Chapter 7.2

Economic and cultural dimensions

The correlation between economic and cultural conservatism among American adults rose from +0.10 in the 1984 ANES to +0.51 in 2024, and is +0.23 among adults with the same seven-point party identification.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 30 September 2026

DraftFigures and numbers may change as the analysis is rebuilt.what is still being checked

To confirm before publication: the full question wording of the traditional-values item (VCF0853) against the ANES codebook; the chapter describes it by its Cumulative Data File label.

Economic and cultural conservatism went together more closely among American adults in 2024 than in 1984, and the correlation between an index of each rose from +0.10 (95% CI 0.05 to 0.15) in the 1984 American National Election Studies (ANES) to +0.51 (95% CI 0.48 to 0.55) in 2024. In a linear fit, adults one standard deviation more conservative on the economic index were on average 0.51 standard deviations more conservative on the cultural index in 2024. If the two kinds of conservatism were one dimension measured without error, the correlation would be one. Correcting the 2024 correlation for measurement error raises it to +0.69 (95% CI 0.65 to 0.74), at which the two indices share 48.1% of their variance, and among adults with the same seven-point party identification the observed correlation is +0.23 (95% CI 0.19 to 0.27).

The ANES asked three economic and three cultural questions in most studies from 1984 to 2024, which gives 15 studies and 33,257 respondents with both indices. This chapter follows the correlation across those studies, asks how much of it remains within party identification and after allowing for measurement error, compares how closely each index goes with liberal-conservative self-placement, and restates the correlation as the share of adults on opposite sides of the medians. The Chapter 7.4, Left and right across countries chapter asks a related question across countries.

Earlier work on two dimensions

Studies of the American public describe a structure that differs from the one the parties offer. Feldman and Johnston found that economic and social ideology are the minimum needed to account for domestic policy preferences [1], and Treier and Hillygus found that many adults hold liberal preferences on one dimension and conservative preferences on the other [2]. Carmines, Ensley and Wagner report that the policy preferences of party elites lie on one dimension while those of citizens lie on two, and that adults whose preferences do not match either party's package are more likely to shift party allegiance [3]. Roll-call voting in Congress is described well by one main dimension [4]. A cross-national study of mass publics that includes the United States found that a positive correlation between cultural and economic conservatism was uncommon across nations and that a negative correlation was more common [5].

How the indices are built

Each index averages up to three ANES answers. Every answer is converted to a z-score, its distance from the item's average in standard deviations, and oriented so that a higher value is more conservative. A respondent needs at least two of an index's three answers. Figure 1 lists the items, their scales and the direction used, including a feeling thermometer rating of gays and lesbians.In the cleaned ANES file the thermometer's top category, which covers a range of degrees near the top of the scale, is stored as the top value. The methods note gives the detail. The direction of each item is read from the value labels of the ANES Cumulative Data File, was not chosen from a correlation with self-placement or with the other index, and the two indices share no answers. The items were not asked in every study, and 4 studies have an index with only two components; those studies are shaded in the trend figures.An index is a respondent's average z-score over the items answered, and the correlation does not depend on its unit. The party means and the self-placement regression use the index re-standardized over all respondents in the analysis sample, and the cross-pressure band and the 2024 cells use the standard deviation of the index within one study. Each figure says which.

Figure 1

Each index averages three ANES items, and a higher value always marks the conservative end

Items in the economic and cultural conservatism indices: ANES variable, scale as fielded, direction used and studies that asked the item, 1984 to 2024

Index	Item	ANES variable	Scale as fielded	Direction used	Studies asking it	Not asked in
Economic	Government guarantee of jobs and a good standard of living	VCF0809	1 = government sees to jobs and living standards; 7 = each person on their own	Kept: a higher code is economic right	15	none
Economic	Government services and spending	VCF0839	1 = many fewer services, reduce spending; 7 = many more services, increase spending	Reversed: a lower code (fewer services) is economic right	15	none
Economic	Government or private health insurance	VCF0806	1 = government insurance plan; 7 = private insurance plan	Kept: a higher code is economic right	12	1986, 1990, 1998
Cultural	Legal status of abortion	VCF0838	1 = never permitted by law; 4 = always a matter of personal choice	Reversed: a lower code (more restriction) is cultural right	15	none
Cultural	Feeling thermometer toward gays and lesbians	VCF0232	0 = coldest feeling; 100 = warmest feeling	Reversed: a lower code (colder) is cultural right	13	1986, 1990
Cultural	Should be more emphasis on traditional values (variable label in the Cumulative Data File)	VCF0853	1 = agree strongly; 5 = disagree strongly	Reversed: a lower code (agreement) is cultural right	14	1984

Show data

Data for Each index averages three ANES items, and a higher value always marks the conservative end
direction	index	missing	scale	studies	topic	vcf
Kept: a higher code is economic right	Economic	none	1 = government sees to jobs and living standards; 7 = each person on their own	15	Government guarantee of jobs and a good standard of living	VCF0809
Reversed: a lower code (fewer services) is economic right	Economic	none	1 = many fewer services, reduce spending; 7 = many more services, increase spending	15	Government services and spending	VCF0839
Kept: a higher code is economic right	Economic	1986, 1990, 1998	1 = government insurance plan; 7 = private insurance plan	12	Government or private health insurance	VCF0806
Reversed: a lower code (more restriction) is cultural right	Cultural	none	1 = never permitted by law; 4 = always a matter of personal choice	15	Legal status of abortion	VCF0838
Reversed: a lower code (colder) is cultural right	Cultural	1986, 1990	0 = coldest feeling; 100 = warmest feeling	13	Feeling thermometer toward gays and lesbians	VCF0232
Reversed: a lower code (agreement) is cultural right	Cultural	1984	1 = agree strongly; 5 = disagree strongly	14	Should be more emphasis on traditional values (variable label in the Cumulative Data File)	VCF0853

Read each row for the item, its scale as fielded and whether the stored code is kept or reversed so that a higher value is more conservative. The last two columns show how many of the 15 studies from 1984 to 2024 asked the item and which years are missing. Orientation follows the value labels of the ANES Cumulative Data File, not any correlation.Source: ANES Cumulative Data File, 1984 to 2024; item wording and coding only, no weights needed

The correlation rose from 2008 to 2020

The correlation was +0.10 in 1984, in a study where the cultural index has only two components and its items agree less closely than in later studies (see the section on measurement error). From 1992 through 2008 the estimates stayed between +0.12 in 2000 and +0.23 in 2008. They then rose to +0.52 in 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, without a claim about any single step. The 2020 interval lies above the 2008 interval. The 2024 estimate of +0.51 contains the 2020 value, so the series has not risen further since 2020. Figure 2 shows every study. Correlation intervals in this chapter are 95% confidence intervals that treat the weighted sample as a simple random sample, which leaves out the ANES sampling design and can move an interval either way.Each interval uses Fisher's z transformation [6] on the Kish effective sample size [7].

