A new Washington Post poll finds that the public souring a bit on Obama’s handling of the economy, while generally disapproving of Obama’s handling of gas prices. The headline: “Gas Prices Sink Obama’s Ratings.” Chris Cilliza and Aaron Blake follow on this:
Make no mistake: The unemployment rate is a telling indicator of Obama’s chances at re-election. But, so too, is the price of a gallon of gas — and whether voters blame the incumbent for the pain at the pump.
Nate Silver tackled this a few weeks ago. He notes that a dramatic shock in gas prices—equivalent to what might happen if, say, the Saudi regime fell—could matter. But absent a shock, here’s his conclusion:
The upshot is this: higher gas prices are important to the extent that they affect things like G.D.P., inflation and unemployment. But there isn’t evidence that they matter above and beyond that.
So how much do gas prices really matter? There is some reason to believe they could: gas prices are a very visible indicator of the economy’s health, one whose impact on voters’ pocketbooks and wallets is easy to see right at the pump. In fact, as research by Stephen Ansolabehere, Marc Meredith, and Erik Snowberg shows, people are pretty accurate at estimating the price of gas in their state. And, unlike other sorts of politically relevant numbers, better educated people aren’t any more accurate than less well-educated people. You don’t have to be following politics closely or be some math whiz to know how much gas costs.
But does the visibility of gas prices translate into real political consequences? My analysis below suggests that it does, although its effects on the election are largely indirect and thus unlikely to rise above the effects of other economic indicators.
Using 50 years the Index of Consumer Sentiment (1960-2011) and over 60 years of presidential approval data (1948-2011), I modeled quarterly consumer sentiment and presidential approval as a function of 6 factors: the previous quarter’s consumer sentiment as well as the percentage change (from the previous quarter) in unemployment, GDP, inflation, the Dow Jones Industrial Average, and oil prices.
The oil prices data are from this series, which is different than what Nate used, but has the advantage of including quarterly data back to the 1940s. For the period from 1976-2011, when quarterly data on gasoline prices is available, there is a 0.73 correlation between the percentage-point changes in the two series from quarter to quarter. That’s fairly high, but it could be higher. So I’ll also report results for 1976-2011 using quarterly gas prices.
1) Do gas prices affect how Americans view the economy?
Oil prices are correlated with economic evaluations, but their effects are very small relative to the other economic variables. For example, a 1 percentage point increase in the unemployment rate is associated with a 2-point drop in consumer sentiment. A 1-point increase in the inflation rate is associated with a 1-point drop in consumer sentiment. A 1-point drop in the Dow is associated with a .34 decrease in sentiment. Gas prices? A 1-point drop is associated with a .06 drop in sentiment. That effect is statistically significant, but just small in magnitude.
If I substitute quarterly gas prices for quarterly oil prices, focusing on the 1976-2011 period, the results are equivocal. A 50-cent increase in the price per gallon from the previous quarter is associated with a –1.3 drop in consumer sentiment, but the effect is not “statistically significant” by conventional standards (t-statistic of –1.13). If one simply uses the actual gas price in that quarter, not the change from the previous quarter, the results suggest that for each additional 50 cents in the price of gas, consumer sentiment falls by –.75. That effect is statistically significant (t-statistic of –1.99)
2) Do gas prices affect presidential approval?
A similar model of presidential approval—one using “fixed effects” for presidential administrations—generates a very similar result. The associations between presidential approval and economic variables like unemployment, GDP, and inflation are large and statistically significant. For example, each 1-point increase in the unemployment rate is associated with a –1.35-point drop in approval (t-statistic=–3.45). The same increase in oil prices has a much smaller apparent effect (–.05-point drop in approval, with a t-statistic of –1.86). Oil prices just aren’t a primary ingredient in approval, once you take into account these other economic variables.
What about gas prices over the period from 1976-2011? Here, the results suggest a stronger effect. Even after controlling for these other economic variables, a 50-cent increase in gas prices from the previous quarter is associated with a –3.5-point drop in approval. Or, to use a relevant contemporary simulation, a 30-cent increase—which is the actual difference in real gas prices in December 2011 and February 2012—is associated with a –2-point drop in approval. (I also find that it is this change in gas prices, not the actual gas price itself, that drives presidential approval.)
3) Do gas prices affect presidential elections?
Nate ran these numbers, and the results were pretty clear.
In a more technical sense, if you put both variables into a regression equation (see example here), there isn’t any additional explanatory power in accounting for gas prices once you’ve already accounted for measures like G.D.P.
My findings, using this different oil prices series, are the same.
However, if we assume that presidential approval plays a role in election outcomes, we can examine a different scenario: how much could higher gas prices affect the election’s outcome by driving down presidential approval?
Take a garden-variety model of the major-party vote in presidential elections from 1948-2008: the percent change in GDP between the first and third quarters of the election year, presidential approval in June of the election year, and a binary indicator of whether the incumbent president is running. This model suggests, unsurprisingly, that the incumbent party does better when the economy is growing, the presidential is popular, and the incumbent himself is running. The estimated effect of approval: a 4-point increase in presidential approval is associated with a 1-point gain in vote share.
So let’s now simulate the potential effect of gasoline prices via its effect on approval.
- Scenario #1: historical high. The real price of gas in February was 3.58. Let’s say that over the next three months or so, that increases to its historical (June 2008) high of 4.27. I use a three-month window since my model of presidential approval was based on quarterly data. This 70-cent increase in gas prices is predicted to drive down approval by 5 points. A 5-point decline in approval would lower Obama’s vote share by about 1.2 points.
- Scenario #2: FIVE DOLLAR GAS. If gas prices spike to $5 per gallon, an increase of $1.42, then the president’s approval rating would decline by about 10 points. That would cost him 2.4 points of vote share.
- Scenario #3: an actual forecast of gas prices. The U.S. Energy Information Agency actually forecasts the price of gas. Over the next 3 months, it is expected to increase by about 40 cents—to a maximum of about $3.96 in May and falling throughout the summer and fall. Let’s assume that it’s the maximum that matters, and not the subsequent decline. A 40-cent increase is equal to about a 2.8 decrease in approval, which would in turn lower Obama’s vote share by about .70 points.
These simulations come with a huge caveat: everything else is being held constant. In essence, I am assuming that only gas prices, and nothing else, are changing. That’s not a very plausible assumption, and it points to a crucial limitation on the effect of gas prices—the same one that Nate identified. Gas prices are unlikely to matter unless (a) other economic indicators are equivocal and (b) gas prices continue to rise sharply. High gas prices are typically accompanied by other economic challenges, making it difficult to blame gas prices if Obama loses in 2012. So if I were advising the president, I’d certainly say that he should keep his eyes on gas prices, but I doubt that his presidency will be won or lost at the pump.
[Cross-posted at The Monkey Cage]
[Photo credit: Tea Party Tribune]
Feed the Political AnimalDonate
Washington Monthly depends on donations from readers like you.