Agenda trends
Eight domains, one topic system, seventy-nine years. Because the press, the public, the parties, the president, Congress and the Court are all coded with the same codebook, the share of each one's attention going to a topic can be set directly beside the others. Pick a topic to see where attention to it rose and fell. Start with any two domains set side by side, then step back to the same topic across all eight.
Compare two domains for agenda priorities
Party platforms and Public (MIP) — Law and Crime as a share of each domain's agenda. The shaded band is the gap between them.
- 2000Public (MIP) gave Law and Crime the larger share — 31.6% against 10.0%.
- 1968Public (MIP) gave Law and Crime the larger share — 19.6% against 7.4%.
- 1988Public (MIP) gave Law and Crime the larger share — 18.3% against 7.1%.
A correlation here says whether two agendas rose and fell together, not whether one caused the other. Both series are shares, so a domain whose total volume grew can still show a falling share. Where the two spans differ, only their overlapping years enter the correlation.
One topic across every domain
Share of each domain's yearly agenda given to Law and Crime, 1945–2025. Panels run in the friction order the rest of this site uses — the signals arriving first, the authoritative outputs last.
All eight panels share one vertical scale, so their heights are directly comparable — a topic that takes a fifth of a State of the Union and a fiftieth of the Supreme Court's docket looks that way here. Each panel is annotated with its own peak year. Sources are the official US Policy Agendas data files; domains cover different spans, and each panel is labeled with its own.
What is being compared
Eight official US Policy Agendas series, 239,945 coded records in all. Every one is a count of items coded to a major topic, converted to that domain's yearly share — except Gallup's Most Important Problem, which is re-scaled so that its twenty policy topics sum to 100% of policy mentions. The New York Times Index series changes its coding in 2009: Immigration and Culture are not coded before then, so shares for those two topics are not comparable across that date.
| Domain | Unit of analysis | Years | Records |
|---|---|---|---|
| Media | front-page and index stories New York Times Index sample, 1946-2016 | 1946–2016 | 43,357 |
| Public (MIP) | share of Gallup respondents Gallup's Most Important Problem, 1947-2023 | 1947–2023 | share series |
| Party platforms | platform quasi-sentences (both parties pooled) Democratic and Republican Party Platforms, 1948-2024 | 1948–2024 | 36,029 |
| Presidential speeches | State of the Union quasi-sentences State of the Union addresses, 1946-2025 | 1946–2025 | 20,633 |
| Executive orders | executive orders Executive Orders, 1945-2025 | 1945–2025 | 4,768 |
| Congressional hearings | hearings Congressional Hearings, 1946-2021 | 1946–2021 | 103,076 |
| Public laws | public laws Public Laws, 1948-2022 | 1948–2022 | 21,968 |
| Supreme Court | cases Supreme Court Cases, 1901-2023 | 1946–2023 | 10,114 |
Built 2026-08-30 from the official files in the Comparative Agendas Project's own store, downloaded and aggregated automatically; nothing on this page is hand-entered. A year is dropped where a domain coded fewer than 25 records, and a pair is left blank where the two overlap for fewer than 15 years. The Supreme Court file carries a few records stamped with future years — those are coding artefacts and are excluded. Files, per domain, are linked from the data domains index; the codebook behind all of them is on Data & methods.
How these numbers are read
Two sets of cutoffs turn coefficients into words on this page, and both are this site's convention rather than a disciplinary standard — there is no agreed one. They are printed here so a reader can disagree with them, and so nobody mistakes a wording choice for a finding.
- ≥ 0.7two domains “move closely together” (absolute value; a negative r is reported as moving in opposite directions)
- ≥ 0.45two domains “move together”
- ≥ 0.25two domains “move loosely together”
- ≥ 0.1two domains “barely track each other”
- < 0.1two domains “do not track each other”
- chainsA chain counts as one where the correlation does not carry when both links reach r ≥ 0.35 and the two ends fall at least 0.2 below the weaker link.
- coverageA year is dropped below 25 coded records, and a pair is left blank below 15 overlapping years.
Both terms are defined for readers meeting them for the first time in the student glossary — correlation and transitivity, each with a worked example from this page.
Public search interest is shown separately, on public attention.