Economic and cultural dimensions

Take, as an illustration, an adult who wants the government to guarantee jobs and wants abortion never permitted by law, a liberal position on one question and a conservative position on another. Split an economic index and a cultural index from the American National Election Studies (ANES) each at its median, the point with equal numbers of adults on either side, and in 1984 46%46.2% percent of adultsEstimandweighted share on opposite sides of the two weighted medians (economic right and cultural left, or the reverse)Interval95% CI 43.8% to 48.6%n1,508Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod1984SourceANES Cumulative Data File of ANES adults sat on the conservative side of one index and the liberal side of the other. In 2024, 30%30.0% percent of adultsEstimandweighted share on opposite sides of the two weighted medians (economic right and cultural left, or the reverse)Interval95% CI 28.0% to 31.9%n4,387Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File did.

The two kinds of conservatism drew together after 2008

The chapter measures the link with a correlationCorrelationA number from minus 1 to plus 1 for how closely two measures move together in a straight-line pattern. Zero means no straight-line pattern; a correlation does not say that one measure causes the other., a number from minus one to plus one for how closely two measures move together along a straight line. Near zero, knowing where an adult stands on the economic index says little about the cultural one; near one, it nearly fixes it. Among adults with at least two answers on each index, the correlation between the economic and cultural indices was +0.10+0.10 correlation coefficient, from −1 to +1Estimandweighted Pearson correlation of the economic and cultural conservatism indices (Fisher z interval)Interval95% CI 0.05 to 0.15n1,508Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod1984SourceANES Cumulative Data File in the 1984 ANES and +0.51+0.51 correlation coefficient, from −1 to +1Estimandweighted Pearson correlation of the economic and cultural conservatism indices (Fisher z interval)Interval95% CI 0.48 to 0.55n4,387Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File in 2024.

Figure 1 follows every study. The correlation stayed at or below +0.23+0.23 correlation coefficient, from −1 to +1Estimandweighted Pearson correlation of the economic and cultural conservatism indices (Fisher z interval)Interval95% CI 0.15 to 0.30n954Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2008SourceANES Cumulative Data File from 1984 through 2008 and then rose through 2020. The intervals for 2008 and 2012 overlap, and so do those for 2012 and 2016, so the rise is best read across the whole period from 2008 to 2020 and not as a single step. The 2024 estimate contains the 2020 value, so the series has not risen further since 2020. A correlation of this size is well short of one, so positions on the two indices can go together more often than in 1984 while 30%30.0% percent of adultsEstimandweighted share on opposite sides of the two weighted medians (economic right and cultural left, or the reverse)Interval95% CI 28.0% to 31.9%n4,387Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File of adults still sit on opposite sides of the medians.

Part of the correlation is a difference between the parties. Comparing adults with the same party identification, on the seven-point ANES scale from strong Democrat to strong Republican, removes differences between the parties. The correlation of what remains is +0.23+0.23 correlation coefficient, from −1 to +1Estimandweighted correlation net of seven-point party identification (Fisher z interval, effective n less six party dummies)Interval95% CI 0.19 to 0.27n4,373Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File in 2024, and in 1984 it was +0.04 (95% CI −0.01 to 0.10)+0.04 correlation coefficient, from −1 to +1Estimandweighted correlation net of seven-point party identification (Fisher z interval, effective n less six party dummies)Interval95% CI −0.01 to 0.10n1,499Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod1984SourceANES Cumulative Data File, an interval that includes zero. The dashed line in the figure follows it. A correlation remains within party identification, and it rose from 2008 to 2024, at a level below the observed correlation in every study since 1988.