Figure 2

The correlation between economic and cultural conservatism was at most +0.23 through 2008 and rose to +0.52 by 2020

Weighted correlation between the economic and cultural conservatism indices by ANES study, 1984 to 2024, adults with at least two of three items on each index; 95% intervals

Show data

Data for The correlation between economic and cultural conservatism was at most +0.23 through 2008 and rose to +0.52 by 2020
Study	Series	Correlation	Interval low	Interval high	Respondents	Effective n	Economic items	Cultural items
1984	As observed	0.102	0.052	0.152	1,508	1,508	3	2
1986	As observed	0.069	0.004	0.134	914	914	2	2
1988	As observed	0.095	0.045	0.144	1,551	1,551	3	3
1990	As observed	0.033	−0.040	0.105	729	729	2	2
1992	As observed	0.156	0.113	0.198	2,005	1,996	3	3
1994	As observed	0.198	0.146	0.248	1,659	1,358	3	3
1996	As observed	0.193	0.136	0.248	1,413	1,139	3	3
1998	As observed	0.162	0.097	0.225	1,055	892	2	3
2000	As observed	0.117	0.041	0.192	811	652	3	3
2004	As observed	0.180	0.113	0.244	1,006	834	3	3
2008	As observed	0.228	0.152	0.301	954	620	3	3
2012	As observed	0.325	0.292	0.357	5,203	2,880	3	3
2016	As observed	0.391	0.356	0.425	3,325	2,297	3	3
2020	As observed	0.517	0.493	0.540	6,737	3,696	3	3
2024	As observed	0.514	0.482	0.546	4,387	2,059	3	3
1984	Same party ID (7-pt)	0.045	−0.006	0.095	1,499	1,499	3	2
1986	Same party ID (7-pt)	0.075	0.010	0.140	907	907	2	2
1988	Same party ID (7-pt)	0.034	−0.016	0.084	1,542	1,542	3	3
1990	Same party ID (7-pt)	−0.011	−0.084	0.062	727	727	2	2
1992	Same party ID (7-pt)	0.066	0.022	0.110	1,997	1,988	3	3
1994	Same party ID (7-pt)	0.107	0.054	0.159	1,652	1,352	3	3
1996	Same party ID (7-pt)	0.072	0.014	0.130	1,408	1,135	3	3
1998	Same party ID (7-pt)	0.090	0.024	0.155	1,047	886	2	3
2000	Same party ID (7-pt)	0.031	−0.046	0.109	806	648	3	3
2004	Same party ID (7-pt)	0.030	−0.039	0.098	993	822	3	3
2008	Same party ID (7-pt)	0.082	0.002	0.160	944	615	3	3
2012	Same party ID (7-pt)	0.106	0.069	0.142	5,189	2,872	3	3
2016	Same party ID (7-pt)	0.155	0.115	0.195	3,314	2,288	3	3
2020	Same party ID (7-pt)	0.227	0.197	0.258	6,725	3,689	3	3
2024	Same party ID (7-pt)	0.227	0.186	0.268	4,373	2,051	3	3
1984	As observed	0.102	0.052	0.152	1,508	1,508	3	2
1986	As observed	0.069	0.004	0.134	914	914	2	2
1988	As observed	0.095	0.045	0.144	1,551	1,551	3	3
1990	As observed	0.033	−0.040	0.105	729	729	2	2
1992	As observed	0.156	0.113	0.198	2,005	1,996	3	3
1994	As observed	0.198	0.146	0.248	1,659	1,358	3	3
1996	As observed	0.193	0.136	0.248	1,413	1,139	3	3
1998	As observed	0.162	0.097	0.225	1,055	892	2	3
2000	As observed	0.117	0.041	0.192	811	652	3	3
2004	As observed	0.180	0.113	0.244	1,006	834	3	3
2008	As observed	0.228	0.152	0.301	954	620	3	3
2012	As observed	0.325	0.292	0.357	5,203	2,880	3	3
2016	As observed	0.391	0.356	0.425	3,325	2,297	3	3
2020	As observed	0.517	0.493	0.540	6,737	3,696	3	3
2024	As observed	0.514	0.482	0.546	4,387	2,059	3	3
1984	Same party ID (3 cat.)	0.051	0.000	0.101	1,499	1,499	3	2
1986	Same party ID (3 cat.)	0.068	0.003	0.133	907	907	2	2
1988	Same party ID (3 cat.)	0.050	0.000	0.099	1,542	1,542	3	3
1990	Same party ID (3 cat.)	0.003	−0.070	0.076	727	727	2	2
1992	Same party ID (3 cat.)	0.070	0.026	0.114	1,997	1,988	3	3
1994	Same party ID (3 cat.)	0.126	0.073	0.178	1,652	1,352	3	3
1996	Same party ID (3 cat.)	0.094	0.036	0.151	1,408	1,135	3	3
1998	Same party ID (3 cat.)	0.101	0.036	0.166	1,047	886	2	3
2000	Same party ID (3 cat.)	0.039	−0.038	0.116	806	648	3	3
2004	Same party ID (3 cat.)	0.050	−0.019	0.118	993	822	3	3
2008	Same party ID (3 cat.)	0.104	0.025	0.181	944	615	3	3
2012	Same party ID (3 cat.)	0.127	0.091	0.163	5,189	2,872	3	3
2016	Same party ID (3 cat.)	0.180	0.140	0.219	3,314	2,288	3	3
2020	Same party ID (3 cat.)	0.241	0.211	0.272	6,725	3,689	3	3
2024	Same party ID (3 cat.)	0.242	0.201	0.282	4,373	2,051	3	3
1984	As observed	0.102	0.052	0.152	1,508	1,508	3	2
1986	As observed	0.069	0.004	0.134	914	914	2	2
1988	As observed	0.095	0.045	0.144	1,551	1,551	3	3
1990	As observed	0.033	−0.040	0.105	729	729	2	2
1992	As observed	0.156	0.113	0.198	2,005	1,996	3	3
1994	As observed	0.198	0.146	0.248	1,659	1,358	3	3
1996	As observed	0.193	0.136	0.248	1,413	1,139	3	3
1998	As observed	0.162	0.097	0.225	1,055	892	2	3
2000	As observed	0.117	0.041	0.192	811	652	3	3
2004	As observed	0.180	0.113	0.244	1,006	834	3	3
2008	As observed	0.228	0.152	0.301	954	620	3	3
2012	As observed	0.325	0.292	0.357	5,203	2,880	3	3
2016	As observed	0.391	0.356	0.425	3,325	2,297	3	3
2020	As observed	0.517	0.493	0.540	6,737	3,696	3	3
2024	As observed	0.514	0.482	0.546	4,387	2,059	3	3
1988	Reliability-corrected	0.181	0.083	0.277	1,551	1,551	3	3
1992	Reliability-corrected	0.273	0.201	0.349	2,005	1,996	3	3
1994	Reliability-corrected	0.332	0.253	0.415	1,659	1,358	3	3
1996	Reliability-corrected	0.316	0.229	0.408	1,413	1,139	3	3
1998	Reliability-corrected	0.317	0.191	0.454	1,055	892	2	3
2000	Reliability-corrected	0.213	0.062	0.364	811	652	3	3
2004	Reliability-corrected	0.276	0.180	0.371	1,006	834	3	3
2008	Reliability-corrected	0.349	0.241	0.461	954	620	3	3
2012	Reliability-corrected	0.471	0.427	0.516	5,203	2,880	3	3
2016	Reliability-corrected	0.550	0.496	0.602	3,325	2,297	3	3
2020	Reliability-corrected	0.680	0.648	0.710	6,737	3,696	3	3
2024	Reliability-corrected	0.693	0.653	0.738	4,387	2,059	3	3
1984	As observed	0.102	0.052	0.152	1,508	1,508	3	2
1986	As observed	0.069	0.004	0.134	914	914	2	2
1988	As observed	0.095	0.045	0.144	1,551	1,551	3	3
1990	As observed	0.033	−0.040	0.105	729	729	2	2
1992	As observed	0.156	0.113	0.198	2,005	1,996	3	3
1994	As observed	0.198	0.146	0.248	1,659	1,358	3	3
1996	As observed	0.193	0.136	0.248	1,413	1,139	3	3
1998	As observed	0.162	0.097	0.225	1,055	892	2	3
2000	As observed	0.117	0.041	0.192	811	652	3	3
2004	As observed	0.180	0.113	0.244	1,006	834	3	3
2008	As observed	0.228	0.152	0.301	954	620	3	3
2012	As observed	0.325	0.292	0.357	5,203	2,880	3	3
2016	As observed	0.391	0.356	0.425	3,325	2,297	3	3
2020	As observed	0.517	0.493	0.540	6,737	3,696	3	3
2024	As observed	0.514	0.482	0.546	4,387	2,059	3	3
1988	All six items answered	0.111	0.052	0.169	1,107	1,107	3	3
1992	All six items answered	0.203	0.154	0.251	1,513	1,507	3	3
1994	All six items answered	0.243	0.187	0.297	1,384	1,132	3	3
1996	All six items answered	0.221	0.158	0.281	1,132	923	3	3
2000	All six items answered	0.154	0.068	0.237	634	512	3	3
2004	All six items answered	0.209	0.135	0.280	803	664	3	3
2008	All six items answered	0.261	0.176	0.342	740	488	3	3
2012	All six items answered	0.370	0.335	0.403	4,422	2,472	3	3
2016	All six items answered	0.435	0.398	0.471	2,719	1,894	3	3
2020	All six items answered	0.563	0.538	0.587	5,374	2,991	3	3
2024	All six items answered	0.560	0.526	0.592	3,516	1,635	3	3
2008	All respondents	0.228	0.152	0.301	954	620	3	3
2012	All respondents	0.325	0.292	0.357	5,203	2,880	3	3
2016	All respondents	0.391	0.356	0.425	3,325	2,297	3	3
2024	All respondents	0.514	0.482	0.546	4,387	2,059	3	3
2008	Face-to-face	0.228	0.152	0.301	954	620	3	3
2012	Face-to-face	0.236	0.174	0.296	1,769	926	3	3
2016	Face-to-face	0.385	0.317	0.448	988	649	3	3
2024	Face-to-face	0.542	0.473	0.605	874	443	3	3
2012	Web	0.368	0.329	0.406	3,434	1,963	3	3
2016	Web	0.396	0.354	0.436	2,337	1,648	3	3
2024	Web	0.473	0.421	0.523	1,589	895	3	3

