared with the second group and vice versa.
Wilcoxon rank-sum test was used for unmatched data, and signed-rank test was used for matched data. Null hypothesis: 2 samples are from populations with the same distribution.
D = discussion group; O = online group; 1 = large-average; 2 = small-average; 3 = small-population; 4 = concrete.
Our first hypothesis was that small gains (small-average, Q2; small-population, Q3) would yield less egalitarian responses than unrealistically large gains (Q1). We found that the large-average and small-population questions gave similar results within both the discussion (D1–D3) and online samples (O1–O3). On the other hand, responses were less egalitarian in the small-average question compared with the large-average question. However, this difference was not statistically significant in the online sample (O1–O2) and only just reached statistical significance in 1 of the 3 tests within the discussion group sample (D1–D2). Furthermore, the size of the difference was never large in any of the large versus small gains comparisons: the largest difference was a 9.2 percentage point difference in the proportion expressing strong egalitarian views (D1–D2). Overall, therefore, there was no clear evidence of a substantial and systematic effect.
Our second hypothesis was that responses would be more egalitarian under population-level (Q2) rather than individual-level (Q3) descriptions of the small gains question. The pattern of responses was in this direction within the discussion group (D2–D3) but did not reach statistical significance in any of the tests. Furthermore, there was no such pattern of responses within the online sample (O2–O3). Again, therefore, there was no evidence of an effect.
Our third hypothesis was that responses might differ (in either direction) between concrete (D4) and abstract (D3) scenarios. In all three tests, we found that discussion group sample responses were significantly more egalitarian under the abstract rather than concrete scenario, providing clear support for the third hypothesis.
Finally, our fourth hypothesis was that discussion group respondents (D) would be more egalitarian than online respondents (O). This was indeed the general pattern of responses, with a substantially larger proportion of the discussion group sample (between 11 and 21 percentage points) expressing strong egalitarian responses in all 3 questions. However, only the small-population question (D3–O3) reached statistical significance. So this provides some, albeit weak, support for our fourth hypothesis.
Figure 3 gives a visual representation of the rank-sum test, using the example of the large-average question comparing the discussion versus online samples (D1 v. O1). Along the horizontal axis are the 11 ranked points at which respondents can be indifferent between the 2 programs. The vertical axis represents the cumulative proportion of respondents who have switched to the less egalitarian program A by that point. The stronger the inequality aversion, the lower the cumulative curve. In Figure 3, the cumulative curves are similar up to 4, but then the online group rises more rapidly, showing a smaller proportion of respondents giving the “strong egalitarian” responses of 3 (“maximin”) and <3 (“strict egalitarian”).
Cumulative distribution of tradeoff points for the large-average question, by mode of administration (D1 v. O1). The trade off point represents the point at which the respondent would be indifferent between the less egalitarian programme A, in terms of the gain in years of life in full health to the poorest fifth in Programme B. In terms of our five principles of health justice, < 3 is “strict egalitarian,” 3 is “maximin,” 3.5 through to 6.5 are “weighted prioritarian,” 7 is “health maximisation” and 7+ is “pro-rich.”
Figure 4 shows a cumulative rank comparison within the discussion group sample, comparing abstract and concrete versions of the small-population presentation (D3 v. D4). This time, the curves start to diverge early on, with a higher proportion of <3 (pro-rich) and 3 (health maximizer) responses under the concrete question frame.
Cumulative distribution of tradeoff points for the small-population versus concrete question for the discussion group (D3 v. D4).
Discussion
A number of studies have attempted to quantify societal value judgments about equity, including value judgments about severity and burden of illness, end of life, and health inequality aversion.3,16,18,19 While these studies elicit public views, most do not evaluate the effect of different scenario presentations and mode of administration that may influence the final results. This article contributes toward filling this gap in the context of health inequality aversion by performing experimental tests contrasting 1) small instead of unrealistically large health inequality reductions, 2) population-level instead of individual-level descriptions of health inequality reductions, 3) concrete instead of abstract intervention scenarios, and 4) online instead of face-to-face mode of administration.
