Solving Elections’ “Snowplow Problem”


[Image courtesy of]

Election Day has come and gone in jurisdictions across the country, and in those places that use paper ballots – usually optical scan – one of the post-election tasks is to handle the (usually considerable) inventory of unvoted ballots. Such ballots must be stored, recycled or destroyed according to the appropriate state and local laws.

The need to deal with unvoted ballots is elections’ own “snowplow problem.” Such problems are commonplace in the field of public administration generally, given the need to balance public demands for services and finite public funds.

The problem takes its name from the choice facing any jurisdiction where snow is a significant but uncertain threat. In places where snow is expected (e.g. Minnesota), snowplows are a necessity; in places where snow is a rarity (e.g.. Miami) they are an extravagance. But in places like my home of Northern Virginia where any given winter can range from mild to Snowpocalyptic, the decision to buy snowplows is fraught with peril for policymakers. Buy them, and a string of mild winters will expose you to criticism for wasteful and unnecessary spending – but refrain and suffer the wrath of snowed-in residents who can’t understand “why we can put a man on the moon but can’t plow my street.”

Elections’ equivalent to the “snowplow problem” is the question of how many ballots to print for an election. Anyone even remotely familiar with voting in America knows that turnout rarely if ever approaches 100%; yet in many jurisdictions election offices are required to print 100% (or even more) of the eligible voter list in order to prevent any voter from having an opportunity to cast a ballot. Such concerns seem laughably improbable until you remember the scorn heaped on election officials in places like Bridgeport, CT – where ballots ran out in 2010 and created chaos in their wake.

The good news for election officials is that turnout – or, more specifically, voter demand for ballots – is far more predictable than snow. That is why we are beginning to see efforts around the nation to modernize ballot production more in order to more closely match the number of ballots to the number of voters needing them. Such efforts right now are focused in two separate (but not mutually exclusive) areas:

  1. using data on past elections and other demographic factors to forecast turnout, which could then be used to create a “confidence band” for the number of ballots to be printed; and
  2. “ballot on demand” systems like the one in Washington County, NY profiled late last week. Such systems use a voter’s address to generate a personalized ballot on the spot.

These are promising developments but widespread adoption of one or both of these solutions will require a number of changes in the current system of voting in this country:

  • + amending state and local laws which require ballots for 100% (or more) of the eligible voter population;
  • + altering or abandoning the relationship with commercial ballot printers, who are often closely tied – if not owned outright – by voting technology vendors; and most importantly.
  • + convincing policymakers and election officials (who will catch the grenade if things go wrong) that the new system will ensure that every voter who wants a ballot on Election Day will get one.

These are significant hurdles to clear, and so I don’t want to overstate the chances that the new approaches will become commonplace anytime soon – but the sooner the election community can find ways to use data and technology to forecast voter demand for ballots, the sooner election offices can begin to eliminate the single most wasteful line item in their budgets: the need to clear away a blizzard of unvoted ballots after Election Day.

Be the first to comment on "Solving Elections’ “Snowplow Problem”"

Leave a comment

Your email address will not be published.