Purpose of this document

This document describes the process for updating Ecdat::USGDPpresidents.

Set working directory

First decide the directory in which we want to work and copy this vignette (*.Rmd file) into that directory. (RStudio does not allow setwd inside code chunks to work as one might naively expect. Therefore, it’s best NOT to try to change the working directory but instead to copy this vignette into the desired working directory.)

Are there new data?

Start by checking the span of years in USGDPpresidents:

library(Ecdat)
## 
## Attaching package: 'Ecdat'
## The following object is masked from 'package:datasets':
## 
##     Orange
(rngYrs <- range(USGDPpresidents$Year))
## [1] 1610 2024

Next download “GDP - US” and “CPI - US” from Measuring Worth. On 2022-02-16 this produced two csv files, which I downloaded and copied into a directory in which we wish to work.

getwd()
## [1] "/private/var/folders/2n/zqk768wj3818l8x2wttbc5kw0000gn/T/RtmpHRNbbn/Rbuild11cd57e99bead/Ecfun/vignettes"
(csv2 <- dir(pattern='\\.csv$'))
## character(0)
(CPIcsvs <- grep('^USCPI', csv2, value=TRUE))
## character(0)
(CPIcsv <- tail(CPIcsvs, 1))
## character(0)
(GDPcsvs <- grep('^USGDP', csv2, value=TRUE))
## character(0)
(GDPcsv <- tail(GDPcsvs, 1))
## character(0)
if((length(CPIcsv)==1) & (length(GDPcsv)==1)){
  Update0 <- TRUE
} else Update0 <- FALSE

We must verify by visual inspection that CPIcsv and GDPcsv are both of length 1 and are the files we want.

Read them:

Update <- FALSE
if(Update0){
  str(USCPI <- read.csv(CPIcsv, skip=2))
  str(USGDP. <- read.csv(GDPcsv, skip=1))
  library(Ecfun)
  USGDP <- asNumericDF(USGDP.)
  print(rngCPIyrs <- range(USCPI$Year) )
  print(rngGDPyrs <- range(USGDP$Year) )
  endYr <- max(rngCPIyrs, rngGDPyrs)
  if(endYr>rngYrs[2]) print(Update <- TRUE)
}

Update

If Update, create a local copy of USGDPpresidents with the additional rows required to hold the new data:

if(Update){
  rowsNeeded <- (endYr - rngYrs[2])
  Nold <- nrow(USGDPpresidents)
  iRep <- c(1:Nold, rep(Nold, rowsNeeded))
  USGDPp2 <- USGDPpresidents[iRep,]
}

Fix the Year and insert NAs for all other columns for the new rows:

if(Update){
  iNew <- (Nold+(1:rowsNeeded))
  USGDPp2$Year[iNew] <- ((rngYrs[2]+1):endYr)
  rownames(USGDPp2) <- USGDPp2$Year
#
  USGDPp2[iNew, -1] <- NA
}

Now replace CPI by the new numbers:

if(Update){
  selCPI <- (USGDPp2$Year %in% USCPI$Year)
  if(any(!is.na(USGDPp2[!selCPI, 2]))){
    stop('ERROR:  There are CPI numbers ', 
         'in the current USGDPpresidents ', 
         'that are not in the new.  ', 
         'Manual review required.')
  }
  USGDPp2$CPI[selCPI] <- USCPI[,2]
}

Does USGDPpresidents.Rd needs to be updated to reflect the proper reference years for the CPI?

if(Update){
  readLines(CPIcsv, n=4)
}

If this says “Average 1982-84 = 100”, it should be good. Otherwise that (and this) should be updated.

Now let’s update GDPdeflator:

if(Update){
  selGDP <- (USGDPp2$Year %in% USGDP$Year)
#
  if(any(!is.na(USGDPp2[!selGDP, 'GDPdeflator']))){
    stop('ERROR:  There are GDPdeflator numbers ', 
         'in the current USGDPpresidents ', 
         'that are not in the new.  ', 
         'Manual review required.')
  }
  selDefl <- grep('Deflator', names(USGDP))
  USGDPp2$GDPdeflator[selGDP] <- USGDP[,selDefl]
  print(names(USGDP)[selDefl])
}

Compare the index year of “GDP.Deflator” with that in USGDPpresidents.Rd: If they are different, fix USGDPpresidents.Rd.