The observed correlation stays at or below +0.23 through 2008, rises to +0.52 in 2020 and is +0.51 in 2024. The dashed line compares adults with the same seven-point party identification, runs from +0.04 in 1984 to +0.23 in 2024 and rises over the same years at a lower level. Each point is one ANES study, the bands are 95% intervals and shaded studies have only two items on an index. The toggles show a coarser party control, a correction for unreliable indices, adults who answered all six items and the interview mode.Note: Shaded studies have only two items on one or both indices: the cultural index in 1984, both indices in 1986 and 1990, and the economic index in 1998.Note: The 2002 study has no cultural index, since only the gay and lesbian thermometer was asked, and is omitted.Source: ANES Cumulative Data File, 1984 to 2024; weighted (VCF0009z), and weight_ftf and weight_web in the mode view; intervals are model based

Two checks support the rise. Adults who answered every item give +0.11 in 1988 and +0.56 in 2024, and the six-item view of the figure shows the same rise. Interview mode is the second check. Web interviews began in 2012, and the step from +0.23 in 2008 to +0.32 in 2012 appears only in the web sample (+0.24 face-to-face, +0.37 web), so the step cannot be told apart from the change of mode and 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, and the pairs for 2016 and 2024 are in the caption of the mode view. The 2020 study has no mode weights, and the step from 2016 to 2020 cannot be split by mode.

Part of the correlation is between the parties

Comparing adults with the same party identification removes differences between the parties. Both indices are compared within each of the seven ANES party categories, which run from strong Democrat to strong Republican and keep leaners apart from weak partisans, and the correlation of what remains is +0.23 in 2024 against the observed +0.51. The same comparison gives +0.04 (95% CI −0.01 to 0.10) in 1984, an interval that includes zero, and +0.08 in 2008. A three-category control (Democrat, independent and Republican, with leaners counted as partisans) gives +0.24 in 2024, close to the seven-point figure. A correlation remains within party identification, and the estimate rose from 2008 to 2024.

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 leaners counted with their party, and for all adults. The Republican minus Democratic gap on the economic index grew from +0.69 in 1984 to +1.29 in 2024, and on the cultural index from +0.30 to +1.27, in standard deviations of the index over all respondents from 1984 to 2024. The lines are group means for each study on one fixed scale, so they compare groups between studies and do not follow individuals.

The cultural panel is less steady in 1984, 1986 and 1990, the studies in which the cultural index has only two components, because the item mix changes: 1984 has no traditional-values item, and 1986 and 1990 have no thermometer, and a two-item index is noisier. Built from abortion and the thermometer alone, the items asked in every study except 1986 and 1990, the cultural gap is +0.28 in 1984 and +1.11 in 2024, the Republican mean went from +0.61 to +0.31 and the Democratic mean from +0.33 to −0.81, so the widening and the fall in both party means do not depend on the third item.

Figure 3

The Republican minus Democratic gap grew from +0.69 to +1.29 on the economic index and from +0.30 to +1.27 on the cultural index

Mean of each index by party identification, leaners counted with their party, and for all adults, in standard deviations of the pooled 1984 to 2024 sample, by ANES study; 95% intervals

Mean of the index (pooled standard deviations)

