Opinion Trends on Moral Issues in the United States and the United Kingdom Explained by the Applicability of Generally Accepted Arguments

Abstract

According to a recent theory, public opinion on moral issues will trend toward opinions that are justified by generally acceptable kinds of arguments. Both in the United States and the United Kingdom, generally acceptable kinds of arguments include concerns about harm, violence, fairness, and liberty. The theory therefore predicts that these countries will exhibit similar opinion trends on moral issues. We test this prediction using polling data on 98 issues in the United States and 108 issues in the United Kingdom, with an overlap of 27 issues on which opinions have been polled in both countries. We find that moral opinion trends are well predicted by the applicability of generally acceptable kinds of arguments and that there is a remarkable similarity between opinion trends in the two countries. These findings lend support to the theory that generally acceptable kinds of arguments play a key role in driving opinion dynamics.

A wide range of moral issues are debated in current society. In discussions of these issues, a variety of specific arguments may be used. Moral psychologists have found it useful to distill these arguments into just a few basic kinds: harm, fairness, authority, ingroup loyalty, and purity (Haidt & Graham, 2007). In recent research, this list was extended by a few additional kinds of arguments: violence, liberty, and government overreach (Vartanova et al., 2021). All kinds of arguments may not be equally effective, however. Extensive studies have asked people to rate how relevant different kinds of arguments are for their moral judgments (Graham et al., 2009, 2011). These studies, complemented by a new study included in this article, find that certain kinds of arguments (harm, violence, fairness, and liberty) are rated as relevant both by liberals and conservatives, while other kinds (authority, ingroup loyalty, purity, and government overreach) are generally rated as less relevant, and especially by liberals. We shall refer to the former kinds of arguments as “generally acceptable” and to the latter as “limitedly acceptable.”

When people are exposed to an argument of a kind that is not relevant to their moral judgments, they should not be influenced. The average person should therefore be less likely to be swayed by limitedly acceptable kinds of arguments than by the generally acceptable ones (Eriksson & Strimling, 2015). This has implications for how public opinion shifts over time on a specific issue, because different kinds of arguments are used to justify different specific opinions. This has been demonstrated in surveys of U.S. and U.K. samples who were presented with a wide range of moral issues and asked which kinds of arguments are used to justify different opinions on each issue (Vartanova et al., 2021). To illustrate, consider the issue of strict gender roles: Is it right that women should take care of the home while men achieve outside it? People who argue against strict gender roles tend to justify this opinion by fairness, which is a generally acceptable kind of argument. By contrast, arguments in favor of strict gender roles tend to be of limitedly acceptable kinds (authority and purity). Arguments against strict gender roles should therefore on average be more effective than arguments in favor of strict gender roles. This reasoning provides an explanation for why support for strict gender roles shows a negative trend over time.

Gender roles is just one of many possible issues to which the theory should apply. In a study of a wide range of moral issues in the United States, Strimling et al. (2019) measured which kinds of arguments are used to argue for different opinions on each issue. The researchers additionally estimated opinion trends from 40 years of opinion polls. Consistent with the theory, those opinions that have become increasingly popular are often characterized by being justifiable by generally acceptable kinds of arguments.

In this article, we assess if this theory can explain the similarities in opinion change in the United States and the United Kingdom. A recent study found that people in these countries have identical views on which kinds of moral arguments apply to which opinions (Vartanova et al., 2021). If it also holds that the same arguments have the same effect on opinion change in both countries, public opinion should change in the same way in the United Kingdom as in the United States. The theory says that it is generally acceptable kinds of arguments that are effective in driving opinion change. Our first research question is therefore whether the same kinds of arguments are generally acceptable in both countries.

Hypothesis 1. In both the United Kingdom and the United States, harm, violence, fairness, and liberty are generally acceptable kinds of arguments while authority, ingroup loyalty, purity, and government overreach are limitedly acceptable kinds of arguments.

If this holds, the theory predicts that the opinions that have more support from generally acceptable kinds of arguments should gain more in popularity over time:

Hypothesis 2. In both the United Kingdom and the United States, those moral opinions that have gained in popularity over the last 30–40 years are characterized by being justifiable by generally acceptable kinds of arguments.

This would also mean that the rate at which the popularity of specific opinions change should correlate between the two countries:

Hypothesis 3. The same opinions should be trending in the United Kingdom as in the United States.

We report two studies designed to test these hypotheses. The first study addresses Hypothesis 1. The second study addresses Hypotheses 2 and 3.

