US GDP per capita with presidents and wars

Description

It is commonly claimed that Franklin Roosevelt (FDR) did not end the Great Depression: World War II (WW2) did. This is supported by the 10.6 percent growth per year in real Gross Domestic Product (GDP) per capita seen in the standard GDP estimates from 1940 to 1945. It is also supported by the rapid decline in unemployment during the war.

However, no comparable growth spurts in GDP per capita catch the eye in a plot of log(GDP per capita) from 1790 to 2024, whether associated with a war or not, using data from Measuring Worth. The only other features of that plot that seem visually comparable are the economic disaster of Herbert Hoover's presidency (when GDP per capital fell by 10 percent per year, 1929-1932), the impressive growth of the US economy during the first seven years of Franklin Roosevelt's presidency (6.4 percent per year, 1933-1940), and the post-World War II recession (when GDP per capita fell by 7.9 percent per year, 1945-1947). (NOTE: The web site for Measuring Worth, https://measuringworth.com/ still works, but has not always been maintained to current internet security standards. Therefore, the link is provided here in text but not as a link.)

Closer inspection of this plot suggests that the US economy has generally grown faster after FDR than before. This might plausibly be attributed to "The Keynesian Ascendancy 1939-1979".

Unemployment dropped during the First World War as it did during WW2. Comparable unemployment data are not available for the U.S. during other major wars, most notably the American Civil War and the Mexican-American War.

This data set provides a platform for testing the effects of presidency, war, and Keynes. It does this by combining the numbers for US population and real GDP per capital dollars from Measuring Worth with the presidency and a list of major wars and an estimate of the battle deaths by year per million population. (As noted above, the web address for measuring worth, https://measuringworth.com/, often gives security warnings but still seems to provide the data as before.)

Usage

data(USGDPpresidents)

Format

A data.frame containing 259 observations on the following variables:

Year

integer: the year, c(seq(1610, 1770, 10), 1774:2024)

CPI

Numeric: U. S. Consumer Price Index per Officer and Williamson (2022), starting in 1774. Average 1982-84 = 100.

GDPdeflator

numeric: Implicit price deflators for Gross Domestic Product with 2017 = 100 per Williamson (2025).

population.K

integer: US population in thousands.

Population figures for 1610 to 1780 came from Springston (2013). The rest came from Johnston and Williamson. (The early population figures reflect only the European settlers in the British colonies that eventually became the US.)

realGDPperCapita

numeric: real Gross Domestic Product (GDP) per capita in 2017 dollars since 1790.

Real GDP = population.K*realGDPperCapita, in thousands.

Current or nominal GDPperCapita = realGDPperCapita*GDPdeflator/100.

executive

ordered: Crown of England through 1774, followed by the "ContinentalCongress" and the "ArticlesOfConfederation" until Washington, who became President under the current base constitution in 1789. Two nineteenth century presidents are not listed here (William Henry Harrison and James A. Garfield), because they died so soon after inauguration that any contribution they made to the economic growth of the nation might seem too slight to measure accurately in annual data like this; their contributions therefore appear combined with their replacements (John Tyler and Chester A. Arthur, respectively). The service of two other presidents is officially combined here: "Taylor-Fillmore" refers to the 16 months served by Zachary Taylor with the 32 months of Millard Fillmore. These modifications make Barack Obama number 41 on this list, even though he's the 44th president of the U.S.

war

ordered: This lists the major wars in US history by years involving active hostilities. A war is "major" for present purposes if it met two criteria:

(1) It averaged at least 10 battle deaths per year per million US population.

(2) It was listed in one of two lists of wars: For wars since 1816, it must have appeared in the Correlates of War. For wars between 1790 and 1815, it must have appeared in the Wikipedia "List of wars involving the United States".

The resulting list includes a few adjustments to the list of wars that might come readily to mind for people moderately familiar with US history.

A traditional list might start with the American Revolution, the War of 1812, the Mexican-American war, the Civil War, the Spanish-American war, World Wars I and II, Korea, and Vietnam. In addition, the Northwest Indian War involved very roughly 30 battle deaths per year per million population 1785-1795. This compares with the roughly 100 battle deaths per year 1812-1815 for the War of 1812.

For present purposes, the Spanish-American War is combined with the lesser-known American-Philippine War: The latter involved 50 percent more battle deaths but over a longer period of time and arguably with less impact on the stature of the US as a growing world power. However, its magnitude suggest it might have impacted the US economy in a way roughly comparable to the Spanish-American war. The two are therefore listed here together as "Spanish-American-Philippine" war.