Now update population:

if(Update){
  selPop <- grep('Population', names(USGDP))
  sPop <- (USGDP[,selPop]/1000)
  quantile(ePop <- ((USGDPp2$population.K[selGDP] /sPop)-1), 
           na.rm=TRUE)
}

Check. Replace.

if(Update){
  USGDPp2$population.K[selGDP] <- sPop
  print(names(USGDP)[selPop])
}

Now realGDPperCapita. This also has a reference year, so we need to make sure we get them all:

if(Update){
  if(any(!is.na(USGDPp2[!selGDP, 'readGDPperCapita']))){
    stop('ERROR:  There are realGDPperCapita numbers ', 
         'in the current USGDPpresidents ', 
         'that are not in the new.  ', 
         'Manual review required.')
  }
  selGDPperC <- grep('Real.GDP.per.c', names(USGDP))
  USGDPp2$realGDPperCapita[selGDP] <- USGDP[,selGDPperC]
  print(names(USGDP)[selGDPperC])
}

Compare the index year of Real.GDP.per.capita with that in USGDPpresidents.Rd: If they are different, fix USGDPpresidents.Rd.

Next: executive:

NOTE: THIS MAY NEED TO BE CHANGED MANUALLY HERE BEFORE EXECUTING, BECAUSE IT IS NOT IN USGDP… BOTH: ** WHO WAS PRESIDENT SINCE THE PREVIOUS VERSION? ** WAS THAT PERSON NOT IN THE PREVIOUS VERSION?

if(Update){
  exec <- as.character(USGDPp2$executive)
  newExec <- 'Biden' 
  exec[is.na(exec)] <- newExec
  lvlexec <- levels(USGDPp2$executive)
  if(!(newExec %in% lvlexec))
    lvlexec <- c(lvlexec, newexec)
  USGDPp2$executive <- ordered(exec, lvlexec)
}

Similarly: war

NOTE: IF THERE HAS BEEN A MAJOR WAR SINCE THE LAST VERSION, THEN THIS TEXT NEEDS TO BE CHANGED, BECAUSE IT ASSUMES THERE HAS NOT BEEN A MAJOR WAR.

if(Update){
  war <- as.character(USGDPp2$war)
  war[is.na(war)] <- ''
  lvlwar <- levels(USGDPp2$war)
  USGDPp2$war <- ordered(war, lvlwar)
}

Next: battleDeaths and battleDeathsPMP:

NOTE: battleDeaths ARE ONLY BATTLE DEATHS IN MAJOR WARS as defined in help(USGDPpresidents).
Otherwise, they are 0.

if(Update){
  USGDPp2$battleDeaths[iNew] <- 0 
#
  USGDPp2$battleDeathsPMP <- with(USGDPp2, 
          1000*battleDeaths/population.K) 
}

Keynes (per help(USGDPpresidents)):

if(Update){
  USGDPp2$Keynes[iNew] <- 0 
}

Unemployment?

Unemployment figures came from different sources for different years. Since 1940 the source has been the Bureau of Labor Statistics (BLS), 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.
Download the most recent years as an Excel file, compute row averages, and transfer the numbers for the most recent years here.

NOTE: When I visited the Current Population Survey of the Bureau of Labor Statistics on 2025-08-22, I found a huge number of options. I clicked, “CPS Data”, then “Data retrieval tools”, then “Labor Force Statistics (Current Population Survey - CPS)” “Top Picks”. Then I found Unemployment Rate - LNS14000000 and checked that. Then near the bottom of that page, I clicked, “Retrieve data”. That produced a table for years 2015:2025 with columns Jan, Feb, …, December, with the data for 2025 incomplete, as it should be. Just above that table it said, “Download: xlsx”. I clicked that. I opened that spreadsheet and added column N = average of columns B - M.

Then I compared those numbers with the numbers in USGDPp2[c('Year', 'unemployment')]. The numbers for 2020 were slightly different: 8.091667 in the previous version and 8.1 for the new number.

Let’s read the new numbers and compare the numbers to confirm that we have read them correctly, then replace the old numbers (including NAs) with the new numbers.

if(Update){
  (xls <- dir(pattern='\\.xlsx$'))
  (BLSxls <- grep('^Series', xls, value=TRUE))
}
library(readxl)
if(Update){
  str(BLS <- read_excel(BLSxls, skip=11))
}

Compute the average unemployment here, so I don’t have to do this separately.

if(Update){
  UNEMP <- as.matrix(BLS[2:13])
  str(unemp <- apply(UNEMP, 1, mean))
}

Store these unemp numbers after checking first.

if(Update){
  selU4GDP <- (USGDPp2$Year %in% BLS$Year)
  selBLS <- (BLS$Year %in% USGDPp2$Year)
  dunemp <- (USGDPp2[selU4GDP, 'unemployment'] - 
          unemp[selBLS])
  cbind(USGDPp2[selU4GDP, c('Year', 'unemployment')], 
        unemp[selBLS], dunemp)
}

As expected. Replace.

if(Update){
  USGDPp2[selU4GDP, 'unemployment'] <- unemp[selBLS]
  USGDPp2$unempSource[iNew] <- USGDPp2$unempSource[
    iNew[1]-1]
  tail(USGDPp2)
}

fedReceipts, fedOutlays

We get fedReceipts and fedOutlays from two different sources. Let’s start with the historical data first.