How the ANES measures two kinds of conservatism

The ANES asked three economic and three cultural questions in most studies from 1984 to 2024, which gives 1515 ANES studiesEstimandcount of studiesWeightsnone (count or constant)Populationstudies with both indicesPeriod1984-2024SourceANES Cumulative Data File studies and 33,25733,257 respondentsEstimandcount of respondents in the analysis sampleWeightsnone (count or constant)Populationadults with at least two of three items answered on each indexPeriod1984-2024SourceANES Cumulative Data File respondents with both indices. The economic index averages answers on guaranteed jobs, government services and health insurance; the cultural index averages answers on abortion, feelings toward gays and lesbians and traditional values. Each answer is put on a common scale and oriented so that a higher value is more conservative, and each index averages the answers an adult gave, at least two of three. Every estimate is weightedSurvey weightsNumbers that scale each respondent so the sample matches the population on known traits such as age, race, education and region. A weighted average counts a respondent with a large weight for more than one with a small weight. so that each study's sample matches the adult population on known traits. The qualification that matters is interview modeSurvey modeHow respondents were interviewed: in person, by phone, on paper or online. Mode can change answers, so a change in mode can look like a change in opinion.. Web interviews began in 2012, and the step from 2008 to 2012 appears only in the web sample (+0.24+0.24 correlation coefficient, from −1 to +1Estimandweighted correlation with the mode-specific weightInterval95% CI 0.17 to 0.30n1,769Weightsweight_ftf or weight_webPopulationadults with at least two of three items answered on each index, interviewed face-to-facePeriod2012SourceANES Cumulative Data File face-to-face, +0.37+0.37 correlation coefficient, from −1 to +1Estimandweighted correlation with the mode-specific weightInterval95% CI 0.33 to 0.41n3,434Weightsweight_ftf or weight_webPopulationadults with at least two of three items answered on each index, interviewed on the webPeriod2012SourceANES Cumulative Data File web). That step cannot be told apart from the change of mode, so the rise is better dated from 2012. In each mode the 2024 correlation is above the 2012 correlation, so the rise after 2012 is present in both. The report gives every estimate with its interval and sample size.

More adults sit in the corners than the correlation implies

A correlation can be restated as a count of people. Figure 2 places the adults with both indices in 2024 on a grid, the economic index across and the cultural index down, each measured in standard deviationsStandard deviationA measure of how far values typically fall from their average. A larger standard deviation means the values are more spread out. from its median. A standard deviation is how far values typically fall from their average. The two darkest cells are the corners where adults are far left on both indices, 8.2%8.2% percent of adultsEstimandweighted share of adults in one cell of the joint distribution (Wilson interval)Interval95% CI 7.1% to 9.5%n353Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File of adults, and far right on both, 8.5%8.5% percent of adultsEstimandweighted share of adults in one cell of the joint distribution (Wilson interval)Interval95% CI 7.3% to 9.8%n416Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File. Two bell-shaped indices with the observed correlation would put 6.4%6.4% percent of adultsEstimandshare in each corner cell (both indices beyond one standard deviation on the same side) for two normally distributed indices centred at the medians with the observed correlationWeightsnone (count or constant)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File in each of those corners, and the 95% intervals of both observed corners lie above that reference. The corner for economic left and cultural right holds 1.4%1.4% percent of adultsEstimandweighted share of adults below the economic median by more than one standard deviation and above the cultural median by at least one (Wilson interval)Interval95% CI 1.0% to 2.0%n47Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File, also above its reference. The opposite corner, economic right and cultural far left, has too few respondents to show, and its blank cells together hold less of the public than the economic-left, cultural-right corner alone. Adults far to opposite sides of the two medians are therefore more often economic left with cultural right than the reverse in 2024.

The reference does not separate the way the two indices go together from the shape of each index. The indices average a few bounded answers, so their own distributions are not exactly bell-shaped, and the excess in the corners may come from that shape as much as from the joint pattern. The excess describes one study and is not evidence of polarization.

The parties moved apart on both indices

Part of the observed correlation sits between the parties, which moved apart on both indices. Figure 3 plots the mean of each index for Democrats and Republicans, with leanersPartisan leanerA respondent who first says they are independent but, when asked, says they lean toward one party. Most figures group leaners with the party they lean toward, and each chapter says which grouping it uses. counted with their party, in standard deviations of the index over all respondents from 1984 to 2024. The Republican minus Democratic gap on the cultural index grew from +0.30+0.30 pooled standard deviations of the index (0 is the mean of all 1984-2024 respondents)Estimanddifference in weighted means of the cultural indexInterval95% CI 0.20 to 0.40n1,332Weightsweight (VCF0009z)PopulationRepublicans minus Democrats, leaners includedPeriod1984SourceANES Cumulative Data File in 1984 to +1.27+1.27 pooled standard deviations of the index (0 is the mean of all 1984-2024 respondents)Estimanddifference in weighted means of the cultural indexInterval95% CI 1.20 to 1.34n4,138Weightsweight (VCF0009z)PopulationRepublicans minus Democrats, leaners includedPeriod2024SourceANES Cumulative Data File in 2024, and the gap on the economic index grew as well. The movement was mostly on the Democratic side. The Republican mean on the cultural index also fell, and so did the mean of all adults, so part of each party's cultural change is a shift of the whole sample. When two groups differ on both indices, a correlation appears in the pooled sample even if the indices are unrelated inside each group. That is party sortingParty sortingThe alignment of party with other traits, such as ideology, religion, race and place, so that people who share a party also tend to share those traits. for these two indices, and Chapter 1.1, Party and ideology follows it for ideology more broadly.

