# The 50 percent problem

What do job approval ratings portend about a president’s reelection prospects?

Quite a lot, according to most analysts, whose mantra is that an incumbent president needs 50 percent approval to be reelected. Gallup’s distinguished editor, Frank Newport, said as much on a recent National Journal panel in which we both participated, and Gallup analyses label the 50 percent mark “determinant for reelection.”

Newport’s colleague Jeffrey Jones intimates the rule might not be so firm, cautioning, “George W. Bush was reelected with a 48 percent approval rating.”

Excavating this 50 percent rule reveals a fairly flimsy foundation.

First, the number of cases is small, and small samples make for bad statistical rules. Since the advent of regular polling, only 11 presidents have sought reelection. In the privacy of my hotel room, I just flipped a coin 10 times. If I flipped it enough times, it would come out 50/50. But after 10 tosses it was nowhere near that— seven tails, three heads. If I told you, based on my experiment, that when flipping a coin you can expect tails 70 percent of the time, you’d say I made a mistake. You’d be right — my sample was just too small. So it might be with approval ratings and election results.

The problems are not just statistical, but substantive. No one, aside from President Bush, was even close to 50 percent, making it difficult to judge exactly what happens to presidents in the mid- to high 40s. The incumbent just below Bush’s 48 percent was his father, way down at 37 percent. The presidents one notch more popular than George W. Bush were and Ronald Reagan, at 58 percent. If all you knew about H2O was that it was water at 55 degrees and ice at 28, you’d be hard-pressed to say what would happen to it at 33 degrees.

Data problems also bedevil the 50 percent rule. Gallup’s numbers are reported differently in different sources. To take but one example, a Gallup document put Gerald Ford’s June rating (its last pre-election) at 32 percent, while the Roper Center archives suggests the same measurement yielded a 45 percent rating. Squishy data yield squishy conclusions.

In a world of simple crosstabs, it’s easy to say only one president with an approval rating below 50 percent has been reelected. (Of course, that depends on when history begins. Gallup starts with Lyndon Johnson. Its data go back to Roosevelt and find Harry Truman was reelected with an approval of just 39 percent in the last pre-election measurement.)

More interesting, though, is how closely votes for incumbents actually match their approval ratings. Going back to Roosevelt and making conservative corrections, on average, presidents get 1 point more in vote than in approval.

But that average masks tremendous variation. Roosevelt’s vote was 17 points lower than his approval. By contrast, George H.W. Bush ran 9 points ahead of his approval rating and Gerald Ford ran either 16 or 3 points ahead, depending on what Gallup actually showed.

Getting inside the approval data offers yet another perspective. Over time, incumbents have garnered the votes of 80-90 percent of those who approve and 10-20 percent of those who disapprove, affording winners a lot of wiggle room on either side of 50 percent approval.

Professor Alan Abramowitz’s statistical model suggests that a 1-point increase in the president’s net approval rating leads to a 0.1 percent increase in vote — meaningful, but hardly the perfect correlation implied by the 50 percent rule.

So based on all the data, what can we say about approval ratings and presidential votes? In short, presidents with approval ratings below 43 percent are quite likely to lose, while those over 55 percent are very likely to win. In between, where President Obama now stands, is the zone of uncertainty, where approval ratings alone provide little guidance.

Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients include the Majority Leader of the Senate and the Democratic Whip in the House.