Skip historical data on fedRecepts and fedOutlays

We manually copied the historical data from series Y 335 and 336 in United States Census Bureau (1975) Bicentennial Edition: Historical Statistics of the United States, Colonial Times to 1970, Part 2. Chapter Y. Government into a LibreOffice *.ods file. We need to read that once and add it to USGDPp:

if(Update){
  (odsFile <- dir(pattern='\\.ods'))
  (odsF <- grep('^hstat', odsFile, value=TRUE))
}
if(Update){
  library(readODS)
  str(hstat <- read_ods(odsF, sheet='Receipts', skip=2))
}
if(Update){
  Hstat <- hstat[!is.na(hstat$Year), 1:3]
  oOld <- order(Hstat$Year)
  head(Hst <- Hstat[oOld, ])
}

Add as new variables to USGDPp2:

if(Update){
  USGDPp2$fedReceipts <- NA 
  USGDPp2$fedOutlays <- NA
  selGDP4Hst <- (USGDPp2$Year %in% Hst$Year)
  USGDPp2[selGDP4Hst, c("fedReceipts", "fedOutlays")] <- 
      (Hst[2:3] / 1000)
  USGDPp2[c('Year', 'fedReceipts', 'fedOutlays')]
}

new data on fedRecepts and fedOutlays

For the latest data on fedReceipts, fedOutlays, and fedSurplus, I went to the website for The White House President’s Budget Historical Tables. 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.

(xls2 <- dir(pattern='\\.xlsx$'))
## character(0)
if(Update){
  (BudgetFiles <- grep('^BUDGET', xls2, value=TRUE))
  (BudgetF2_1 <- grep('2-1', BudgetFiles, value=TRUE))
  (BudgetFile <- (if(length(BudgetF2_1)>0) 
    tail(BudgetF2_1, 1) else tail(BudgetFiles, 1)))
}

Confirm that BudgetFile is what we want.

From opening this file in spreadsheet software, it appears that we want tab hist01z1.

if(Update){
  Budget <- read_excel(BudgetFile, 
          sheet='hist01z1', skip=3)
  head(Budget)
  tail(Budget)  
}

Let’s use only the most recent 40 years, because there are anomalies in these data like “-*” for a number that is “$500 thousand or less” and TQ for “transitional quarter” when the US had its fiscal year change from starting July 1 to October 1. We also drop the last two row, because they are comments. And keep only columns 1:4:

if(Update){
  library(Ecfun)
  nBudg0 <- nrow(Budget)
  iBudg <- sort(seq(to=nBudg0-2, length=40))
  str(Budg <- asNumericDF(Budget[iBudg, 1:4]))
  tail(Budg)
}
if(Update){
  selGDP4budg <- (USGDPp2$Year %in% Budg[, 1])
  selBudg <- (Budg[, 1] %in% USGDPp2$Year)
  dfedR <- (USGDPp2[selGDP4budg, 'fedReceipts'] 
              - Budg[selBudg, 2])
  dfedO <- (USGDPp2[selGDP4budg, 'fedOutlays'] 
              - Budg[selBudg, 3])
  dfedS <- (USGDPp2[selGDP4budg, 'fedSurplus'] 
              - Budg[selBudg, 4])
  tail(cbind(USGDPp2[selGDP4budg, c('Year', 
          'fedReceipts', 'fedOutlays', 'fedSurplus')], 
       Budg[selBudg, 2:4], dfedR, dfedO, dfedS), 10)
  matplot(cbind(dfedR, dfedO, dfedS), type='l')
}

There are tiny changes in the years since 2017. There may also be a few in earlier years, but we will ignore the earlier years.