Show data

Data for The Republican minus Democratic gap grew from +0.69 to +1.29 on the economic index and from +0.30 to +1.27 on the cultural index
Study	Index	Group	Mean	Interval low	Interval high	Respondents	Effective n	Items in the index
1984	Economic index	All adults	0.08	0.03	0.12	1,508	1,508	3
1986	Economic index	All adults	0.01	−0.05	0.07	914	914	2
1988	Economic index	All adults	0.06	0.02	0.11	1,551	1,551	3
1990	Economic index	All adults	−0.05	−0.12	0.01	729	729	2
1992	Economic index	All adults	−0.05	−0.08	−0.01	2,005	1,996	3
1994	Economic index	All adults	0.14	0.09	0.19	1,659	1,358	3
1996	Economic index	All adults	0.14	0.09	0.20	1,413	1,139	3
1998	Economic index	All adults	−0.07	−0.13	0.00	1,055	892	2
2000	Economic index	All adults	0.01	−0.06	0.07	811	652	3
2004	Economic index	All adults	−0.11	−0.17	−0.05	1,006	834	3
2008	Economic index	All adults	−0.06	−0.14	0.01	954	620	3
2012	Economic index	All adults	0.20	0.16	0.23	5,203	2,880	3
2016	Economic index	All adults	0.09	0.05	0.13	3,325	2,297	3
2020	Economic index	All adults	−0.11	−0.15	−0.08	6,737	3,696	3
2024	Economic index	All adults	−0.20	−0.25	−0.16	4,387	2,059	3
1984	Economic index	Democrats	−0.24	−0.30	−0.18	699	699	3
1986	Economic index	Democrats	−0.26	−0.34	−0.17	459	459	2
1988	Economic index	Democrats	−0.26	−0.32	−0.19	731	731	3
1990	Economic index	Democrats	−0.32	−0.41	−0.22	368	368	2
1992	Economic index	Democrats	−0.35	−0.40	−0.31	975	971	3
1994	Economic index	Democrats	−0.21	−0.27	−0.15	768	621	3
1996	Economic index	Democrats	−0.20	−0.26	−0.14	734	593	3
1998	Economic index	Democrats	−0.33	−0.42	−0.25	544	464	2
2000	Economic index	Democrats	−0.27	−0.36	−0.19	414	331	3
2004	Economic index	Democrats	−0.50	−0.58	−0.43	476	389	3
2008	Economic index	Democrats	−0.51	−0.59	−0.42	553	338	3
2012	Economic index	Democrats	−0.33	−0.37	−0.30	2,722	1,397	3
2016	Economic index	Democrats	−0.45	−0.50	−0.40	1,540	1,039	3
2020	Economic index	Democrats	−0.85	−0.89	−0.81	3,209	1,672	3
2024	Economic index	Democrats	−0.80	−0.85	−0.75	2,196	1,022	3
1984	Economic index	Republicans	0.45	0.39	0.51	633	633	3
1986	Economic index	Republicans	0.40	0.31	0.49	334	334	2
1988	Economic index	Republicans	0.43	0.37	0.49	664	664	3
1990	Economic index	Republicans	0.28	0.18	0.38	291	291	2
1992	Economic index	Republicans	0.38	0.32	0.43	790	787	3
1994	Economic index	Republicans	0.59	0.52	0.66	726	605	3
1996	Economic index	Republicans	0.64	0.56	0.71	561	454	3
1998	Economic index	Republicans	0.37	0.27	0.47	395	333	2
2000	Economic index	Republicans	0.48	0.39	0.57	304	254	3
2004	Economic index	Republicans	0.35	0.26	0.44	425	361	3
2008	Economic index	Republicans	0.52	0.40	0.64	282	214	3
2012	Economic index	Republicans	0.85	0.81	0.90	1,802	1,097	3
2016	Economic index	Republicans	0.77	0.72	0.83	1,389	983	3
2020	Economic index	Republicans	0.72	0.67	0.76	2,813	1,642	3
2024	Economic index	Republicans	0.49	0.43	0.55	1,942	920	3
1984	Cultural index	All adults	0.48	0.44	0.53	1,508	1,508	2
1986	Cultural index	All adults	0.20	0.14	0.26	914	914	2
1988	Cultural index	All adults	0.51	0.47	0.55	1,551	1,551	3
1990	Cultural index	All adults	0.21	0.15	0.27	729	729	2
1992	Cultural index	All adults	0.28	0.24	0.31	2,005	1,996	3
1994	Cultural index	All adults	0.46	0.41	0.50	1,659	1,358	3
1996	Cultural index	All adults	0.35	0.30	0.40	1,413	1,139	3
1998	Cultural index	All adults	0.23	0.17	0.28	1,055	892	3
2000	Cultural index	All adults	0.29	0.23	0.36	811	652	3
2004	Cultural index	All adults	0.17	0.11	0.24	1,006	834	3
2008	Cultural index	All adults	0.09	0.02	0.16	954	620	3
2012	Cultural index	All adults	−0.01	−0.05	0.02	5,203	2,880	3
2016	Cultural index	All adults	−0.23	−0.27	−0.19	3,325	2,297	3
2020	Cultural index	All adults	−0.33	−0.36	−0.29	6,737	3,696	3
2024	Cultural index	All adults	−0.33	−0.38	−0.29	4,387	2,059	3
1984	Cultural index	Democrats	0.33	0.27	0.40	699	699	2
1986	Cultural index	Democrats	0.19	0.10	0.28	459	459	2
1988	Cultural index	Democrats	0.42	0.36	0.48	731	731	3
1990	Cultural index	Democrats	0.10	0.01	0.19	368	368	2
1992	Cultural index	Democrats	0.09	0.03	0.14	975	971	3
1994	Cultural index	Democrats	0.27	0.19	0.34	768	621	3
1996	Cultural index	Democrats	0.18	0.10	0.25	734	593	3
1998	Cultural index	Democrats	0.03	−0.05	0.12	544	464	3
2000	Cultural index	Democrats	0.13	0.03	0.22	414	331	3
2004	Cultural index	Democrats	−0.10	−0.20	−0.01	476	389	3
2008	Cultural index	Democrats	−0.17	−0.27	−0.07	553	338	3
2012	Cultural index	Democrats	−0.41	−0.46	−0.37	2,722	1,397	3
2016	Cultural index	Democrats	−0.69	−0.74	−0.63	1,540	1,039	3
2020	Cultural index	Democrats	−0.89	−0.93	−0.85	3,209	1,672	3
2024	Cultural index	Democrats	−0.93	−0.99	−0.88	2,196	1,022	3
1984	Cultural index	Republicans	0.63	0.56	0.71	633	633	2
1986	Cultural index	Republicans	0.19	0.10	0.29	334	334	2
1988	Cultural index	Republicans	0.63	0.58	0.69	664	664	3
1990	Cultural index	Republicans	0.29	0.19	0.39	291	291	2
1992	Cultural index	Republicans	0.52	0.47	0.58	790	787	3
1994	Cultural index	Republicans	0.65	0.59	0.72	726	605	3
1996	Cultural index	Republicans	0.61	0.54	0.69	561	454	3
1998	Cultural index	Republicans	0.48	0.39	0.56	395	333	3
2000	Cultural index	Republicans	0.51	0.41	0.60	304	254	3
2004	Cultural index	Republicans	0.49	0.40	0.57	425	361	3
2008	Cultural index	Republicans	0.41	0.29	0.52	282	214	3
2012	Cultural index	Republicans	0.41	0.36	0.46	1,802	1,097	3
2016	Cultural index	Republicans	0.30	0.25	0.35	1,389	983	3
2020	Cultural index	Republicans	0.30	0.26	0.34	2,813	1,642	3
2024	Cultural index	Republicans	0.33	0.28	0.38	1,942	920	3

The gap between the Republican and Democratic lines widens on both panels, mostly because the Democratic line falls. The Republican minus Democratic gap on the economic index went from +0.69 in 1984 to +1.29 in 2024 and on the cultural index from +0.30 to +1.27. The grey dashed line is the mean of all adults. On the cultural panel it also falls, from +0.48 to −0.33, and so does the Republican line, so part of each party's fall is a shift of the whole sample. Positive values are more conservative than the average of all respondents from 1984 to 2024.Note: A party-by-study cell needs at least 100 respondents and an effective sample of 100; pure independents fall below that in most studies and are counted only in the line for all adults.Note: Shaded studies have only two items on one or both indices: the cultural index in 1984, both indices in 1986 and 1990, and the economic index in 1998. The cultural panel is noisier in those studies because an index of two items has lower alpha.Source: ANES Cumulative Data File, 1984 to 2024; weighted (VCF0009z); intervals are model based

The movement was mostly on the Democratic side. Democrats went from −0.24 to −0.80 on the economic index and from +0.33 to −0.93 on the cultural index. Republicans went from +0.45 to +0.49 on the economic index and from +0.63 to +0.33 on the cultural index, so the Republican mean on the cultural index also fell. The mean of all adults on the cultural index fell as well, from +0.48 to −0.33, so part of each party's cultural change is a shift of the whole sample and the gap measures the distance between the parties. 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 sorting for these two indices, and the Chapter 1.1, Party and ideology chapter 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.

