Research manuscript · Science and political trust

Science Confidence and Ideology

More Than Relabeling: Ideology and Confidence in Science, 1975–2024

In the 1970s, Americans identifying as conservative expressed comparatively high confidence in the scientific community. By the 2020s, that relationship had reversed. This study asks whether the shift simply reflects who now calls themselves conservative—or whether evaluations of scientific institutions changed within political groups.

Explore the project →All research projects

PUBLIC TRUST · IDEOLOGY · 1975–2024

THE IDEOLOGICAL REVERSAL

What changed across five decades?

≈0.39 SDOverall movement in the standardized gap

31General Social Survey waves

41,438Respondents in the analyzed series

These are the two reported period estimates, not a line interpolated through unobserved annual values. The interpretation follows the manuscript’s standardized outcome coding.

Reported standardized contrasts from the science-confidence manuscript. The two periods are shown without inventing intervening annual estimates.

RESEARCH IN DEPTH

What the reversal rules out—and what it does not

Beyond the abstract: the argument, study design, evidence, and limits of inference.

01 / RESEARCH

Sorting versus changing evaluations

The paper uses 31 General Social Survey waves, 1975–2024, with 41,438 respondents. Its key outcome asks about confidence in the people running the scientific community, not whether people believe a particular scientific fact. Political categories may acquire different members over decades. To test that account, the analysis allows party affiliation and moral conservatism to have period-specific associations with institutional confidence instead of assuming these relationships never change.

02 / RESEARCH

What the adjustments reveal

Changes in partisan and moral composition account for roughly two-fifths of the long historical movement, leaving a significant ideological trend. The conservative–moderate gap also changes within birth-decade cohorts observed across multiple survey periods, which a story based solely on one generation replacing another cannot explain. Elevated recent contrasts are visible in 2018, 2022, and 2024 rather than only during one pandemic survey.

03 / RESEARCH

The measurement problem

The paper compares the scientific-community confidence item with support for government research spending and frontier research. Those questions ask respondents to evaluate different objects—an institution, a spending priority, and public investment—and they show different ideological structures. Tests of specific formative crises and rival interpretations do not yield a robust cause of the remaining reversal. That uncertainty is reported rather than obscured.

Research basis: More Than Relabeling unpublished GSS manuscript; statistical results described as manuscript findings.

VERIFIED RESEARCH DETAILS

The estimates behind the reversal

These are design-based estimates from the GSS manuscript. A positive gap indicates conservatives report less confidence than moderates.

Survey year	Respondents	Conservative–moderate gap (SD)	SE
2016	1,753	+0.162	0.081
2018	1,384	+0.226	0.072
2021	2,285	+0.254	0.063
2022	2,010	+0.327	0.077
2024	1,881	+0.260	0.062

The pooled late-1970s estimate is −0.113 SD; the 2020–24 estimate is +0.274 SD. The recent increase is large, but a separate curvature test is not significant (p = .18), so acceleration is not established.

Within birth cohorts

For respondents born in the 1940s, the modeled gap moves from −0.075 (1975–89) to +0.296 (2015–24). The repeated cross-sections do not follow the same people.

Partisan and moral composition

On the common 1977–2024 sample (N = 17,703), adding the two time-varying adjustments reduces the ideology × year coefficient from +0.00898 to +0.00544 (p = .00027). This is not a causal decomposition.

What “science trust” measures

Among 7,834 respondents, the liberal contrast between general science-research spending and federally funded frontier research is −0.151 SD (p = .0002); conservatives show no detectable corresponding contrast (+0.028 SD, p = .51).

More Than Relabeling, current blinded manuscript; Tables 2, 3, 4 and 6. Unpublished.

Read critically

Scope, status, and limits.

Repeated survey cross-sections do not track the same individuals. Within-cohort differences describe group-level movement, and model adjustments cannot eliminate every unobserved historical change. This is an unpublished manuscript.

Research source

Unpublished research manuscript, 1975–2024 GSS analysis.