Let’s replace the numbers for fedReceipts, fedOutlays, and fedSurplus for 2017:2024.

if(Update){
  table(sel2017_2024 <- (USGDPp2$Year %in% 2017:2024))
  table(s2017_2024 <- (Budg[, 1] %in% 2017:2024))

  USGDPp2[sel2017_2024, c('fedReceipts', 'fedOutlays', 
      'fedSurplus')] <- Budg[s2017_2024, 2:4]
  tail(USGDPp2)
}

Let’s plot these budget numbers before proceeding.

if(Update){
  Xlim <- c(1790, max(USGDPp2$Year, na.rm=TRUE))
  plot(fedReceipts ~Year, USGDPp2, log='y', type='l', 
     xlim=Xlim, las=2)
  Xlim <- c(1790, max(USGDPp2$Year, na.rm=TRUE))

  plot(fedOutlays ~Year, USGDPp2, log='y', type='l', 
     xlim=Xlim, las=2)

  plot(fedSurplus ~Year, USGDPp2, type='l', 
     xlim=Xlim, las=2)
}

fedDebt is not the negative of a simple cumulative sum fedSurplus. The sources of the discrepancies are not clear.However, some outlays are “Off-budget” including a “black budget” that is not revealed to many (and perhaps all) members of the US Congress. It’s not obvious, at least to this researcher, if interest on the national debt is included in the official budget.

fedDebt are available as “Historical Debt Outstanding” from the US Treasury. On 2025-08-23 we requested “Date Range (Record Date): All”, then “CSV” and “Download CSV File”. The result was HstDebt_17900101_20240930.csv.

(csv3 <- dir(pattern='\\.csv$'))
## character(0)
if(Update){
  (debtFiles <- grep('^HstDebt', csv3, value=TRUE))
  tail(HstDebt <- read.csv(debtFiles))
  (HstDebt6 <- head(HstDebt))
  tail(USGDPp2[c('Year', 'fedDebt')])
}

Visual inspection suggests that the numbers match for 2019:2021. Let’s compute the difference to confirm.

if(Update){
  nobs <- nrow(USGDPp2)
  (endRows <- seq(nobs, by=-1, length=6))
  (dHstDebt6 <- (USGDPp2$fedDebt[endRows]-HstDebt6[, 2]))
}

Roundoff error. Let’s replace those numbers.

if(Update){
  (USGDPp2$fedDebt[endRows] <-HstDebt6[, 2])
  tail(USGDPp2)
  plot(fedDebt ~Year, USGDPp2, type='l', log='y',
     xlim=Xlim, las=2)
}

Finally: fedOutlays, … fedDebt as a percent of GDP.

For *_pGDP, I’m getting discrepancies that seem a little more than roundoff error. Let’s look at the numbers since 1843, which was the year the US first adopted a fiscal year different from the calendar year.

if(Update){
  selEnd <- (USGDPp2$Year>1843) 
  currentGDP <- with(USGDPp2[selEnd, ], 
      1000 * population.K * realGDPperCapita 
          * GDPdeflator / 100)
  plot(USGDPp2$Year[selEnd], currentGDP, 
       log='y', type='l', las=2)
  tail(currentGDP)
}

GDP for 2024 is just over 29 trillion. Confirmed. And the plot also looks plausible. Continue.

if(Update){
  plot(fedReceipts~Year, USGDPp2[selEnd, ], log='y', 
     type='l', las=2)
}

Plausible.

if(Update){
  fedR_p <- (1e6*USGDPp2$fedReceipts[selEnd] / 
             currentGDP)
  plot(USGDPp2$Year[selEnd], fedR_p, type='l', 
     las=2, log='y')
  matplot(USGDPp2$Year[selEnd], 
        cbind(USGDPp2$fedReceipts_pGDP[selEnd], fedR_p), 
        type='l', las=2, log='y')
}

Good. Ratio?

if(Update){
  plot(USGDPp2$Year[selEnd], 
      USGDPp2$fedReceipts_pGDP[selEnd] / fedR_p, 
        type='l', las=2, log='y')
}

The new numbers differ by less than 3 percent from the previous numbers. I don’t think I care.

Use the new numbers.

if(Update){
  USGDPp2$fedReceipts_pGDP[selEnd] <- fedR_p
  tail(USGDPp2)
}

Next fedOutlays_pGDP.

if(Update){
  fedO_p <- (1e6*USGDPp2$fedOutlays[selEnd] / 
             currentGDP)
  matplot(USGDPp2$Year[selEnd], 
        cbind(USGDPp2$fedOutlays_pGDP[selEnd], fedO_p), 
        type='l', las=2, log='y')
}

Good, similar to Receipts. Ratio?

if(Update){
  plot(USGDPp2$Year[selEnd], 
      USGDPp2$fedOutlays_pGDP[selEnd] / fedO_p, 
        type='l', las=2, log='y')
}

Like Receipts. Store.

if(Update){
  USGDPp2$fedOutlays_pGDP[selEnd] <- fedO_p
  tail(USGDPp2)
}

Good. Surplus?