Correcting for measurement error

Indices built from a few answers measure their topic with error, and error in either index lowers the observed correlation. The standard summary of how closely the items of one index agree is Cronbach's alpha [8], which runs from zero, when the items are unrelated, to one, when they agree perfectly. Figure 4 gives alpha for each study. The economic index went from 0.60 in 1984 to a peak of 0.83 in 2020 and 0.79 in 2024, and the cultural index from 0.44 in 1988, the first study in which the cultural index has three components, to 0.69 in 2024. The cultural index has the lower alpha in 14 of the 15 studies, the exception being 1998, and each index has its lowest alpha in a study where it has only two components. Alpha is computed only on adults who answered every item of an index. The correlation also uses adults with two of three answers, who are measured less reliably than alpha suggests: on the economic index they are 58.3% of the 1984 sample, because the health-insurance item was answered by only part of that sample, and 15.4% of the 2024 sample.

Figure 4

Items in each index agree more in recent studies, and the cultural items agree less than the economic items in 14 of 15 studies

Standardized Cronbach alpha of each index by ANES study, adults who answered every item of the index that year, 1984 to 2024; 95% intervals from resampling respondents

Show data

Data for Items in each index agree more in recent studies, and the cultural items agree less than the economic items in 14 of 15 studies
Study	Index	Alpha	Interval low	Interval high	Respondents with every item	Items in the index
1984	Economic index	0.60	0.53	0.66	629	3
1986	Economic index	0.43	0.33	0.51	914	2
1988	Economic index	0.62	0.57	0.66	1,171	3
1990	Economic index	0.47	0.37	0.56	729	2
1992	Economic index	0.62	0.59	0.66	1,588	3
1994	Economic index	0.67	0.63	0.71	1,451	3
1996	Economic index	0.66	0.62	0.71	1,185	3
1998	Economic index	0.47	0.37	0.55	1,055	2
2000	Economic index	0.60	0.53	0.66	675	3
2004	Economic index	0.69	0.64	0.73	832	3
2008	Economic index	0.73	0.68	0.77	765	3
2012	Economic index	0.76	0.74	0.78	4,486	3
2016	Economic index	0.79	0.77	0.81	2,772	3
2020	Economic index	0.83	0.82	0.84	5,691	3
2024	Economic index	0.79	0.77	0.81	3,743	3
1984	Cultural index	0.35	0.28	0.41	1,508	2
1986	Cultural index	0.37	0.29	0.44	914	2
1988	Cultural index	0.44	0.38	0.49	1,462	3
1990	Cultural index	0.17	0.04	0.29	729	2
1992	Cultural index	0.52	0.48	0.56	1,900	3
1994	Cultural index	0.53	0.48	0.57	1,582	3
1996	Cultural index	0.56	0.51	0.60	1,346	3
1998	Cultural index	0.56	0.51	0.60	989	3
2000	Cultural index	0.50	0.43	0.57	759	3
2004	Cultural index	0.61	0.57	0.66	971	3
2008	Cultural index	0.58	0.52	0.64	924	3
2012	Cultural index	0.62	0.60	0.65	5,118	3
2016	Cultural index	0.64	0.61	0.67	3,251	3
2020	Cultural index	0.69	0.68	0.71	6,354	3
2024	Cultural index	0.69	0.67	0.72	4,109	3

Alpha is higher in recent studies for both indices: the economic index went from 0.60 in 1984 to a peak of 0.83 in 2020 and 0.79 in 2024, and the cultural index from 0.44 in 1988, the first study with three cultural items, to 0.69 in 2024. Alpha runs from zero, when the items are unrelated, to one, when they agree perfectly. Values are lower in the shaded studies, where an index has only two items.Note: Alpha uses adults who answered every item of the index that year.Note: Shaded studies have only two items on one or both indices: the cultural index in 1984, both indices in 1986 and 1990, and the economic index in 1998.Source: ANES Cumulative Data File, 1984 to 2024; weighted (VCF0009z); intervals are model based

Alpha rises when the items of an index measure the same thing with less noise, and it also rises when opinions on the items become more consistent across adults. A rise in alpha can therefore come from a fall in error, from a rise in the consistency of opinion, or from both, and the corrected series below inherits that mix.

Dividing the observed correlation by the square root of the product of the two alphas gives the correlation the indices would show if their items were error-free, a correction proposed by Spearman [9]. In 2024 it gives +0.69, against +0.35 in 2008 and +0.18 in 1988. The corrected view of Figure 2 draws it only where both alphas are at least 0.4, which leaves out 1984, 1986 and 1990, studies in which the cultural index has only two components and a lower alpha. The corrected series rises as well and lies above the observed series in every study where it is drawn, and even the corrected indices leave the rest of the variation in each attitude unshared.

Self-placement follows the economic and cultural indices about equally

The seven-point liberal-conservative scale, on which adults place themselves, is a label and not a bundle of positions. Regressing self-placement on both indices at once shows how closely each goes with the label when the other is held fixed. For 1984 to 1996, adults one standard deviation apart on the economic index differ by 0.50 points on the scale, and adults one standard deviation apart on the cultural index by 0.53 points, with the other index held fixed. For 2012 to 2024 the coefficients are 0.66 and 0.67. The cultural minus economic difference is +0.02 (95% CI −0.03 to 0.08) in the early period and +0.01 (95% CI −0.04 to 0.06) in the recent one, both intervals include zero, and neither extends beyond 0.08 points in either direction (Figure 5). Both coefficients are larger in the recent period, and the intervals for the two periods do not overlap. The studies from 1998 to 2008 are left out so that the two periods do not touch. The data do not separate the two indices in how closely each goes with the label.

Two cautions apply to that reading. An index with lower alpha carries more measurement error, which lowers its coefficient, and the cultural index has the lower alpha, so equal coefficients are consistent with a cultural attitude that goes with the label somewhat more closely than the economic attitude does. Alpha also rose for both indices between the two periods, so part of the rise in both coefficients may be less measurement error, as the section on measurement error describes.The regression includes a fixed effect for each study and uses HC3 robust standard errors. The Chapter 7.1, The conservative label (planned) chapter separates the label from policy preferences.

