In two studies, we asked participants from the UK and USA their views about differences in life expectancies for distinct bases of categorization. There is already considerable study of the UK public’s view of health inequality (2, 10) and we wished to include respondents from another English-speaking country with high levels of inequality. In line with previous work,(2, 11, 30) we asked participants to indicate the priority with which these life expectancy differences should be addressed (Study 1) and agreement that something should be done about them (Study 2), as well as about the overall acceptability of life expectancy inequality (Studies 1 and 2). We present participants with one of six bases of categorization: genetics, lifestyle choices, income, education, neighbourhood and social class. The latter four categorizations were selected to capture a broad range of SES indicators employed in epidemiologic literatures of socioeconomic health inequalities. We also examined views on health inequalities by lifestyle choices following evidence that health behaviours explain a considerable proportion of social gradients in health and mortality,(32) as well as concerns that health inequality policy often focuses on tackling social gradients in lifestyle choices and behaviours at the expense of upstream interventions that might address socioeconomic inequalities themselves.(33) Whilst the genetic basis of socioeconomic health inequality is unclear and controversial,(34) we included this as a basis because people view genes as a key determinant of ill-health(35) and because we wanted to examine responses for a category likely to invoke biological explanations.
Studies 1 and 2 were conducted separately and sequentially but are reported together in the interests of brevity and in order to clarify the motivating logic for key changes between the two studies. Although primarily exploratory, we pre-registered both study designs and analysis protocols prior to data collection: https://osf.io/4cez8/?view_only=886245f44632487fbb152a5c47ba142b (Study 1) and https://osf.io/nbvc7/?view_only=6c8f0f9c6b17483ab5b4cdfc7559ee70 (Study 2). Divergence from planned analyses is acknowledged where relevant. Both studies received ethical approval from the ethics committee at [institution redacted] (/#11312/sub2/R(A)/2022/Dec/BLSSFAEC).
Participants (Study 1 and 2)
The data reported here are from 602 participants in Study 1 and 719 in Study 2. Samples in both studies were recruited through the UK and US Prolific panels (1: UK = 301, US = 301; 2: UK = 364, US = 355). Use of platforms like Prolific is widespread in experimental psychology. They can rapidly produce large datasets, are more socioeconomically diverse than student participant pools and have been shown to replicate the results of in-person studies.(36) A priori power analysis was precluded by the exploratory nature of the study design; nonetheless we sought to obtain a minimum of 100 participants in each of the categorization conditions.
Mean age was comparable in the two studies (1 = 38.35, SD = 13.76, range = 18–83; 2 = 39.93, SD = 13.89, 18–83), however, gender distribution was less balanced in Study 1 than 2. 61.96% (n = 373) of participants in Study 1 identified as female, 37.54% as male (n = 226) and the remainder (n = 3) as non-binary, compared to 48.95% (n = 352) identifying as female in the Study 2. The remainder of Study 2 respondents identified as male (n = 355; 49.37%), as genderfluid (n = 1), non-binary (n = 6) or preferred not to say (n = 6). Participants from Study 1 were not able to take part in Study 2. During recruitment, participants with missing data on more than two of the key study questions were dropped and replaced.
Design and Materials
Study 1
Participants were randomly allocated to one of six categorization bases (income, neighbourhood, education, social class, genetics, lifestyle choices). All participants were presented with an initial text stating that In Britain/the USA today, people in the longest-lived percentile live 15 years longer than those in the 1st percentile, accompanied by a graphical depiction of life expectancy by percentile (adapted from (37)). Participants were then informed that these differences in life expectancy are mainly due to differences in the corresponding categorization condition. Figure S1 presents a screenshot of how this information and ancillary questions appeared to participants in the social class condition.
Participants were next presented with four questions in the following order. First, they were asked how acceptable these differences in life expectancy are (on a 0 to 100 slider, where 0 = not at all acceptable and 100 = entirely acceptable). Second, they were asked how much of a priority they thought addressing these differences should be (where 0 = not at all a priority and 100 = a very high priority). The final two questions of this section asked how easy they thought it would be to address these differences in life expectancy due to the basis of inequality (where 0 = not at all easy and 100 = very easy) and how inevitable they thought these differences in life expectancy are (where 0 = not at all inevitable and 100 = entirely inevitable).
