The Historical Analysis of Presidential Electability: Does Approval Trump All?

What is the effect of presidential approval leading up to a general election on the likelihood a president is reelected or defeated? Further, when dealing with lame ducks, does a negative approval thus punish the sitting president’s party’s nominee? If the president has a net approval or an approval rating near or above fifty percent, he or she should be reelected, or the nominee of their party should win. Further, the higher the net approval, the larger the electoral victory should be. If voters use any form of retrospective voting, they should not reward a president or his or her party with an electoral victory if their approval is underwater. To test this, we will analyze approval rating data of presidents since Franklin Delano Roosevelt (independent variable) to determine if voters reward or punish the president or party (dependent variable) in electoral victory or defeat. Since the 1944 presidential election, the hypothesis holds in all cases except three: the general elections of 1976, 2000, and 2016. We will try to answer this based upon historical context of events happening during those times that may impugn the accuracy of the hypothesis during certain circumstances.

The data were collected from four sources. For approval polling, the Gallup poll was utilized with data being retrieved from UC Santa Barbara’s The American Presidency Project data file (Woolley and Peters 2018). The file contained approval data since 1944-2018, it was supplemented with the Gallup approval up until present via FiveThirtyEight’s polling data collection (Silver 2019). The Electoral College data was gathered from a Wikipedia table that analyzed presidential elections by Electoral College margin (Wikipedia 2019). According to the citation, the data was from Michael Sheppard, a MIT researcher. Finally, for Gross Domestic Product (GDP), data was gathered from the Federal Reserve Economic Data bank via the Federal Reserve Bank of St. Louis (U.S. Bureau of Economic Analysis 2019). These data will help find correlational or causal data that supports or negates the hypothesis. All the sources used have used the data for other empirical projects, thus, to a level of certainty we can deem it accurate and trustworthy. The approval data provides the start and end dates of polling, approval, and disapproval—net approval is calculated from these two metrics. The Electoral College data provides the election date, winner, winner’s total, and runner up’s total. The GDP data includes the quarterly dates and GDP growth percentage.

This plotting highlights the main hypothesis argument, that is, if the president has a net approval or an approval rating near or above fifty percent, he or she should be reelected, or the nominee of their party should win. The first vertical dashed abline represents the 1944 presidential election and the eighteen following it represent every presidential following it until the present. As interpreted from the plotting, there are three instances where the data does not fit the hypothesis. We shall discuss why perhaps it does not and why the hypothesis may only show a factor of electoral success and may not be monolithic. The first abline line that contradicts the hypothesis is the election of 1976. Here we saw incumbent President Gerald Ford lose to peanut farmer, Jimmy Carter. One factor that the model does not directly consideration but would be indirectly factored into approval would be scandal. President Ford had pardoned former President Nixon for his crimes and many certainly had a sour tongue

because of that. However, he did experience approvals above fifty percent—perhaps because of economic success or other factors. Secondly, the election of 2000. Here was saw current Vice President Al Gore lose to Texas Governor, George W. Bush. Out of the three divergent elections this one is the hardest to explain. Perhaps Bush’s father played a role in the election or that the country did not see Clinton’s successes being directly correlated to the policies of the Democratic Party. Finally, the election of 2016, where former Secretary of State Hillary Clinton lost to business mogul, Donald J. Trump. Many factors certainly apply to this divergence—the negative view of both candidates, the shaky approval of President Obama, and Hillary Clinton’s involvement and interactions with the government. Other than those three instances, we see the hypothesis hold across the board.

As hypothesized, the higher the net approval, the larger the electoral victory should be. If voters use any form of retrospective voting, they should not reward a president or his or her party with an electoral victory if their approval is underwater. Thus, periods leading up to elections where the president had either high net approval or disapproval has boosted the electoral success of his party. Further, analyzing the three instances

where the hypothesis does not align from the last plot, we can see their Electoral College victories are in fact among some of the lower end victories, with 1976 and 2000 within the first quartile and minimum score, respectively, and 2016 just slightly above the first quartile.

This visualization offers a counterview to purely approval rating and attempts to separate some of the factors that make up an approval rating, e.g. economic performance, scandal, consumer confidence, foreign policy, etc. As we can interpret, the trends in this correlation are less succinct than the approval rating. At times, GDP could be high but approval low. Thus, it is not a positive to positive relationship but are connected. Let us analyze the two of three elections where the data does not match the hypothesis—removing 1976 because of the scandalous nature of the Nixon and Ford Administrations. In 1999 leading into 2000, the economy began to stagnant and GDP growth plummeted. In the words of Clinton strategist, “It’s the economy, stupid.” Therefore, if approval is a strong predictor but not monolithic of simply who wins the presidency, we can expect other factors like GDP growth rate to be a telling predictor of the winner and loser. Further, in 2016, we see a similar story. President Obama had a lukewarm

approval rating with below par GDP growth—a key talking point of the Trump campaign. As we know, Hillary Clinton lost that election even with President Obama having a net approval rating.

From this analysis, we have learned presidential approval rating has a strong predictor on whether a president or his or her party wins the next presidential election. This at least suggests voters use retrospective voting and typically do not reward a president or his or her party with an electoral victory if their approval is underwater. Simply, if a voter drives on potholes every day, they would have a negative view of their mayor or city council and would likely vote against them unless the potholes improve. Thus, voters reward results that improve their lives with approval and then cast votes for those they approve of. We are in a challenging and turbulent time as a country. However, it is certainly apparent that the argument that out democracy or constitutional republic is broken is simply not true if three out of nineteen presidential elections have resulted in voters acting in a retrospective way.

Further, building from the hypothesis, we have seen that the higher a president’s approval or disapproval is, the higher the Electoral College victory is. This idea is complementary with the main hypothesis and highlights again how voters are responsive to stimuli in the political process. This analysis was one in which I thought many questions that become overly complicated could be answered, partially or fully, with a simple hypothesis. Many presidential election predictors use a model of a dozen factors to make their prediction. Rather, as we have laid out here, the vast majority of times, if the president’s approval is above fifty percent, we would expect his or her party to prevail in the upcoming presidential election. Hopefully because of the importance of empirical polling data in circumstances like these, we continue to see vast improvements in the field of analytical political science research.

Works Cited

Wikipedia 2019. “List of United States Presidential Elections by Electoral College
Margin.” Wikipedia Foundation. Retrieved December 8, 2019 https://en.wikipedia.org/wiki/List_of_United_States_presidential_elections_by_Electoral_ College_margin

Woolley, John and Gerhard Peters. 2018. “Presidential Job Approval.” The American Presidency Project. Retrieved December 8, 2019 https://www.presidency.ucsb.edu/statistics/data/presidential- job-approval

Silver, Nate. 2019. “How Popular Is Donald Trump?” FiveThirtyEight. Retrieved December 8, 2019 https://projects.fivethirtyeight.com/trump-approval-ratings/

U.S. Bureau of Economic Analysis 2019. “Real Gross Domestic Product.” FRED. Retrieved December 8, 2019 https://fred.stlouisfed.org/series/A191RL1Q225SBEA#0