Methodology & Caveats

Sources, definitions, limits

Author

Sam Caldwell

What this analysis measures

Three loosely-related signals that get conflated in casual talk about “sentiment toward a president”:

Section Asks Source Refresh
Approval (page) “Do you approve of the job President X is doing?” Gallup aggregate (hand-curated) On demand
Economic (page) “How do you feel about the economy?” U. Michigan Consumer Sentiment (FRED UMCSENT) Daily
Media (page) “What tone is the news media using to describe current events?” Not yet implemented —

These measure different things and often diverge. A president can enjoy strong consumer sentiment (good economy) while having weak approval (wars, scandals, polarization) — or vice versa.

Approval — how it’s collected

  • Gallup surveys roughly 1,000 US adults at a time by phone, asking the job approval question worded identically since 1938.
  • The values here are term averages computed by Gallup from all polls taken during each president’s term.
  • These aggregates are facts in the Feist v. Rural sense and are redistributable; exact poll-by-poll microdata is not.
  • Stored in data/sentiment/gallup_approval.csv. Values are hand-entered from Gallup’s published summaries and updated when the user refreshes them — not automatically.

Why hand-curated? No free programmatic interface exists for Gallup’s historical series. Modern polling APIs (FiveThirtyEight’s old feed, Pew, etc.) either no longer exist, are paywalled, or restrict redistribution.

Economic sentiment — how it’s collected

  • U. Michigan’s Surveys of Consumers asks households five questions about: personal finances (two), business conditions (two), and major-purchase attitudes (one). Responses are indexed against a 1966 baseline of 100.
  • Published monthly; available from FRED as series UMCSENT with full history back to 1952.
  • The value is computed from consumer responses — it’s what the public feels about the economy, not what economists model.

Administration attribution

Each monthly observation is attributed to whichever president was in office on the 14th of that month — same convention used in the Presidential Economies analysis. Inaugurations on January 20 mean the outgoing president keeps January of transition years.

Baselines

“Baseline” in this report = mean across the 1999-to-present window of the administrations actually displayed. It’s a simple reference line, not a statistical null.

Why you should not read these as president scorecards

All six caveats from the Presidential Economies methodology apply here, plus:

  1. Approval has a rally effect. Presidents get bumps during crises regardless of performance (Bush 43 hit 90% after 9/11 without changing any policy). Term averages smooth some of this but not all.

  2. Economic sentiment is dominated by inflation perception and gas prices. Multiple studies find that household feelings about the economy track gas prices and grocery-aisle inflation more than GDP or unemployment. Whoever happens to be in office during an inflation episode inherits bad sentiment that is not primarily about them.

  3. Polarization. Partisan sort over the past 25 years means that a large fraction of the population gives the president of their own party ~90% approval and the other party ~5%, regardless of events. This compresses every modern president’s possible approval range — Trump I was the first president to never hit 50% because partisan floors and ceilings have tightened, not because he was uniquely disliked.

  4. Event-date attribution is fuzzy. The world-events list is editorial — a non-exhaustive set of high-salience moments chosen for illustration, not a statistical fixture.

Data sources

Public-domain (safe to redistribute)

Aggregate (fact-like, redistributable)

Editorial

  • World-events list: hand-curated by the site author, maintained in data/sentiment/events.csv. Corrections welcome via the repository.

Code license

MIT — see the LICENSE file in the repository.

Disclaimer

Nothing in this analysis constitutes political, investment, or financial advice. Aggregates shown are descriptive summaries of publicly-available historical data; they say nothing about future performance or about the policies that produced them.