if(Update){
  fedS_p <- (1e6*USGDPp2$fedSurplus[selEnd] / 
             currentGDP)
  matplot(USGDPp2$Year[selEnd], 
        cbind(USGDPp2$fedSurplus_pGDP[selEnd], fedS_p), 
        type='l', las=2)
}

Good, similar to Receipts and Outlays. Ratio?

if(Update){
  plot(USGDPp2$Year[selEnd], 
      USGDPp2$fedSurplus_pGDP[selEnd] / fedS_p, 
        type='l', las=2)
  quantile(rSup <- (USGDPp2$fedSurplus_pGDP[selEnd] / fedS_p), 
           na.rm=TRUE)
}

Good, similar to Receipts and Outlays. Store.

if(Update){
  USGDPp2$fedSurplus_pGDP[selEnd] <- fedS_p
  tail(USGDPp2)
}

fedDebt?

if(Update){
  fedD_p <- (USGDPp2$fedDebt[selEnd] / 
             currentGDP)
  matplot(USGDPp2$Year[selEnd], 
        cbind(USGDPp2$fedDebt_pGDP[selEnd], fedD_p), 
        type='l', las=2, log='y')
}

Good, similar to Receipts, Outlays and Surplus. Ratio?

if(Update){
  plot(USGDPp2$Year[selEnd], 
      USGDPp2$fedDebt_pGDP[selEnd] / fedD_p, 
        type='l', las=2)
}

As before. Store.

if(Update){
  USGDPp2$fedDebt_pGDP[selEnd] <- fedD_p
  tail(USGDPp2)
}

Plot US federal outlays

if(Update){
  USGDPpresidents <- USGDPp2

  sel <- !is.na(USGDPpresidents$fedOutlays_pGDP)
  plot(100*fedOutlays_pGDP~Year, 
     USGDPpresidents[sel,], type='l', log='y', 
     xlab='', ylab='US federal outlays, % of GDP')
  abline(h=2:3)
  War <- (USGDPpresidents$war !='')
  abline(v=USGDPpresidents$Year[War], 
    lty='dotted', col='light gray')
  abline(v=c(1929, 1933), col='red', lty='dotted')
  text(1931, 22, 'Hoover', srt=90, col='red')
}

How about the same plot of Deficit = -fedSurplus_pGDP?

if(Update){
  selD <- !is.na(USGDPpresidents$fedSurplus_pGDP)
  plot(-100*fedSurplus_pGDP~Year, 
     USGDPpresidents[sel,], type='l', 
     xlab='', ylab='US federal deficit, % of GDP')
  abline(h=2:3)
  abline(v=USGDPpresidents$Year[War], 
    lty='dotted', col='light gray')
  abline(v=c(1929, 1933), col='red', lty='dotted')
  text(1931, 22, 'Hoover', srt=90, col='red')
}

What about inflation = diff(log(CPI))?

if(Update){
  selI <- (USGDPpresidents$Year>1789)
  quantile(diff(USGDPpresidents$Year[selI]))
}
if(Update){
  infl <- 100*diff(log(USGDPpresidents$CPI[selI]))
  yr2 <- USGDPpresidents$Year[selI][-1]
  plot(yr2, infl, type='l', las=2)
  abline(h=c(-2, 0, 2, 10))
  abline(v=USGDPpresidents$Year[War], 
    lty='dotted', col='light gray')
  abline(v=c(1929, 1933), col='red', lty='dotted')
  text(1931, 22, 'Hoover', srt=90, col='red')
}
if(Update){
  infl2 <- 100*diff(log(
    USGDPpresidents$GDPdeflator[selI]))
  plot(yr2, infl2, type='l', las=2)
  abline(h=c(-2, 0, 2, 10))
  abline(v=USGDPpresidents$Year[War], 
    lty='dotted', col='light gray')
  abline(v=c(1929, 1933), col='red', lty='dotted')
  text(1931, 22, 'Hoover', srt=90, col='red')
}
if(Update){
  plot(battleDeathsPMP~Year, USGDPpresidents,  
       type='l', las=2, xlim=Xlim)
  abline(h=100)
  
  plot(1+battleDeathsPMP~Year, USGDPpresidents,  
       type='l', las=2, xlim=Xlim, log='y')
  abline(h=100)
  abline(v=USGDPpresidents$Year[War], 
    lty='dotted', col='light gray')
  abline(v=c(1929, 1933), col='red', lty='dotted')
  text(1931, 22, 'Hoover', srt=90, col='red')
}

Done: Save

if(Update){
  save(USGDPpresidents, file='USGDPpresidents.rda')
  getwd()
}

Now copy this file from the current working directory to ~Ecdat\data, overwriting the previous version.