Study 1

Method

Participants

Participants were invited using the Prolific platform (prolific.co). Prolific provides a large international subject pool for online studies, which yields high-quality data (Eyal et al., 2021). Volunteers to participate in such studies sign up to the platform and provide demographic information that researchers may use for screening of participants. We set up a study with Prolific with a recruitment goal of 100 liberals from the United States, 100 conservatives from the United States, 100 left-wing participants from the United Kingdom, and 100 right-wing participants from the United Kingdom. The final sample consisted of 201 participants from the United States (age M = 34.2 years, SD = 13.0; 54% women; 47% identifying as liberals and 48% identifying as conservatives in our study) and 201 participants from the United Kingdom (age M = 37.7 years, SD = 13.7; 58% women; 49% identifying as left-wing and 46% identifying as right-wing in our study). Data used were collected between the 2nd and 4th of November 2020. Respondents were paid £0.39 for their participation in the study.

Procedure

Following Graham et al. (2011), participants were asked “When you decide whether something is right or wrong, to what extent are the following considerations relevant to your thinking?” A total of 24 considerations were rated, designed to cover eight kinds of arguments: harm, violence, fairness, liberty, authority, ingroup loyalty, purity, and government overreach (Vartanova et al., 2021). For example, the violence kind of arguments included whether violence is used, whether someone is killed, and whether someone is physically harmed. Each kind of argument was similarly covered by three items, see Supplementary Table 1 for the full list. Ratings were given on a 6-point scale between not at all relevant (coded 0) and extremely relevant (5).

In addition to rating the relevance of different arguments, participants reported their gender, age, and political ideology, and other sociodemographic characteristics. To measure political ideology, the U.S. study used an established liberal-conservative scale (extremely liberal, liberal, slightly liberal, moderate, slightly conservative, conservative, extremely conservative, or libertarian). In the United Kingdom, ideology was instead measured on a left-right scale from 0 to 10. We coded responses from 0 to 4 as left and from 6 to 10 as right.

Results

The internal consistency of ratings of each kind of argument was typically adequate, see Supplementary Table 2 for details. Figure 1 presents results on relevance ratings aggregated for each kind of argument; for five out of eight kinds, the figure also includes corresponding results from Graham et al. (2011). The left panel shows the acceptability of each kind of argument in the population at large, estimated by the average of the mean rating among liberals (or left) and the mean rating among conservatives (or right). Consistent with Hypothesis 1, results were the same in the United States and the United Kingdom, as well as across our study and the previous study by Graham et al. Specifically, four kinds of arguments (ingroup loyalty, authority, purity, and government overreach) showed limited acceptability, with mean ratings well below 3, and particularly low ratings among liberals/left. The other four kinds of argument (harm, violence, fairness, and liberty) all showed general acceptability, with mean ratings of 3 or above, and slightly higher ratings among liberals/left.

Study 2

In Study 1, we established that the same kinds of moral arguments are generally acceptable in the United States and the United Kingdom. In the second study, we examine whether we can predict opinion trends in both countries, using measures of which kinds of arguments are used to justify specific moral opinions. For a wide range of moral opinions, such measures are available from prior research (Vartanova et al., 2021). For these moral opinions, we here estimate opinion trends using polling data from the General Social Survey (GSS) in the United States and British Social Attitudes survey (BSA) in the United Kingdom. The GSS is a biannual survey asking demographic, behavioral, and attitudinal questions to representative samples of American respondents since 1972 (Smith et al., 2019). The BSA is an annual survey asking similar questions to representative samples of the British population since 1982 (NatCen Social Research, 2019).

Method

Selection of moral opinions

We use Vartanova et al.’s (2021) selection of 98 items from the GSS and 108 items from the BSA. This selection was based on content (the item must correspond to a “moral” opinion, as coded by a research assistant) and availability of data (must be available for at least three different years with a time span of at least 8 years from the first data collection to the last). For the wordings of these items, see Supplementary Tables 3 and 4.

Measures of which kinds of arguments are used to justify which opinions

To measure which kinds of arguments are used to justify which opinions on the selected items from the GSS and the BSA, Vartanova et al. (2021) surveyed U.S. and U.K. samples. Every item (e.g., “Homosexual couples should have the right to marry one another”) was judged by more than 100 participants. For each of two opposite opinions on the item (“agree” and “disagree”), these participants ticked all arguments they believe apply to justify that opinion. As a measure of the applicability of, say, harm arguments, Vartanova et al. used the proportion of the sample that reported harm arguments as applying to that opinion. Thus, for every kind of argument, an applicability measure between 0 and 1 was obtained (available at https://github.com/irinavrt/moral-args-appli). These measures were largely independent of whether they were calculated among women or men, among young or old, among liberals or conservatives, or among people with or without higher education (Vartanova et al., 2021).