The Correlates of War (COW) data include multiple US uses of military force during the Vietnam War era. It starts with "Vietnam Phase 1", 1961-65, with 506 battle deaths in the COW data base. It includes the "Second Laotian" war phases 1 and 2, plus engagement with a "Communist Coalition" and Khmer Rouge as well as actions in the Dominican Republic and Guatemala. The current data.frame includes only "Vietnam", referring primarily to COW's "Vietnam War, Phase 2", 1965-1973. The associated battle deaths include battle deaths from these other, lesser concurrent conflicts.

The COW data currently ends in 2007. However, the post-2000 conflicts in Afghanistan and Iraq averaged less than 1,000 battle deaths per year or roughly 3 battle deaths per year per million population. This is below the threshold of 10 battle deaths per year per million population. This in turn suggests that any impact of those conflicts on the US economy might be small and difficult to estimate.

battleDeaths

numeric: Numbers of battle deaths by year estimated by allocating to the different years the totals reported for each major war in proportion to the number of days officially in conflict each year. The totals were obtained (in August-September 2015) from The Correlates of War data for conflicts since 1816 and from Wikipedia for previous wars back to 1774, as noted above.

battleDeathsPMP

numeric: battle deaths per million population = 1000*battleDeaths/population.K.

Keynes

integer taking the value 1 between 1939 and 1979 and 0 otherwise, as suggested by the section entitled "The Keynesian Ascendancy 1939-1979" in the Wikipedia article on John Maynard Keynes.

unemployment

Estimated US unemployment rate in percent per the sources cited in the Historical unemployment rate charts section of the Wikipedia article on Unemployment in the United States .

unempSource

ordered giving the source for US unemployment:

1610-1799

<NA>

1800-1889

Lebergott

1890-1929

Romer

1930-1939

Coen

1940-present

BLS

Clearly, the more recent numbers should be more accurate.

fedReceipts, fedOutlays, fedSurplus

Receipts and Outlays of the US federal government in millions of current dollars.

For data beginning with 1901, these are from the US federal budget from The White House (2022). Earlier data are from series Y 335-337 in US Census Bureau (1975). As of 2022-02-22 the data from The White House included aggregations for 1789-1849 and 1850-1900, which matched the totals of Y 335-337 for those two sets of years. The numbers from 1901 to 1933 are the same in both sources.

On 2022-02-22, we used The White House (2022) for the more recent numbers with one exception: Between 1976 and 1977 the fiscal year was changed from starting July 1 to October 1. July, August, and September, 1976, is called the "transitional quarter", and has been deleted from this dataset.

On 2025-01-22, the comparable data was found on a different web address but seemed to be largely the same except that it said it was updated to 2023-03-28, just over a month after the previous update. That gave data starting in 2022 as "estimates". Since we could not find better numbers, we used those.

NOTES:

The numbers for 1843 are for only the first half of the year, January 1 through June 30. This explains why the numbers for 1843 are only roughly half of the corresponding values for 1844 and 1845.

Also, the numbers for 1791 are actually for 1789-1791. However, those numbers seem comparable to those for 1792 and 1793, so it is listed as only for one year rather than three.

fedDebt

US federal government debt in current dollars per FiscalData (2025). This matches Y 338 in United States Census Bureau (1975) 1921-1939 but not earlier, and Y 338 ends with 1939. Between 1921 and 1939 these numbers are as of June 30. Between 1843 and 1920 they are as of July 1. The earlier numbers are as of January 1.

These numbers are NOT a simple cumsum of fedSurplus. It's not clear the sources of the discrepancies, but the following seem likely: First, the published budget may not include interest on the national debt. Second, there are likely expenditures / outlays that are "Off-budget" and perhaps otherwise hidden from the public.

FiscalData (2025) includes debt for both January 1 (20 million) and July 1 (33 million) for 1843. For present purposes, we omit the January 1 number. This overstates the volatility of the national debt during that period, showing it rising from 14 million in 1842 (January 1) to 33 million in 1843 (July 1), being 18 not 12 months. The alternative would be to delete the 33 million, but that would understate the volatility of the debt during that period.