Figure 5

Self-placed conservatism follows the economic and cultural indices about equally, and both coefficients grew

Coefficient on each index in a regression of 7-point liberal-conservative self-placement on both indices (in pooled standard deviations) with study fixed effects, 1984 to 1996 and 2012 to 2024; 95% intervals

1984 to 1996
2012 to 2024

Points on the seven-point self-placement scale per standard deviation

Show data

Data for Self-placed conservatism follows the economic and cultural indices about equally, and both coefficients grew
early	early_hi	early_lo	index	n_early	n_recent	recent	recent_hi	recent_lo
0.502	0.536	0.467	Economic index	7,815	17,452	0.664	0.690	0.638
0.526	0.558	0.495	Cultural index	7,815	17,452	0.673	0.698	0.647

Both coefficients rose, the economic one by 0.16 and the cultural one by 0.15 points per standard deviation, and the two rows sit almost on top of each other, with a cultural minus economic difference of +0.02 in the early period and +0.01 in the recent one. Both difference intervals include zero and neither extends beyond 0.08 in either direction. Each dot is the difference in self-placement, on the seven-point scale, between two adults one standard deviation apart on one index and equal on the other. The studies from 1998 to 2008 are left out so that the two periods do not touch.Source: ANES Cumulative Data File, 1984 to 1996 and 2012 to 2024; weighted (VCF0009z); HC3 intervals

Adults on opposite sides of the medians

A correlation can be restated as a count of people. Splitting each index at its weighted median within a study puts equal numbers of adults on each side of each index. An adult is cross-pressured when right of the median on one index and left of it on the other, that is, conservative on one index and liberal on the other. That share was 46.2% (95% CI 43.8% to 48.6%) in 1984 and 30.0% (95% CI 28.0% to 31.9%) in 2024 (Figure 6), and it moves with the correlation.

For two bell-shaped indices with the observed correlation, the share on opposite sides of the medians would be 46.8% in 1984 and 32.8% in 2024.A bell-shaped (normal) pair with a given correlation is the reference case in which the correlation alone fixes how the two indices are spread jointly. The methods note gives the formula. The observed share is within 3.3 points of the implied value in every study, so a median split gives little evidence beyond the correlation itself. The difference has the same sign from 2008: the observed share is 2.1 to 3.3 points below the implied share in each study from 2008, and in 2012, 2016, 2020 and 2024 the implied share lies above the observed 95% interval. More adults sit on the same side of both medians than the correlation implies, which is the same feature as the excess in the corner cells shown below.

Figure 6

The share of adults on opposite sides of the medians fell from 46.2% to 30.0% and stays within 3.3 points of the share the correlation implies

Weighted share of adults by ANES study, 1984 to 2024; 95% intervals

Opposite sides of the medians
Strongly cross-pressured
Implied by the correlation

Show data

Data for The share of adults on opposite sides of the medians fell from 46.2% to 30.0% and stays within 3.3 points of the share the correlation implies
conf_high	conf_low	estimate	n	series	year
48.62	43.82	46.22	1,508	Opposite sides of the medians	1984
50.19	44.14	47.17	914	Opposite sides of the medians	1986
49.48	44.54	47.01	1,551	Opposite sides of the medians	1988
54.07	47.22	50.65	729	Opposite sides of the medians	1990
46.19	41.84	44.01	2,005	Opposite sides of the medians	1992
45.58	40.36	42.97	1,659	Opposite sides of the medians	1994
45.85	40.11	42.98	1,413	Opposite sides of the medians	1996
51.25	44.80	48.02	1,055	Opposite sides of the medians	1998
51.60	44.00	47.80	811	Opposite sides of the medians	2000
48.17	41.47	44.82	1,006	Opposite sides of the medians	2004
44.43	36.75	40.59	954	Opposite sides of the medians	2008
38.80	35.35	37.08	5,203	Opposite sides of the medians	2012
36.76	32.92	34.84	3,325	Opposite sides of the medians	2016
30.84	27.95	29.39	6,737	Opposite sides of the medians	2020
31.92	28.00	29.96	4,387	Opposite sides of the medians	2024
20.12	16.22	18.17	1,508	Strongly cross-pressured	1984
19.74	14.83	17.29	914	Strongly cross-pressured	1986
19.77	15.95	17.86	1,551	Strongly cross-pressured	1988
22.21	16.47	19.34	729	Strongly cross-pressured	1990
18.40	15.13	16.77	2,005	Strongly cross-pressured	1992
16.09	12.37	14.23	1,659	Strongly cross-pressured	1994
15.96	11.93	13.95	1,413	Strongly cross-pressured	1996
16.74	12.12	14.43	1,055	Strongly cross-pressured	1998
19.30	13.60	16.45	811	Strongly cross-pressured	2000
19.11	14.06	16.58	1,006	Strongly cross-pressured	2004
17.02	11.51	14.27	954	Strongly cross-pressured	2008
13.10	10.73	11.92	5,203	Strongly cross-pressured	2012
11.09	8.65	9.87	3,325	Strongly cross-pressured	2016
7.35	5.75	6.55	6,737	Strongly cross-pressured	2020
7.69	5.54	6.62	4,387	Strongly cross-pressured	2024
n/a	n/a	46.75	n/a	Implied by the correlation	1984
n/a	n/a	47.79	n/a	Implied by the correlation	1986
n/a	n/a	46.99	n/a	Implied by the correlation	1988
n/a	n/a	48.96	n/a	Implied by the correlation	1990
n/a	n/a	45.02	n/a	Implied by the correlation	1992
n/a	n/a	43.66	n/a	Implied by the correlation	1994
n/a	n/a	43.83	n/a	Implied by the correlation	1996
n/a	n/a	44.83	n/a	Implied by the correlation	1998
n/a	n/a	46.26	n/a	Implied by the correlation	2000
n/a	n/a	44.25	n/a	Implied by the correlation	2004
n/a	n/a	42.69	n/a	Implied by the correlation	2008
n/a	n/a	39.47	n/a	Implied by the correlation	2012
n/a	n/a	37.21	n/a	Implied by the correlation	2016
n/a	n/a	32.71	n/a	Implied by the correlation	2020
n/a	n/a	32.80	n/a	Implied by the correlation	2024

The share of adults on opposite sides of the medians falls from 46.2% in 1984 to 30.0% in 2024 and stays within 3.3 points of the dashed line, the 46.8% in 1984 and 32.8% in 2024 that the correlation implies for two bell-shaped indices. From 2008 the observed share is 2.1 to 3.3 points below the implied share. The lower line counts only adults at least 0.5 standard deviations from each median, 18.2% in 1984 and 6.6% in 2024.Note: Each index is split at its weighted median within the study, so each index divides adults into two equal groups. The implied share is the value for two normally distributed (bell-shaped) indices with the observed correlation. It has no interval because it is a function of the correlation.Note: Strongly cross-pressured means at least 0.5 standard deviations right of the median on one index and at least that far left of it on the other, in standard deviations of the index within the study. A band of 0.25 gives 27.5% in 1984 and 15.9% in 2024, and a band of 0.75 gives 10.3% and 3.2%.Source: ANES Cumulative Data File, 1984 to 2024; weighted (VCF0009z); intervals are model based

A median split also counts as cross-pressured an adult who is barely across a median, which overstates how many hold strong views in opposite directions. A stricter definition puts a tolerance band around each median: an adult is strongly cross-pressured when the adult is at least 0.5 standard deviations right of the median on one index and at least that far left of it on the other. That share was 18.2% in 1984 and 6.6% in 2024, and a narrower band and a wider band give the same fall (see the notes under the figure). In 2024 70.0% of adults are on the same side of both medians. A median split also makes the two split groups equal in size, and the two consistent groups equal in size, so shares from the split cannot show whether cross-pressure runs more toward economic left with cultural right or the reverse.

In 2024, more adults sit in the corners than expected

Figure 7 divides the 4,387 adults with both indices in 2024 into cells 0.5 standard deviations wide on each index, measured from each index's median, with open-ended outer bins that collect every adult beyond 1.0 standard deviations. The corner cell for adults more than 1.0 standard deviations left of the median on both indices holds 8.2% of adults, and the corner cell for adults at least that far right on both holds 8.5%. Two bell-shaped indices with the observed correlation would put 6.4% of adults in each corner, and the 95% intervals of both observed corners lie above that reference. The reference uses the same open bins, so the bin width does not explain the difference.