Comparing within party identification is a descriptive comparison. Adults may take positions from the party they support, or support the party that matches their positions, and these data cannot separate the two. A rising correlation fits an account in which the public has moved toward the alignment the parties present, and the party gaps are consistent with it, but nothing in these data tests that account. Earlier studies of the American public describe a structure that differs from the one the parties offer. Treier and Hillygus found that many adults hold liberal preferences on one dimension and conservative preferences on the other [1]. A cross-national study found a positive correlation between the two kinds of conservatism uncommon across nations [2], and Chapter 1.3, Left and right across countries asks a related question across countries. Measurement error also holds the observed correlation down, since indices built from a few answers measure their attitude imperfectly. Corrected for that error, the 2024 correlation is +0.69+0.69 correlation coefficient, from −1 to +1Estimandobserved correlation divided by the square root of the product of the two alphas (bootstrap percentile interval)Interval95% CI 0.65 to 0.74n4,387Weightsweight (VCF0009z)Populationadults with at least two of three items answered on each indexPeriod2024SourceANES Cumulative Data File, and even the corrected indices leave the rest of the variation in each attitude unshared. The correction assumes that the errors in the two indices are unrelated to each other and that the items of an index measure the same attitude.

What two fixed indices cannot tell you

The correlation is between two indices built in advance from a fixed set of answers, so it measures how closely two given attitudes go together and does not estimate how many dimensions the public holds. A model fitted to a wider set of questions could find a different number, and issues outside both indices, such as immigration and race, may form a third dimension. Each study interviews different adults, so the series cannot show whether individuals' attitudes moved together or whether the population changed as cohorts replaced one another. The intervals treat each weighted sample as a simple random sample and ignore the ANES sampling design.

Go deeper

• The report: every estimate with its interval and sample size.
• Tables: every figure of the report as a table.
• World values: the explorer for this part.

Reading paths that include this chapter:

• The core argument in eight chapters

Data sources

• ANES Cumulative Data FileANES Time Series Cumulative Data File, studies 1984 to 2024: jobs (VCF0809), services and spending (VCF0839), health insurance (VCF0806), abortion (VCF0838), gay and lesbian thermometer (VCF0232), traditional values (VCF0853), party identification (VCF0301), liberal-conservative self-placement (VCF0803)Years: 1948-2024 (33 study years; even years only after 2004; no 2006, 2010 or 2014 study) · Weights: weight (VCF0009z); weight_ftf and weight_web for the interview-mode comparison · License: Free download after accepting the ANES terms of use; cite the release.

All data sources

References

1. Treier, Shawn; Hillygus, D. Sunshine (2009). The Nature of Political Ideology in the Contemporary Electorate. Public Opinion Quarterly 73(4): 679-703. academic.oup.com/poq/article-abstract/73/4/679/1829142
2. Malka, Ariel; Lelkes, Yphtach; Soto, Christopher J. (2019). Are Cultural and Economic Conservatism Positively Correlated? A Large-Scale Cross-National Test. British Journal of Political Science 49(3): 1045-1069. www.cambridge.org/core/journals/british-journal-of-political-science/article/are-cultural-and-economic-conservatism-positively-correlated-a-largescale-crossnational-test/83AFEDEA5E004CF23631C5388E7C9F67

How to cite

@misc{singh2026economicandculturaldimensions,
  author = {Singh, Vishal},
  title = {Economic and cultural dimensions},
  year = {2026},
  url = {https://vishalsingh.org/politics/economic-and-cultural-dimensions},
  note = {In The Public Mind. Published 30 September 2026.}
}