We find no clear evidence of systematic and substantial effects of 1) small versus unrealistically large health inequality reduction scenarios or 2) population-level descriptions. However, we do find clear evidence of 3) an inequality aversion-reducing concrete scenario effect and weak evidence of 4) an inequality aversion-reducing online mode of administration effect. Within our discussion group sample, the proportion of nonegalitarians rose substantially and significantly when using a concrete rather than abstract scenario (by 12 percentage points), and the proportion expressing extreme inequality aversion fell substantially and significantly (by 19 percentage points). Finally, respondents were substantially more likely to give strict egalitarian responses in our discussion group sample compared with our online sample (by between 11 and 21 percentage points), although this pattern of findings reached statistical significance only for 1 of the 3 questions. So we conclude cautiously that there is only weak evidence of an inequality aversion-reducing online mode of administration effect.
Despite these 3 effects, however, median responses always implied substantial health inequality aversion, and the implied weight to health gains in the worst off group never fell below 2.57 in any of our experimental conditions or subgroups.
One possible explanation for the concrete scenario effect is that the intervention in question (a reminder letter to promote uptake of bowel cancer screening) was an “agentic” intervention to promote individual health behavior change, rather than a “structural” intervention to alter the social determinants of health. Agentic interventions may make concepts of individual responsibility for health behavior and outcomes more salient in respondents’ minds, thus reducing aversion to health inequality. If so, abstract scenarios may tend to produce higher estimates of public concern for reducing health inequality than concrete scenarios based on agentic interventions.
Finally, the mode of administration effect may suggest a social desirability bias whereby face-to-face administration tends to elicit pro-egalitarian “Sunday best” responses. On the other hand, the (slightly) less egalitarian social attitudes reported by the discussion group sample suggests that this may be because face-to-face discussion provides a better opportunity for careful deliberation. Either way, it is important that decision makers are aware of the nature and magnitude of potential biases, so that they can appropriately assess and interpret evidence about public views.
One of the considerations in inequality aversion surveys is the choice of terminology used to describe the socioeconomic groups being compared. Our survey describes these groups as the richest and poorest fifth. This is both technically correct and easy to understand for the participants. Other terms such as most/least deprived and most advantaged/disadvantaged have been used in the literature. However, although the choice of terminology may influence the point of indifference within a scenario description, it is unlikely that the effects found in our study between scenario descriptions and between modes of administration may be due to the choice of this terminology.
We note that a number of valuation tools exist to elicit equity-efficiency tradeoffs 29 ; these may or may not provide different estimates of inequality aversion. However, the cognitive effects identified in our study are also likely to be relevant to other valuation tools. A potential weakness of the questionnaire instrument used in this study is that the visual aids used focus on health gains and not the final distributions of health. In this respect, this study may be criticized for favoring gain egalitarianism over outcome egalitarianism. 30 However, since identical framing with a fixed ratio of gains has been maintained across all questions, we expect this potential effect to cancel out in the comparisons across pairs of questions that we carried out. Another potential criticism of the study may be that respondents may favor the first (or the left-hand side) option over the second alternative in pairwise comparisons (commonly known as “response order effect”); similarly, the titration sequence in which questions are presented (compared with random-order presentation) may influence choice behavior. However, Abasolo and Tsuchiya 2 specifically evaluated both effects and found no evidence of significant difference due to response order or titration sequence.
Another potential criticism is that the order of questions was not randomized, which may influence value judgments because of a learning effect. However, we did not find a consistent pattern of responses across questions to suggest such effect. Finally, we acknowledge that the sample size of this study is relatively modest although large enough to detect at least 2 statistically significant effects. As a suggestion for future research in this area, we recommend that the biases considered in this study should be investigated in larger samples.
Another consideration in this study is the generalizability of our findings from a small sample to the general population. We think that although the absolute level of inequality aversion may vary between population groups, the cognitive effects found in this study are likely to hold true in other settings. However, further research is needed to investigate these biases in larger and more diverse samples.
Overall, however, the basic finding of substantial health inequality aversion is surprisingly robust. There was no substantial or clearly significant effect of using small rather than unrealistically large health gains or using population-level rather than individual-level presentations. And although a concrete scenario and online mode of administration reduced the proportion of respondents expressing infinite inequality aversion, median responses in all the variations explored still displayed a substantial degree of health inequality aversion.
Footnotes
This research work was carried out at the Centre for Health Economics, University of York. This work was presented at Health Economics Study Group (HESG) meeting in Sheffield, 9–10 January 2014.
Financial support for this study was provided by the C2D2 program, an interdisciplinary initiative supported by the University of York and the Wellcome Trust. The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report.
The following authors are employed by University of York, who partly funded this study: Shehzad Ali, Miqdad Asaria, and Richard Cookson.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
