Re-Assessing Obama’s “Under-Performance” in the Polls

One of the more puzzling questions in election polling this year is why isn't Barack Obama doing better in the polls? In a year when outgoing president George W. Bush has an approval rating around 30%, when less than 20% of Americans are satisfied with conditions in the United States, when Democrats have had the largest advantages in party ID they have enjoyed in recent memory, and when Americans have a much more positive view of the Democratic Party than of the Republican Party, Democrat Obama has held only a modest lead over Republican John McCain, averaging just 3 points among registered voters since early June.


It's possible that it's not Obama's performance in the polls that is lacking, but that the expectations for how he should be doing are too high. The high expectations for Obama are based largely on an assessment that the political environment is very favorable for the Democrats, but maybe that will not be as big a factor in this election as in other elections.

I'm certainly a believer in the power of the political environment to explain why elections turn out the way they do. A president's party usually loses seats in Congress in midterm elections, but these losses tend to be greater when the president is unpopular, the economy is bad, and Americans are dissatisfied with the state of the country. We have only to go back to 2006 for a terrific example of this. And elections in which an incumbent president is running have proven to be quite predictable based on the president's approval rating -- not just in terms of who will win but also how close the election might be.

My expectation going into the year was that Democrats seemed like a safe bet to win the election, and probably by a comfortable margin in the popular vote given the political environment. While I still believe the Democrats are the favorites, I'm not convinced that Obama will win by a very big popular vote margin if he does in fact prevail. I began to have some doubts about that after seeing a presentation by a former professor of mine, Helmut Norpoth of Stony Brook University. He presented his simple election forecasting model at the American Association for Public Opinion Research meetings this spring, and predicted a close-as-can-be election between Obama and McCain, based in part on the New Hampshire primary results and in part on the party holding the White House going into the election.

Part of the explanation he gave is that non-incumbent elections tend to be very close. That made me question whether the political environment applies to non-incumbent election years, and if it does, does it apply as strongly? What do the data suggest?

The historical evidence is thin. There have been only five non-incumbent presidential elections in the modern polling era. I'm largely evaluating the political environment for each using the outgoing president's approval rating, since Gallup has data for that going all the way back to the 1940s, and we know approval is correlated with other political environment ratings such as satisfaction and ratings of the economy.

In 1952, Democratic incumbent Harry Truman had historically low approval ratings and Republican Dwight Eisenhower easily defeated Democrat Adlai Stevenson for president. The political environment explanation would predict a solid Republican win, and that is what happened.
In 1960, Eisenhower was very popular when he left office after eight years (58% approval in the last measurement before the election), but John F. Kennedy and the Democrats won a closely contested election (winning the popular vote by less than a percentage point) to replace Eisenhower instead of choosing his vice president, Richard Nixon. The political climate model does not seem to apply to this election.
In 1968, the Democrats were trying to win an election with an unpopular president (Lyndon Johnson) waging an unpopular war (in addition to other problems going on in the country), but Republicans only narrowly won. A political conditions interpretation would suggest a big Republican win in 1968, not a 1-point win. It's possible that George Wallace's strong third-party showing made things closer than they otherwise would have been.
In 1988, Ronald Reagan was a popular incumbent (51% approval), though he was not as popular as Eisenhower in 1960 or Bill Clinton in 2000. The elder George Bush won what many perceived as a third Reagan term, and by a healthy margin.
In 2000, Clinton had high approval ratings (57% at the time of the election) and national satisfaction was near record levels, yet Al Gore won the popular vote by only the narrowest of margins.
Whereas a political climate explanation seems to work so well in incumbent presidential elections and midterm elections, on the surface it doesn't seem to explain the outcomes of non-incumbent elections that well, in terms of either the margin or the winner.

So what matters? One thing that definitely comes into play is voters' reluctance to keep a party in power for more than two consecutive terms. Since 1952, this has occurred just once (in 1988). So this can clearly be a countervailing and perhaps more powerful force in non-incumbent elections than the political climate.

Of course, idiosyncratic candidate factors can also influence the outcome and the margin, and seem to provide at least some guidance. While 1952 seems to fit the political explanation model well, perhaps the big Republican margin had as much to do with Eisenhower's popularity as the climate. And maybe Kennedy won an election Democrats might otherwise have lost because voters saw him as a much more attractive candidate than Nixon. Perhaps the elder George Bush's election had more to do with the weakness of Democratic challenger Michael Dukakis than the political climate. It would seem to, given the closeness of the 1960 and 2000 elections, in which the outgoing presidents were more popular than Reagan in 1988.

One other possibility is that voters approach each election differently. When deciding to re-elect a president to four more years in office, it makes sense for a voter to evaluate his performance to date, so the choice is largely backward-looking and, thus, the state of the country should be a major influence on that choice.

However, when an incumbent is not running, the election is arguably more forward-looking, and how the country is doing matters less than voters' projections as to how the two major candidates would do as president if elected. If so, then candidate-specific factors such as experience and policy positions (where candidates want the country to go) should matter more than voters' perceptions of how the country is doing or how the incumbent president or his party has performed (where the country has been).

So in past elections in which no incumbent was running (as is the case this year), the political environment hasn't seemed as important to how the election eventually played out in an obvious way. Thus, maybe it's not that Obama is underperforming what he "should" be doing. Rather, pundits' expectations about how he should be doing may just be too high because they are relying on patterns that -- though well-established -- may not apply to all election years equally.