What is the minimum no. of votes that could have swung previous US elections?

elec_marg.R

# Context: https://bsky.app/profile/gmcd.bsky.social/post/3lzvu75zygk2f

library(rvest)
library(data.table)

elec_marg = function(year = 2020) {
  wurl = paste0("https://en.wikipedia.org/wiki/", year, "_United_States_presidential_election")
  ev = read_html(wurl)  |>
    html_element("table.wikitable:nth-child(1)") |>
    html_table()

  h = ev[1, ] |> as.vector()
  h[1] = ""
  names(h)[1] = "state"
  idx = grepl("^state|Democratic$|Republican$", names(h)) & !grepl("%", h)
  h = h[idx]

  nms = c(
    names(h)[1],
    paste0(tolower(h[-1]), "_", gsub(".*(Democratic|Republican)", "\\1", names(h[-1])))
  )

  ev = as.data.table(ev)[-1, ..idx] |> 
    setnames(nms)
  suppressWarnings(
    ev[, names(.SD) := lapply(.SD, \(s) as.numeric(gsub(',', '', s))), .SDcols = -1]
  )
  setnafill(ev, type = "const", fill = 0, cols = c("ev_Democratic", "ev_Republican"))

  # housecleaning (optional but nicer printing)
  ev[, state := gsub(".*Tooltip |\\[.*", "", state)]
  ev[, state := gsub("(\\w+)'s (\\d+)\\w+ congressional district", "\\1 \\2", state)]
  # remove total case and any remaining NA cases (artifact of html tables)
  ev = na.omit(ev[state != "Total"])
  # remove special congression cases (handled directly)
  ev = ev[!grepl("^Maine |^Nebraska ", state)]
 
  # reshape (using the special measure(value.name) keyword to efficiently
  # reshape and split across party-vote types, i.e. votes_<party>, ev_<party>.)
  ev = melt(ev, measure.vars = measure(value.name, party, sep = "_"))

  # get the total number of EVs and order by winning party
  ev_tot = ev[, lapply(.SD, sum), .SDcols = c("votes", "ev"), by = party][order(-ev)]

  # carry over to main dataset for ordering (needed for diff calc below)
  ev[, party := factor(party, levels = ev_tot$party)]
  setorder(ev, party)

  ev_diff = ev[
    order(state, party),
    lapply(.SD, diff),
    by = state,
    .SDcols = c("votes", "ev")
  ]

  # how many EVs are required to flip the result?
  req_ev = 270 - ev_tot[, min(ev)]

  # calculate min req votes based on most efficient potential gains (votes/ev)
  flippable = ev_diff[ev < 0][order(abs(votes/ev))]
  fidx = which(cumsum(abs(flippable$ev)) >= req_ev)[1]
  min_votes = sum(abs(flippable$votes[1:fidx]))
  min_states = flippable$state[1:fidx]

  min_votes_perc_tot = paste0(signif(min_votes / sum(ev_tot$votes) * 100, 1), "%")
  min_votes_perc_flip = paste0(signif(min_votes / ev[state %chin% min_states][, sum(votes)] * 100, 1), "%")

  cat("Winner:", ev_tot$party[1], "\n")
  cat("Min votes to flip ", ev_tot$party[2], ": ",
      prettyNum(min_votes, big.mark = ","),
      "\n", sep = "")
  cat("  ", min_votes_perc_tot, " of total votes\n", sep = "")
  cat("  ", min_votes_perc_flip, " of votes from ", fidx, " states",
      " (", paste(flippable$state[1:fidx], collapse = ", "), ")",
      "\n", sep = "")
  return(invisible(min_votes))
}

## Examples

elec_marg(2016)
#> Winner: Republican 
#> Min votes to flip Democratic: 77,744
#>   0.06% of total votes
#>   0.6% of votes from 3 states (Michigan, Pennsylvania, Wisconsin)

elec_marg(2020)
#> Winner: Democratic 
#> Min votes to flip Republican: 123,473
#>   0.08% of total votes
#>   0.7% of votes from 4 states (Georgia, Arizona, Wisconsin, Pennsylvania)

elec_marg(2024)
#> Winner: Republican 
#> Min votes to flip Democratic: 344,866
#>   0.2% of total votes
#>   2% of votes from 4 states (Wisconsin, Michigan, Pennsylvania, Georgia)