From these measures, the “agree” opinion’s advantage with respect to, say, harm arguments is calculated as the applicability of harm arguments to justify the “agree” opinion minus the applicability of harm arguments to the “disagree” opinion. In this way we obtain advantage measures with respect to eight different kinds of arguments. We averaged the advantage measures across the four generally acceptable kinds of arguments (harm, violence, fairness, and liberty) to obtain a specific opinion’s HVFL argument advantage. We similarly averaged the advantage measures across the four limitedly acceptable kinds of arguments (authority, ingroup loyalty, purity, and government overreach) to obtain a specific opinion’s AIPG argument advantage. See Supplementary Tables 3 and 4 for these values.

Estimation of opinion trends

In the GSS and the BSA, some of the selected items have dichotomous responses (e.g., “agree” or “disagree”), some include neutral responses (such as “neither agree nor disagree”), and some include graded responses (such as “slightly agree” and “strongly agree”). To obtain estimates of opinion trends that are comparable across items, we dichotomize all items by omitting neutral responses and by combining graded responses. The dichotomized opinion data consist of 1’s and 0’s representing “agree” and “disagree,” respectively. We use all available data from the GSS and the BSA, amounting to a total of between 1,857 and 69,605 respondents per item depending on how many times the item has been included in the survey. Opinion change rates were estimated using logistic regression with time, measured in decades, as a single predictor. (Change rate estimates are virtually identical if age and gender are included as covariates, r = .997.) By this procedure we obtained change rates in terms of average change in log odds per 10 years, where “odds” refer to the proportion of 1’s divided by the proportion of 0’s. For example, if the proportion of 1’s has increased from 50% to 75% over a time span of 30 years, the odds have changed from 0.50/(1–0.50) = 1 to 0.75/(1–0.75) = 3, which is a factor of 3. In log odds the change is log 3 = 1.10 over 30 years, which corresponds to an average change rate per 10 years of 1.10/3 = 0.37. Note that if the proportion of 1’s had decreased from 75% to 50% over 30 years, the change rate would be −0.37 instead. Thus, the sign of the change rate tells us whether the opinion has increased or decreased in popularity.

Change rates with standard errors are reported in Supplementary Tables 3 and 4. Across items there was a large span in change rates: between −0.85 and 1.02 for the 98 GSS items, and between −1.17 and 1.07 for the 108 BSA items.

Analysis plan

Hypothesis 2 says that those opinions that are characterized by applicability of generally acceptable kinds of arguments have generally become more popular in both the United States and the United Kingdom. In both countries we therefore perform linear regressions of opinions’ change rates with their HVFL argument advantage as a predictor. We also include the opinions’ AIPG advantage as a second predictor to examine the claim that it is specifically the generally acceptable kinds of arguments that drive opinion change. We report standardized coefficients.

Results

Note that the unit of analysis is items representing opinions on moral issues: 98 items in the GSS and 108 items in the BSA. A linear regression predicting opinion change rates in the United States yielded a very large positive effect of HVFL advantage, β = 0.73, 95% CI [0.45, 1.01], and a negligible effect of AIPG advantage, β = 0.00, 95% CI [−0.28, 0.28]. The total proportion of variance explained (R squared) was 53%. The strong relation between an opinion’s HVFL argument advantage and its popularity trend in the United States is illustrated in the left panel of Figure 2.

The corresponding analysis in the United Kingdom yielded similar results: a very large positive effect of HVFL advantage, β = 0.78, 95% CI [0.51, 1.04], and a nonsignificant effect of AIPG advantage, β = 0.19, 95% CI [−0.07, 0.46]. The total proportion of variance explained (R squared) was 39%. The strong relation between an opinion’s HVFL argument advantage and its popularity trend in the United Kingdom is illustrated in the right panel of Figure 2. In sum, Hypothesis 2 was supported: In both the United States and the United Kingdom, trending moral opinions are characterized by being justifiable by generally acceptable kinds of arguments.

Comparing U.S. and U.K. opinion trends for similar issues

The selected issues in the United States and the United Kingdom have some overlap that can be used for making direct comparisons between the two countries. Among the selected issues in the United States and the United Kingdom, Vartanova et al. (2021) identified 27 pairs as equivalent, see Supplementary Tables 3 and 4. The left panel of Figure 3 illustrates that the HVFL argument advantage measures for these 27 moral opinions were practically identical in the two countries, r = .96, 95% CI [0.92, 0.98]. The right panel of Figure 3 illustrates that, consistent with Hypothesis 3, also opinion trends in the two countries were extremely highly correlated, r = .85, 95% CI [0.70, 0.93].

Discussion

Public opinion on moral issues shifts over time. What characterizes those opinions that are gaining in popularity? According to a recent theory of argument-driven opinion dynamics, opinions gain in popularity over time if they are justifiable by kinds of arguments that are acceptable to both liberals and conservatives (Eriksson & Strimling, 2015; Strimling et al., 2019). Here we tested this prediction across two countries: the United States and the United Kingdom.

