Timothy T. Taylor Home Page
Journal of Economic Perspectives
Articles and Writing
Economics Textbook
Classroom Teaching
The Teaching Company
High School Pedagogy

Articles and Writing

March 30, 1992
"Professor's Model is Usually Right - Do the Math and Bush Wins"
San Jose Mercury News

By Timothy Taylor
<< Back to 1992 menu

FOR ABOUT a year now, the conventional political wisdom has held that the recession will drag down President Bush, and give the Democrats a real shot at the White House.

But whenever conventional wisdom lives more than a few months, it's ripe for overturning. The historical statistics say that the economy is not nearly shaky enough for voters to turn out an incumbent president.

Ray Fair, a professor of economics at Yale University, has derived a formula for relating the state of the economy to the election returns. According to his calculations, the Democratic share of the total votes cast for the two main parties (thus ignoring ballots cast for third parties) can be predicted in this way:

Start with 48.49 percent, which can be thought of as the baseline Democratic presidential vote.

Next, add 0.53 percent if the incumbent is a Democrat, but subtract that amount if the incumbent is a Republican.

Then, add 4.24 percent if one candidate is an incumbent Democrat who is running for re-election, but subtract that amount if one candidate is an incumbent Republican running for re-election.

Add 1.04 times the annual percentage growth rate of gross national product, measured on a per capita basis and adjusted for inflation, in the second and third quarters of the election year if the incumbent is a Democrat, or subtract that amount if the incumbent is Republican.

Finally, add 0.31 times the annual percentage rate of inflation in the two years prior to the election if the incumbent is a Republican, but subtract that amount if the incumbent is a Democrat.

With the oddly numbered percentages and multipliers, the calculation may appear more complex than it really is. The precise numbers come from a statistical test that Fair performed on economic data and the results of all presidential elections since 1916. (For those who care, it's an ordinary least squares regression, with a time trend.)

In the 19 elections since then, Fair's formula has predicted the winner incorrectly only three times. Those were very close elections, like the 1976 Carter-Ford contest, or the 1960 Kennedy-Nixon race.

The average prediction error for the elections since 1968 is only 1.1 percentage points. In the 1988 election, Dukakis was predicted to get 46.8 percent of the votes that went to either Democrats or Republicans, and he actually received 46.1 percent.

The fact that it is possible to predict election results so closely based on nothing but incumbency, recent inflation, and two quarters of growth statistics is a bit astonishing. Clearly, economic factors can swing elections.

But the results also emphasize that the Democrats will have a very, very tough job unseating George Bush. If the second and third quarters of 1992 looked like they would bring, oh, 7 percent inflation and the economy shrinking at about 4 percent, then the Democrat would have a good chance at the presidency, according to Fair's formula.

But consider the more likely scenario. Bush's political gurus are just now assembling a blizzard of campaign ads that will give him credit for taking down the Berlin Wall, driving Iraq out of Kuwait, bringing home the hostages, dismantling the Soviet Union, and ending the Cold War. Not to mention ads attacking the Democratic candidate.

On the economic front, inflation looks to remain at less than 3 percent. The economy has expanded, although just barely, since the second quarter of last year. Signs of a sturdier economic recovery have been sprinkled through the news of the last few months, including pickups in home and car sales, and rising levels of business activity.

Even if these signs fizzle out by late this year or early 1993, remember that according to Fair's calculations, the period from April through September is what most affects the election. Even if growth stays sluggish, Fair's formula predicts a sizable Bush victory in November.

The countdown to the general election still has more than seven months to go, which is one-seventh of Bush's entire term of office. Seven months is an eternity in politics. If you don't believe that, remember, back in 1988, that Dukakis was well ahead in the polls after the Democratic convention, just four months before the election.

Or remember that seven months ago, the big issue was how the Soviet Union -- which no longer exists -- would survive the after-effects of the coup against Mikhail Gorbachev -- who is no longer in power. When election day finally arrives in November, the dismal economic news of the past year is likely to seem every bit as distant as that coup.

Americans traditionally don't throw out their incumbent presidents for being average, or even below-average. The advantages of Bush's incumbency have been overshadowed in recent weeks by the excitement over the Republican protest vote for Patrick Buchanan, and the Democratic in-fighting. But before the voters evicted Jimmy Carter from the White House in 1980, it took dozens of U.S. hostages in Iran, along with double-digit inflation and unemployment. Unless an economic or political disaster of those near-Biblical proportions befalls Bush in the next few months, any odds-maker would have to like his chances for re-election.

Working through Ray Fair's model with realistic assumptions for growth in GNP, 2 percent, and for inflation, 3 percent, November looks like a bad month for the Democratic candidate:
  • Begin with 48.49 percent as the baseline Democratic vote.
  • Subtract 0.53 percent because the incumbent is a Republican.
  • Subtract 4.24 percent because the incumbent is a Republican running for re-election.
  • Subtract 1.04 times 2 percent (the assumed growth rate for GNP) because the incumbent is a Republican.
  • Add 0.31 times 3 percent (the assumed annual inflation rate) because the incumbent is a Republican.

Total: The Democratic candidate will get 42.57 percent of the votes cast for the two main parties.

<< Back to 1992 menu