Methodology · The presidential record · presidents-v1

How the presidential record is measured.

The presidential record shows how 21 published measures of the U.S. economy moved while each president since 1953 was in office. It describes what happened. It does not explain it, rank anyone or reach a conclusion.

The measures, formulas, windows and the treatment of gaps were written down and committed to Palanor’s code repository on 2 November 2026, before any figure by party was computed. Before that, while the work was being scoped on 1 November 2026, payroll jobs, the CPI, the unemployment rate, production-worker real hourly pay, federal debt as a share of GDP and payrolls with a one-year lag had been computed by presidency. Any later change is a new, dated version that says what changed and why.

What it can and cannot say

• Correlation, not cause. A figure shows what happened during a term, not what a president caused.
• Presidents don’t act alone. Congress writes budgets and laws, the Federal Reserve sets interest rates, and the world economy moves on its own. The page lists the House and Senate majorities and the Fed chair for every presidency.
• Lags differ by measure. Interest rates and markets move within days, payrolls within months, the budget within a fiscal year or more, and poverty and household income are measured once a year. That is why a second window, starting one year later, is always shown beside the first.
• Shown as published. No adjustment is made for wars, oil shocks, recessions or the pandemic, for anyone. Recessions as the NBER dates them are shaded on the charts.
• No measure has a single agreed good direction. Higher rates help savers and cost borrowers; a trade deficit is the mirror of capital flowing in. No figure is coloured as good or bad.

The window rule

A calendar month belongs to whoever was president on its 1st. A presidency’s change runs from the reading of the last month before its first month to the reading of its last month: Eisenhower from January 1953 to January 1961, Kennedy from January 1961 to November 1963, Johnson from November 1963 to January 1969, Nixon from January 1969 to August 1974, Ford from August 1974 to January 1977. Years in office are those months divided by 12.

• Quarterly series use the quarter that holds each boundary month; averages use the quarters after the starting one through the last one.
• Annual calendar-year series are read as year-end values: a boundary month from January to June uses the year before, July to December that year.
• Federal budget measures use fiscal years. A fiscal year belongs to whoever was president on its first day (1 October; 1 July before fiscal 1977). A new president’s first budget year therefore belongs to the predecessor, and a departing president keeps the fiscal year running when they leave. Only completed fiscal years count.
• Starting one year later: every boundary moves 12 months later (fiscal and calendar years by one, quarters by four). Each term’s last year in this window is its successor’s first.
• The open term, and any window that runs past the latest reading, stops at the latest published reading and is marked partial.
• Gaps: a figure is shown only when the series covers both ends of the window. Otherwise it reads “no data”. Nothing is estimated, spliced or interpolated.

The party view

One figure per party for each measure and window, weighted by time in office, with the years it covers. Changes per year are the sum of the presidencies’ changes divided by the sum of their years; compound rates are the product of their ratios raised to one over the sum of their years; averages are the mean of every period attributed to the party’s presidencies. The open term counts with its partial years. There is no count of measures, no overall figure and no ordering of the parties. Party is the president’s party, as a plain label.

The measures

Coverage and limits by measure

• Real hourly pay starts in January 1964, so Eisenhower, Kennedy and Johnson read “no data”. It uses the CPI-U, not the CPI-W of BLS’s own real earnings release, so it matches the consumer price measure.
• Real disposable income per person starts in 1959, real median household income in 1984, the official poverty rate in 1959, and the 30-year mortgage rate in April 1971.
• Gasoline after inflation divides the gasoline price index by the all-items index, both not seasonally adjusted, because the adjusted gasoline index starts only in 1967. Boundaries in months other than January carry some seasonal movement.
• Real job growth subtracts the growth of the civilian population aged 16 and over, the population the jobs survey covers. BLS resets that population each January without smoothing earlier months, so it steps in some Januaries.
• Federal debt added counts total public debt, including what one part of the government owes another. Today’s dollars restate each fiscal-year-end level with the CPI-U, so a figure below zero means the debt was worth less in today’s dollars at the end of the window than at the start.
• Real S&P 500 total return uses Aswath Damodaran’s year-end series (NYU Stern, histretSP), which includes dividends, deflated by the December CPI-U. Year-end readings mean a presidency’s window starts at the year-end nearest its first month.
• For 2013 and 2017 Census lists two poverty figures; the first one Census lists is used, as in its own national series.
• Removed from the original plan: the share without health insurance. No consistent national series spans enough presidencies (the survey questions were redesigned in 1987, 1999, 2013 and 2017, and the ACS series starts in 2008), and no splice is made.

Context columns

House and Senate majorities come from the House party divisions and the Senate party division tables, including the Senate’s three changes of control in 2001. Federal Reserve chairs come from the Board’s membership list.

Pages for each president

Each presidency has its own page. Beside each per-year figure it shows the median of the completed presidencies since 1953 that have a figure in that window (the term under way and partial windows are left out), a reference line and not a rank, and the predecessor’s and successor’s figures. Its four charts show the same window for every president: from 1 January three years before the term to 31 December three years after it. Terms come from one table of heads of government, built so that other countries can be added.

Elections

Results by state for every presidential election since 1952. Votes for 1976–2020 come from the MIT Election Data + Science Lab (Harvard Dataverse, released under CC0 1.0); votes for 1952–1972 and 2024 from the Clerk of the House’s Statistics of the Presidential and Congressional Election; electoral votes from the National Archives. A candidate’s share counts every ballot line for them (such as New York’s Liberal and Conservative lines, which the Clerk lists as separate national columns) over all votes cast for president. Counting only the lines the Clerk labels Democratic or Republican, the state rows match the Clerk’s national totals exactly. New York’s 1968 Liberal and Courage lines count for Humphrey and Wallace, as the state’s own 1968 Electoral College proceedings name them; its 1972 Conservative and Liberal lines are shown as the Clerk prints them, without a candidate, because no primary source naming them could be reached. The state map shades each state by the election winner’s margin over the next candidate on a single grey scale keyed to that winner, never party colours. Johnson (1963) and Ford took office by succession, with no election for that start.

Approval over the term

The Palanor approval average (method approval-v2, unchanged), read on the last day of each month in office. Before 2017 the rule registered in approval-v3-history applies to every president alike: a month gets a line only when at least three independent pollsters published a national poll in the 30 days before it, the average’s own minimum and window; in other months each poll is shown as a dot. Each poll’s figures come from its pollster’s own release (or the Internet Archive copy of it), linked per poll. Gallup, YouGov and Ipsos polls are never used, and neither is any aggregator, so terms before the 1990s have few or no polls. Approval is opinion, not an economic outcome, and is not part of the record’s measures.

Presidents on long charts

Any U.S. chart covering eight years or more can show the presidential terms (Show → U.S. presidents): two alternating light greys by the president’s number, the surname along the top edge, the name and dates on hover. With recessions shaded too, terms become thin dividers. It changes the picture only; the figures and data files are the same with it on or off.

Where it appears

On palanor.com/presidents and the Terminal only. It is not used by any automated writing, social post or forecast: the page and its charts are kept out of Palanor Foresight, Studio, the Finding Desk, The Read and the Council, and Numen states its figures by measure without saying who did better overall.

Sources: U.S. Bureau of Labor Statistics; U.S. Bureau of Economic Analysis; U.S. Census Bureau; U.S. Department of the Treasury, Bureau of the Fiscal Service; Congressional Budget Office; Office of Management and Budget via FRED (Federal Reserve Bank of St. Louis); Freddie Mac; Aswath Damodaran, NYU Stern (histretSP).