Figure 7

In 2024 the two corner cells with both indices far left or far right hold more adults than the correlation implies

Share of adults in each cell (columns: economic index; rows: cultural index), in standard deviations from each index's 2024 median, ANES 2024; 95% intervals

0.0%9.0%

No estimate

Heat map of 2024 adults by economic index in columns and cultural index in rows. Shading is deepest in the top-right and bottom-left corner cells, where adults are far right or far left on both indices, with 8.5% and 8.2%. The top-left corner, where an adult is far left on the economic index and far right on the cultural index, is light at 1.4%. The bottom-right corner, far right on the economic index and far left on the cultural index, has no estimate because it has too few respondents.
Cultural	below −1	−1 to −0.5	−0.5 to 0	0 to 0.5	0.5 to 1	above 1
above 1	1.4%, interval 1.0% to 2.0%, n = 47	1.0%, interval 0.7% to 1.5%, n = 39	1.9%, interval 1.4% to 2.6%, n = 70	2.5%, interval 1.9% to 3.2%, n = 116	3.6%, interval 2.9% to 4.5%, n = 144	8.5%, interval 7.3% to 9.8%, n = 416
0.5 to 1	0.9%, interval 0.6% to 1.4%, n = 37	1.2%, interval 0.8% to 1.8%, n = 49	1.7%, interval 1.2% to 2.4%, n = 64	3.3%, interval 2.6% to 4.1%, n = 114	2.0%, interval 1.5% to 2.7%, n = 93	4.0%, interval 3.2% to 4.9%, n = 213
0 to 0.5	2.2%, interval 1.6% to 2.9%, n = 96	3.0%, interval 2.4% to 3.8%, n = 118	3.3%, interval 2.6% to 4.1%, n = 141	4.8%, interval 3.9% to 5.8%, n = 205	4.1%, interval 3.3% to 5.0%, n = 158	4.0%, interval 3.3% to 5.0%, n = 200
−0.5 to 0	3.0%, interval 2.4% to 3.9%, n = 127	2.4%, interval 1.9% to 3.2%, n = 105	3.4%, interval 2.7% to 4.2%, n = 150	3.4%, interval 2.7% to 4.3%, n = 155	1.8%, interval 1.3% to 2.4%, n = 71	1.8%, interval 1.3% to 2.5%, n = 78
−1 to −0.5	2.7%, interval 2.1% to 3.5%, n = 102	2.3%, interval 1.8% to 3.1%, n = 112	2.1%, interval 1.6% to 2.8%, n = 91	2.5%, interval 1.9% to 3.2%, n = 121	0.8%, interval 0.5% to 1.3%, n = 35	0.7%, interval 0.4% to 1.2%, n = 31
below −1	8.2%, interval 7.1% to 9.5%, n = 353	5.4%, interval 4.5% to 6.5%, n = 244	3.6%, interval 2.8% to 4.5%, n = 148	1.9%, interval 1.4% to 2.6%, n = 106	No estimate	No estimate

Show data

Data for In 2024 the two corner cells with both indices far left or far right hold more adults than the correlation implies
Cultural index (SD from median)	Economic index (SD from median)	Share of adults (%)	Interval low	Interval high	Normal reference (%)	Observed minus reference (points)	Respondents
below −1	below −1	8.2	7.1	9.5	6.4	+1.8	353
below −1	−1 to −0.5	5.4	4.5	6.5	3.5	+1.9	244
below −1	−0.5 to 0	3.6	2.8	4.5	3.0	+0.6	148
below −1	0 to 0.5	1.9	1.4	2.6	1.8	0.0	106
−1 to −0.5	below −1	2.7	2.1	3.5	3.5	−0.8	102
−1 to −0.5	−1 to −0.5	2.3	1.8	3.1	3.1	−0.8	112
−1 to −0.5	−0.5 to 0	2.1	1.6	2.8	3.4	−1.3	91
−1 to −0.5	0 to 0.5	2.5	1.9	3.2	2.6	−0.2	121
−1 to −0.5	0.5 to 1	0.8	0.5	1.3	1.5	−0.6	35
−1 to −0.5	above 1	0.7	0.4	1.2	0.8	−0.1	31
−0.5 to 0	below −1	3.0	2.4	3.9	3.0	+0.1	127
−0.5 to 0	−1 to −0.5	2.4	1.9	3.2	3.4	−0.9	105
−0.5 to 0	−0.5 to 0	3.4	2.7	4.2	4.3	−1.0	150
−0.5 to 0	0 to 0.5	3.4	2.7	4.3	4.0	−0.6	155
−0.5 to 0	0.5 to 1	1.8	1.3	2.4	2.6	−0.9	71
−0.5 to 0	above 1	1.8	1.3	2.5	1.8	0.0	78
0 to 0.5	below −1	2.2	1.6	2.9	1.8	+0.4	96
0 to 0.5	−1 to −0.5	3.0	2.4	3.8	2.6	+0.4	118
0 to 0.5	−0.5 to 0	3.3	2.6	4.1	4.0	−0.7	141
0 to 0.5	0 to 0.5	4.8	3.9	5.8	4.3	+0.5	205
0 to 0.5	0.5 to 1	4.1	3.3	5.0	3.4	+0.7	158
0 to 0.5	above 1	4.0	3.3	5.0	3.0	+1.1	200
0.5 to 1	below −1	0.9	0.6	1.4	0.8	+0.1	37
0.5 to 1	−1 to −0.5	1.2	0.8	1.8	1.5	−0.2	49
0.5 to 1	−0.5 to 0	1.7	1.2	2.4	2.6	−0.9	64
0.5 to 1	0 to 0.5	3.3	2.6	4.1	3.4	−0.1	114
0.5 to 1	0.5 to 1	2.0	1.5	2.7	3.1	−1.1	93
0.5 to 1	above 1	4.0	3.2	4.9	3.5	+0.5	213
above 1	below −1	1.4	1.0	2.0	0.3	+1.1	47
above 1	−1 to −0.5	1.0	0.7	1.5	0.8	+0.2	39
above 1	−0.5 to 0	1.9	1.4	2.6	1.8	+0.1	70
above 1	0 to 0.5	2.5	1.9	3.2	3.0	−0.5	116
above 1	0.5 to 1	3.6	2.9	4.5	3.5	+0.1	144
above 1	above 1	8.5	7.3	9.8	6.4	+2.1	416

Two corners are dark, adults far left on both indices at the bottom left and far right on both at the top right, and the corner for economic left and cultural right at the top left is light. Each cell is the share of 2024 adults, and the second view shows each share minus the share for two bell-shaped indices with the observed correlation.Note: Columns run from the most liberal economic bin on the left to the most conservative on the right, and rows from the most conservative cultural bin at the top to the most liberal at the bottom. The middle cut of each axis is the median. The outer bins are open-ended and collect every adult beyond 1.0 standard deviations, so they are larger than the interior bins, which are 0.5 standard deviations wide. The normal reference uses the same bins.Note: Cells with fewer than 30 respondents are left blank: 2 cells, holding 0.5% of adults. Both are in the bottom-right corner region, where the cultural index is far left and the economic index is right of the median.Note: Bins are measured from the 2024 weighted median of each index, in units of that index's 2024 standard deviation.Source: ANES Cumulative Data File, 2024; weighted (VCF0009z); intervals are model based

The off-diagonal corners hold adults far to opposite sides of the two medians. The corner for economic left and cultural right holds 1.4% of adults, against 0.3% in the normal reference. 2 cells have fewer than 30 respondents and are left blank, and both lie where the cultural index is far left and the economic index is right of the median. Together they hold 0.5% of adults, less than the economic-left, cultural-right corner alone, so adults far to opposite sides are more often economic left with cultural right than the reverse in 2024.