Participants were then asked to rate four causal explanations (biological, psychological, sociocultural and chance factors) for differences in life expectancy. Full description and analysis of these items can be found in Supplemental Materials. We do not report further on these responses here because aspects of the analysis indicated that participants had difficulty interpreting these items. For instance, the genetics category was rated higher for psychological than biological explanations. It was also the category rated lowest on chance factors, which was counter to our expectations. To help address these concerns, in Study 2 we reverted to the approach employed in the study that motivated the inclusion of these causal explanation items in the first place.(19) This meant that we asked participants to rate more specific causal explanations, i.e., hormones rather than superordinate categories (i.e., biological explanations) which may be harder to interpret. This is described in further detail below.
Demographic questions at the end of the survey were restricted to gender and age. Along with standard information on study details, withdrawal, contact and data storage, at the point of debrief we took care to ensure that participants were informed that it is not the case that differences in health and life expectancy are driven by any one factor and that in reality, health and life expectancy are influenced by many different determinants at the same time. Participants were invited to learn more at an international webpage on the determinants of health.
Study 2
This study was designed to replicate and address some limitations of Study 1. There were a number of changes to the design and materials. Firstly, we included a seventh “no-basis” condition, intended to serve as a neutral reference category. In this condition, we asked participants to make acceptability ratings of inequality where inequality was presented without any basis of categorization. This condition allows us to compare the influence of specific bases of categorization against a condition with no category.
We were also cognizant that in Study 2 we could not differentiate participants’ judgments about the basis of categorization itself (e.g., social class) from life expectancy differences related to that basis (e.g., life expectancy differences by social class). It may be, for example, that social class is viewed as a more intractable issue than life expectancy by social class (or vice versa), or that some participants interpreted the question as asking about the former whilst others answered about the latter. To determine whether this mattered, we manipulated whether participants were asked to rate the basis of categorization per se or life expectancy by that basis of categorization (we henceforth refer to this variable as explanation target). Table S3 presents an overview of the study design and the final sample size per cell.
The initial text and question on acceptability were as in Study 1. As it may have assessed ratings of the urgency of the issue as opposed to general support for intervention, and in order to more directly compare with previous studies,(2, 11) the question on how much of a priority intervening should be was replaced with a question asking about support for intervention: How much do you agree that something needs to be done to address these differences in life expectancy [by [category]]? on a scale from − 50 (strongly disagree) to + 50 (strongly agree). The response scale was changed from 0-100 to -50 to + 50, to help participants differentiate agreement vs. disagreement.
Participants were next asked four questions (in a randomized order) designed to gauge perceptions of inevitability, malleability and complexity. As these are relatively abstract concepts to rate and participants may have struggled with this in Study 1, we reverted to asking participants to agree or disagree with simpler statements.(19) Accordingly, we asked the extent to which participants agreed or disagreed (on a 100-point slider where − 50 = strongly disagree and 50 = strongly agree), that differences: can be easily changed; are inevitable; have a simple cause; and are complex. Participants in the categorization as explanation target condition were asked to make these ratings about the basis of categorization (e.g., differences in social class have a simple cause), whilst participants in the life expectancy by categorization condition were asked about life expectancy by the basis of categorization (e.g., differences in life expectancy by social class have a simple cause).
Subsequently, participants were asked to rate the contribution of 12 causal factors (in a randomized order) on a scale from 0 (not at all driven by) to 100 (strongly driven by). The response scale and phrasing of nine of these causal explanation items were adapted from Nettle and colleagues where ratings of explanations that relate to genetics, hormones and evolutionary advantage were found to cluster together (‘biological’ cluster), ratings of choice, motivation and psychological traits clustered together (‘psychological’ cluster), as did ratings of culture, social role, childhood experiences (‘sociocultural’ cluster).(19) Items were rephrased to fit the current paradigm (e.g., rate the extent to which differences are driven by factors related to childhood experiences). A further three items were added to capture features of what might be understood to be the case if something is driven by chance: factors that cannot be known, factors that cannot be controlled, factors that cannot be predicted. These were included to determine whether these differentiated from biological causes, that are often conflated with chance attributions(4) and because views on chance and luck are known to be determinative in people’s judgments about fair health outcomes.(38) Due to an error, all participants were presented with two identically-worded items relating to social roles (instead of the 13th pre-registered item) and so responses to these two identical items were averaged. We therefore collected data on 12 causes rather than the pre-registered 13.
On the final page, participants were asked analogous questions about the extent to which differences should be driven by each of the 12 causes. Henceforth, these items are referred to as measures of ideal causation in order to differentiate them from measures of perceived causation described above and reported below. Full reporting and analysis of ideal causation responses can be found in Supplemental Materials.