The Scientific Basis for Allan Lichtman's "13 Keys to the White House"

Allan Lichtman's 13 Keys as applied to presidential elections since 1860. Note the races where the minimum 8 keys favor the incumbent — 1888, 1948, 1996, 2000. In two of these races, the Electoral College went the other way, as it also did in 2016. We may be headed for another such race (source; click to enlarge).

by Thomas Neuburger

I've recently written two pieces that look ahead to the 2020 presidential election. The first, "A Way-Too-Early Handicapping of the 2020 Presidential Race," examined the composition of the electorate — independent, Democratic, Republican; change-wanting, Trump-wanting, or return-to-"normal"–wanting — and determined that if the Democrats nominate Bernie Sanders or someone like him, they would win a strong popular vote victory, something on the order of 54-45%. If, however, they nominate someone not like Sanders, they're risking a much closer race.

The second piece, "A Way-Too-Early Look at Allan Lichtman's 13 Keys to the White House in 2020," looked at historian Allan Lichtman's famous "13 Keys to the White House" with an eye to predicting the 2020 outcome by that method.

There again, the result showed that if the Democrats nominate Bernie Sanders or someone equally inspiring (note, that's not the same as "someone like him"), they would have 9 of the 13 keys, enough to ensure both a popular vote victory and an Electoral College win based on the historical record of the system. If, however, they nominate someone less inspiring, they're again taking their chances.

What that means is this: If not much changes between now and election day to change the keys (for example, if the administration achieves no new foreign policy success) and the Democrats nominate someone uninspiring (denying them the critical 9th key), they would have the bare minimum — 8 of 13 keys — needed to win the popular vote, but the Electoral College would be up for grabs.

In 2016, for example, Lichtman's system predicted that Clinton, with 8 keys, would win the popular vote, which she did, but the Electoral College voted the other way, as it has done in several previous contests with eight-key-only winners, such as Grover Cleveland in 1888 and Al Gore in 2000 (see graphic at top).

The Scientific Basis for Lichtman's Keys

I've since been asked about the scientific basis of Lichtman's method. Is this more than just lucky guessing? The answer is no. What follows is the simplest explanation of why the system works.

It started with a dinner party at Cal Tech in the early 1980s, at which historian Allan Lichtman was seated next to a Russian geophysicist, Vladimir Keilis-Borok, a man who had just devised a system for successfully predicting earthquakes using pattern recognition. Soon the two began comparing notes. Could they find a pattern that marks every election in which the incumbent party retains or loses the White House? Could such a pattern be used for prediction?

Out of those discussions came their initial paper, "Pattern recognition applied to presidential elections in the United States, 1860-1980: Role of integral social, economic, and political traits," which attempted to apply Keilis-Borok's method to U.S. elections.

After a fair amount of testing, they found a pattern that correctly "predicted" (matched the results of) the popular vote in every presidential election from 1864 through 1980. Since then the method has been modified slightly — 13 keys, and slightly different ones, instead of the original 12. Lichtman has since used it to successfully predict the popular vote winner of every election from 1984 to the present. (Lichtman discusses the science behind his keys at greater length here.)

The method works like this. Instead of trying to identify causes — in seismic terms, the geophysical forces that move the earth; in electoral terms, the polling results and issues discussed — the method identifies patterns that indicate general conditions of stability or instability. Politically this means determining if the party in power is in a stable or unstable position relative to the mood of the electorate.

Thus the method involves responding to a series of true-or-false statements like, "The incumbent candidate is the sitting president" (key 3) and "The incumbent administration is untainted by major scandal" (key 9). A yes for both questions shows strength for the incumbent party, and indicate stability. A No indicates some weakness.

Lichtman explained this to the Washington Post in 2008:

"We reconceptualized presidential politics in geophysical terms," said Lichtman, who teaches at American University. "We didn't look at it as Reagan versus Carter or Republicans versus Democrats or liberals versus conservatives. Rather, we looked at elections as stability versus upheavals."

Stability, according to their definition, is when the party that is incumbent in the White House -- in this case, the Republican Party -- wins the next presidential election. Upheavals are when the opposition party wins elections.

The researchers found that four markers or "keys" correctly predicted every presidential election over 120 years. These keys were whether the incumbent party's candidate won the presidential nomination on the first ballot with at least two-thirds of the delegate count, the absence of a third-party candidate who won 5 percent or more of the overall vote, the absence of a recession, and the presence of a major policy victory in the previous term. ...

Given that they wanted to play it safe, Keilis-Borok and Lichtman selected nine other keys that increased the confidence with which they could have predicted all the elections between 1860 and 1980.

Note: The original paper tested 12 yes-or-no questions. The method has evolved into 13 statements that could be assessed as true or false. Most discussions of this research refer to the final 13 keys in true-or-false statement form. The definitive version appears in Lichtman's 1991 book, The Keys to the White House.

Why These 13 Keys?

Where did Lichtman and Keilis-Borok get these particular keys? In an article in Analytics Magazine, writer Douglas Samuelson explains (emphasis mine):

Lichtman is quick to point out that his method is based on a solid statistical model that incorporates a test of competing theories of politics, and the prediction results validate some of the theories and contradict others. His method is based on a statistical pattern recognition algorithm for predicting earthquakes, implemented by Russian seismologist Vladimir Keilis-Borok. The highest plurality of the popular vote, not the electoral vote that actually decides the presidency, is the criterion ...

Out of nearly 200 questions, which were all binary (“yes” or “no”) variables, the algorithm picked those that displayed the greatest difference between the proportion of the time the variable was “yes” for years when the incumbent party won and the corresponding proportion for years when the challenging party won, using all U. S. elections from 1860 through 1976 as the training set.

"Displayed the greatest difference between the proportion of the time the variable was “yes” for years when the incumbent party won and the corresponding proportion for years when the challenging party won" means this: Determine which questions show the greatest tendency to be right when the incumbent won and also right when the challenger won.

This is a bit simplified, but imagine that the answer to a given question (a key) was right in 12 elections when the incumbent won and also right in 7 elections when the challenger won. You could assign a differential score of 19 (12+7) to that key. The keys with the highest differential scores would then be selected as, together, identifying the pattern they were looking to find.

Thus the 13 keys in their final form. So far, right every time.