Prior to 1844, the national debt was reported for January 1. Between 1843 and 1920, it was reported for July 1. From 1921 to 1976, it was reported for June 28, 29 or 30, usually for June 30; four years in that period it was reported for June 29 and once for June 28. Since 1977, it has been reported for September 28, 29 or 30, usually for September 30, only once for September 28 and once for September 29.

fedReceipts_pGDP, fedOutlays_pGDP, fedSurplus_pGDP, fedDebt_pGDP

numeric = fedReceipts, fedOutlays, fedSurplus, and fedDebt as a portion of GDP, which is population.K * realGPDperCapita / (.01*GDPdeflator).

For the single year 1843, fedReceipts, fedOutlays, and fedSurplus were for only the first six months; to compute *_pGDP for these numbers for 1843 only, the denominator was cut in half.

Details

rownames(USGDPpresidents) = Year

Author(s)

Spencer Graves

Source

Robert M. Coen (1973) Labor Force and Unemployment in the 1920's and 1930's: A Re-Examination Based on Postwar Experience", The Review of Economics and Statistics, 55(1): 46-55.

FiscalData (2024) "Historical Debt Outstanding", accessed 2025-01-22.

Louis Johnston and Samuel H. Williamson, "What Was the U.S. GDP Then?", Measuring Worth, accessed 2022-02-22. (NOTE: This came from https://www.measuringworth.org/usgdp/. this web link generally works as of 2022-02-22. However, in the past it has sometimes returned a warning, e.g., "SSL certificate problem". The web site seems to be good but not maintained to current security standards.)

Stanley Lebergott (1964). Manpower in Economic Growth: The American Record since 1800. Pages 164-190. New York: McGraw-Hill. Cited from Wikipedia, "Unemployment in the United States", accessed 2016-07-08.

Lawrence H Officer and Samuel H. Williamson, 'The Annual Consumer Price Index for the United States, 1774-Present,' MeasuringWorth, 2022-02-22.

Christina Romer (1986). "Spurious Volatility in Historical Unemployment Data", The Journal of Political Economy, 94(1): 1-37.

Sarkees, Meredith Reid; Wayman, Frank (2010). "The Correlates of War Project: COW War Data, 1816 - 2007 (v4.0)", accessed 2015-09-02.

The White House (2022). On 2025-08-22, I saw "Historical Tables" in 3 places on that page. I clicked on the bottom one and got, "BUDGET-2026-HIST.xlsx". The file I got doing this on 2025-01-22 included "Table 1.1-Summary of Receipts, Outlays, and Surpluses or Deficits (-): 1789-2026" included budget forecasts. The version of this table I got 2025-08-22 included data through 2024 but no forecasts.

United States Census Bureau (1975) Bicentennial Edition: Historical Statistics of the United States, Colonial Times to 1970, Part 2. Chapter Y. Government, accessed 2022-02-22.

Wikipedia, "List of wars involving the United States", accessed 2015-09-13.

Wikipedia, "Unemployment in the United States". See also https://en.wikipedia.org/wiki/User_talk:Peace01234#Unemployment_Data. Accessed 2016-07-08.

The unemployment data since 1940 are from series LNS14000000 from the Current Population Survey. These data are available as a monthly series from the Current Population Survey of the Bureau of Labor Statistics.