In both countries, we found the generally acceptable kinds of arguments to be those that relate to harm, violence, fairness, and liberty (Study 1). This study was limited by the small size of the sample (200 respondents per country) and its potential lack of representativeness; however, results were consistent with previous studies based on much larger samples in multiple countries (Graham et al., 2011). We then examined opinion trends on 98 moral issues in the United States and 108 moral issues in the United Kingdom (Study 2). In both countries, justifiability by generally acceptable kinds of arguments served to characterize the trending opinions. Opinion trends were based on surveys of large representative samples, but measures of opinions’ justifiability by generally acceptable kinds of arguments were based on relatively small samples. This may not be a serious limitation, however, as prior research has shown that these justifiability measures are robust across gender, age, and educational levels.

Consistent with the theory being applicable in both countries, opinion trends in the United States and the United Kingdom were almost identical for those moral issues that had been polled in both countries. This is a remarkable finding, because there are a range of differences between the two countries that could potentially have affected their moral opinion dynamics in different ways. For example, compared to the United Kingdom, the United States has a greater political focus on moral issues (Bafumi & Shapiro, 2009; Evans et al., 1996; Green, 2007), a higher level of religiosity (Oppenheimer et al., 2014; Solt et al., 2011), and more politicized TV news (Kaid & Strömbäck, 2008). Despite these differences, we found the two countries had undergone highly similar patterns of opinion change across a range of moral issues. Our interpretation is that, in the long run, it is mainly the applicability of generally acceptable arguments that determine opinion trends.

Note that the crucial role of arguments does not preclude the importance of other mechanisms for political opinion change, such as social movements, critical moments in a country’s political history, and the influence from the political elites and media (e.g., Clifford et al., 2015; Costain & Majstorovic, 1994; Hill et al., 2013; Pollock III, 1994). It indicates, however, that such mechanisms may be aligned with the moral argument process. For instance, political elites and media could be the first to move toward opinions based on generally acceptable moral arguments, so that their influence pulls in the same direction. The influence of critical moments may be large when they occur but, in the long run, such short-term influences are likely to be dominated by the constant force on opinions exerted by moral arguments.

The empirical pattern is consistent with the theory that long-term shifts in public opinion on moral issues are driven by which kind of arguments justify which opinion on the issue (Eriksson & Strimling, 2015; Strimling et al., 2019). The pattern is too strong to be coincidental, especially as it is observed in two different countries. Moreover, we believe reverse causality is unlikely. Namely, prior research has found that people with different opinions on an issue nonetheless tend to agree on which kinds of arguments justify which opinion (Vartanova et al., 2021). We conclude that it is the applicability of generally acceptable kinds of arguments that somehow drives opinion trends.

This could happen through several pathways. Assuming an argument must be acceptable to be persuasive, persuasion in social interactions should lead to a tendency for opinion change to go mostly in the direction supported by generally acceptable kinds of arguments (Strimling et al., 2019). Another possibility is that people change opinions due to their own deliberations, in which they give more weight to the kinds of arguments they find acceptable. It could also be that the acceptability of different kinds of moral arguments changes over time. If arguments related to harm, violence, fairness, and liberty have become increasingly acceptable over time, it would have facilitated the observed opinion change, whether it has occurred through persuasion or deliberation or, most likely, both. Note, however, that in Study 1 we found no clear indication of change in acceptability compared to data collected a decade earlier (Graham et al., 2011).

Across a broad range of moral issues, we here discovered a distinct pattern of opinion trends that was almost identical in the United States and in the United Kingdom. The extent to which this finding generalizes to other countries is an open question to be settled in future research. We expect the same pattern to be found in another country as long as three conditions are met: (a) the same kinds of arguments are generally acceptable as in the United States/United Kingdom; (b) to justify specific opinions, the same kinds of arguments are used as in the United States/United Kingdom; and (c) there is free speech as in the United States/United Kingdom, so that people are exposed to arguments for both sides. All these conditions can be empirically examined in future research.

Supplementary Data

Supplementary Data are available at IJPOR online.

Funding

This work was supported by Knut and Alice Wallenberg Foundation [grant number 2017.0257]; and the Swedish Research Council [grant number 2019-02759]

Conflict of Interest

None declared.

References

Kimmo Eriksson PhD is a professor of Mathematics at Mälardalen University and a researcher at the Institute for Futures Studies. His main research area is mathematical modeling and cross-cultural studies.

Irina Vartanova PhD is a researcher at the Institute for Futures Studies. Her main research interest is statistical analysis.

Pontus Strimling PhD is a research director at the Institute for Futures Studies. His main research area is social norm change.