The normal reference does not separate the way the two indices go together from the shape of each index. The indices average a few bounded, discrete 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 Chapter 7.3, Class, church and the two axes (planned) chapter takes up how income, education and church attendance sort the two axes.

What these data cannot show

The correlation is between two indices built in advance from a fixed set of items, so these data cannot show how many dimensions the public holds. A factor model or an item-response model fitted to a wider set of questions could find a different number, and a different choice of items would change the level of the correlation. The chapter also leaves out issues that fall outside both indices, such as immigration and race, which may form a third dimension.

The corrected correlation assumes that the errors in the two indices are unrelated to each other and that alpha measures reliability, which holds only if the items of an index measure the same attitude. It is an upper bound for the correlation between the underlying attitudes only if errors are also unrelated across the items of one index. A shared response format or response style would raise alpha, make the correction too small and remove the bound, and the corrected series inherits the ambiguity in what a rise in alpha means. The interval for the corrected value comes from resampling respondents within a study and is wide in the early studies, where the indices have two components and low alpha.

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 comparison within party identification uses party identification measured in the same interview and shaped by the same attitudes, so it does not measure an effect of party. Interview mode changed over the period, and the interview-mode comparison covers only 4 studies. The intervals are model based and ignore the ANES clustering and stratification. The 2002 study has no cultural index, and 4 studies have two components in one or both indices.

A comparison with the Cooperative Election Study is not included. Its economic index contains only health-care questions and its cultural index mixes abortion, guns, immigration and the environment, so a higher correlation there would confound item content with survey. A matched comparison needs an index built from the same kinds of questions in both surveys.

A rising correlation fits an account in which the public has moved toward the alignment that the parties present, and the party gaps in Figure 3 are consistent with it, but nothing in these data tests that account.

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

Methods

Sample: respondents in the ANES Cumulative Data File with a positive weight and at least two of three items on each index, 15 studies from 1984 to 2024 (33,257 respondents). The 2002 study has no cultural index, since only the gay and lesbian thermometer was asked. Indices: each item is oriented so that a higher value is more conservative (the direction comes from the value labels of the Cumulative Data File, see the items table, never from a correlation), converted to a z-score with the weighted mean and standard deviation over all ANES years in which it was asked, and averaged over the items the respondent answered (the share of adults whose index averages two answers is the weighted share of a study's respondents who answered two of the three items). The economic index uses guaranteed jobs (VCF0809), services and spending (VCF0839) and health insurance (VCF0806); the cultural index uses abortion (VCF0838), the gay and lesbian thermometer (VCF0232) and the traditional-values item (VCF0853). The cleaned ANES file stores the Cumulative Data File thermometer code 97 (97 to 100 degrees) as 100. An index is the mean of the standardized items a respondent answered; the correlation does not depend on its unit, and each figure names its own yardstick: the party means and the self-placement regression use the index re-standardized to mean 0 and standard deviation 1 over all respondents in the analysis sample, and the cross-pressure band and the 2024 joint distribution use the standard deviation of the index within the study. Estimand: the survey-weighted Pearson correlation of the two indices within each study, weight VCF0009z. Net of party: both indices are residualized on the party identification categories within the study (weighted group means; three categories, and seven points with leaners kept apart, VCF0301) and the residuals are correlated; the Fisher z interval uses the Kish effective sample size less the number of party dummies. Reliability: standardized Cronbach alpha from the weighted mean inter-item correlation among respondents who answered every item of the index in that study (respondents with two of three items enter the correlation but not alpha, so alpha overstates their reliability); the corrected correlation divides the observed correlation by the square root of the product of the two alphas and is shown where both alphas are at least 0.4; intervals for alpha and the corrected correlation are 95% percentile intervals from 1000 bootstrap resamples of respondents within each study (each resample draws respondents with replacement and recomputes the estimate; seed 12345 plus the study year). Cross-pressure: an adult is cross-pressured when right of the median on one index and left of it on the other; each index is split at its weighted median within the study, respondents tied at the median are divided so that each side holds exactly half of the weighted population, and the tolerance band is 0.5 standard deviations of the index within the study around the median; the implied share is 50% minus arcsin(r) divided by pi, the value for two normally distributed (bell-shaped) indices with correlation r. Party means use party identification with leaners counted with their party and are expressed in standard deviations of the pooled analysis sample; the line for all adults is the weighted mean of every respondent in the analysis sample, independents without a lean included; the common-item check on the cultural party means and gap uses abortion and the thermometer only, which were asked in every study except 1986 and 1990. The reference for each cell of the 2024 joint distribution is the share of a bivariate normal distribution with the observed correlation, centred at the medians, inside the same bin; the second view of the figure subtracts it from the observed share and from both ends of the observed interval. Self-placement: weighted least squares of seven-point liberal-conservative self-placement (VCF0803) on both standardized indices with study fixed effects and HC3 errors, for 1984 to 1996 and 2012 to 2024; the rise in a coefficient between the two periods is the difference of the two coefficients with a normal interval from the two HC3 standard errors, treating the periods as independent samples. The slope of the standardized cultural index on the standardized economic index equals their correlation. Interview mode: face-to-face and web estimates use weight_ftf and weight_web, which exist for 2008, 2012, 2016 and 2024; web interviews began in 2012 and the 2020 study mixes modes without mode weights. Intervals: 95% Fisher z intervals on the Kish effective sample size, normal intervals for shares (Wilson intervals for cells of the joint distribution); all are model based and ignore clustering and stratification. Suppression: a study-level cell needs at least 150 respondents and an effective sample of 100; a party-by-study cell needs 100 and 100; a cell of the joint distribution needs 30 respondents.

References

Feldman, Stanley; Johnston, Christopher (2014). Understanding the Determinants of Political Ideology: Implications of Structural Complexity. Political Psychology 35(3): 337-358. doi:10.1111/pops.12055
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
Carmines, Edward G.; Ensley, Michael J.; Wagner, Michael W. (2012). Who Fits the Left-Right Divide? Partisan Polarization in the American Electorate. American Behavioral Scientist 56(12): 1631-1653. doi:10.1177/0002764212463353
Poole, Keith T.; Rosenthal, Howard (2007). Ideology and Congress: A Political Economic History of Roll Call Voting. Transaction Publishers (second edition).
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
Fisher, Ronald A. (1921). On the Probable Error of a Coefficient of Correlation Deduced from a Small Sample. Metron 1(4): 3-32.
Kish, Leslie (1965). Survey Sampling. John Wiley and Sons.
Cronbach, Lee J. (1951). Coefficient Alpha and the Internal Structure of Tests. Psychometrika 16(3): 297-334. doi:10.1007/BF02310555
Spearman, Charles (1904). The Proof and Measurement of Association between Two Things. American Journal of Psychology 15(1): 72-101. doi:10.2307/1412159

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.}
}

Reading paths

The core argument in eight chapters