Chuck Springston, "Population of the 13 Colonies 1610-1790", October 28, 2013

Samuel H. Williamson, "What Was the U.S. GDP Then?" MeasuringWorth, 2025

Examples

data(USGDPpresidents)
##
## FUNCTION TO plot GDP, Presidents and Wars 
## 
plotUSGPDpresidents <- function(y, Ylab, Scale=1, 
        Ylim=NA, Log='y', x='Year', Xmin=NA, Type='l', 
        epocs = c(1929, 1933, 1945), epocy=c(.2, .8), 
        epocNames = c('Hoover', 'FDR'), epocex=c(0.9, 1.1), 
        war='war', warShade=c(lty='dotted', col='grey'), 
        Data=USGDPpresidents, ...){
  ##
  ## 1. y
  ##
  Y <- (Scale* Data[, y])
  if(any(is.na(Ylim))){
    if(is.na(Ylim[1])) {
      if(length(grep('y',Log)>0)){
        Y0 <- (Y<=0)
        n0 <- sum(Y0, na.rm=TRUE)
        if(n0>0){
          cat(n0, ' nonpositive observations ignored/n')
          Y[Y0] <- NA
        }
      } 
      Ylim[1] <- min(Y, na.rm=TRUE)
    }
    if(length(Ylim)<2 || is.na(Ylim[2])){
      Ylim[2] <- max(Y, na.rm=TRUE)
    }
  }
  ##
  ## 2. Xmin > Xlim
  ##
  if(is.na(Xmin)){
    imin <- which.max(!is.na(Y))
    Xmin <- Data[imin, x]
  }
  Xlim <- c(Xmin, max(Data[, x]))
  ##
  ## 3. plot
  ##
  plot(Data[, x], Y, log=Log, type=Type, 
       xlim=Xlim, xlab='', ylim=Ylim,
       ylab=Ylab, las=1, ...)  
  ##
  ## 4. epocs
  ##
  abline(v=epocs, lty='dashed')
  if(length(grep('y',Log)>0)){
    lYlim <- log(Ylim)
    yep <- exp(epocy*lYlim[1]+(1-epocy)*lYlim[2])
  } else {
    yep <- (epocy*Ylim[1]+(1-epocy)*Ylim[2])
  }
  for(i in 1:length(epocy)){
    ix <- mean(epocs[i+0:1])
    text(ix, yep[i], epocNames[i], srt=90, 
         cex=epocex[i])
  }
  ##
  ## 5. wars
  ##
  if(!is.null(war) && !is.na(war) &&
     !is.na(match(war, names(Data))))  {
    selWar <- (Data[, war]!='')
    abline(v=Data[selWar, x], lty=warShade['lty'], 
           col=warShade['col'])
  }
}
##
##** Plot the first four variables
##
plotUSGPDpresidents('CPI', Ylab='Consumer Price Index')
plotUSGPDpresidents('GDPdeflator', Ylab='GDP deflator')
plotUSGPDpresidents('population.K', Ylab='population (M)', 
                   Scale=0.001)
plotUSGPDpresidents('realGDPperCapita', 
                   Ylab='GDP per capita (2017K$)', Scale=0.001)

##
##** Table exective and war
##
table(USGDPpresidents[, 'executive'])
table(USGDPpresidents[, 'war'])

##
##** plot battleDeaths 
##** battledeathsPMP = per million population
##
plotUSGPDpresidents('battleDeaths', Ylab='battle deaths (K)', 
                   Scale=0.001, lwd=3)

(MAw <- grep('Mex', USGDPpresidents[, 'war']))
USGDPpresidents[MAw,]

plotUSGPDpresidents('battleDeathsPMP', 
                   Ylab='battle deaths per million pop', lwd=3)
##
##** plot Keynes
##
plotUSGPDpresidents('Keynes', Log='', Ylab='Keynes')
selKeynes <- (USGDPpresidents[, 'Keynes']==1)
range(USGDPpresidents[selKeynes, 'Year'])
##
##** plot unemployment
##
plotUSGPDpresidents('unemployment', Ylab='unemployment (%)')
# Estimated US unemployment rate in percent
# per the sources cited in the section on 
# "Historical unemployment rate charts"
# in the Wikipedia article on 
# "Unemployment in the United States"

# table unemployment source
table(USGDPpresidents[, 'unempSource'])

##
##** plot budget analyses
##
plotUSGPDpresidents('fedReceipts', Ylab="fed receipts ($M)")
plotUSGPDpresidents("fedReceipts", Ylab="fed receipts ($B)", 
                   Scale=0.001)
abline(h=1)
plotUSGPDpresidents("fedOutlays", Ylab="fed outlays ($B)", 
                   Scale=0.001)
abline(h=1)

plotUSGPDpresidents("fedSurplus", Ylab="fed outlays ($B)",
                   Log='', Scale=0.001)

Surplus <- with(USGDPpresidents, fedReceipts - fedOutlays)

rSurp <- with(USGDPpresidents, fedSurplus/Surplus)
plot(USGDPpresidents[, 'Year'], rSurp)
# Discrepancies less than 1 percent

plotUSGPDpresidents('fedDebt', 'fed debt ($B)',
                   Scale=1e-9)

plotUSGPDpresidents('fedReceipts_pGDP', 
                   'fed receipts % of GDP', Scale=100)
plotUSGPDpresidents('fedOutlays_pGDP', 
                   'fed outlays % of GDP', Scale=100)
plotUSGPDpresidents('fedSurplus_pGDP', 
                   'fed surplus % of GDP', Scale=100, Log='')

plotUSGPDpresidents('fedDebt_pGDP', 
                   'fed debt % of GDP', Scale=100)
(debtPrng <- range(USGDPpresidents[, 'fedDebt_pGDP'], na.rm=TRUE))
plotUSGPDpresidents('fedDebt_pGDP', 
                   'fed debt % of GDP', Scale=100, 
                   Ylim=c(1, 100*debtPrng[2]))
abline(h=100)