The ubiquity of computers and data available on-line have made it possible for statistically savvy non-politicians to engage in numerical political science. Recently, Professor Pollkatz has drawn well-deserved attention to the relationship between Bushs presidential approval rating and gasoline prices. As gasoline prices rise, President Bush’s job approval rating decreases. As the prices fall, Bushs approval rises. Of course, the mere fact that there is a correlation between the two does not prove a casual link. Nonetheless, it is reasonable to suppose that gasoline prices affect the popular perception of how the economy is fairing and consequently the approval of the president. If gasoline prices are high, people are reminded weekly at the gas pump.

This relationship takes on contemporary importance since gasoline prices are currently falling. If the relationship holds, Bushs approval should rise and perhaps affect the prospects for Republicans in the mid-term elections about a month-and-half away. Indeed, as prices have declined there does appear to be a modest improvement in presidential approval over the last week.

To make the gas-price-presidential-approval relationship clear, Pollkatz plots a composite presidential approval index as a function of time on the same graph as the scaled reciprocal gasoline price. As the gasoline price goes up, his index goes down. This allows the presidential approval and Pollkatzs price index to track each other on similar numerical axes. The observation that the two quantities track is quite clear, but Pollkatz does not provide (or I could not find) the actual correlation statistics at his site.

To perform my own statistical analysis, I pulled down Pollkatzs composite approval data which he based on a combination of a number of publicly available polls. I also retrieved semi-monthly prices for regular-grade unleaded gas from the Department of Energy. Rather than plotting both presidential approval and gasoline prices as a function of time, the graph below shows a scatter plot of presidential approval as a function of gasoline price.

This way of displaying the data re-enforces some intuitive notions. First, there is general relationship between gasoline prices and presidential approval. Second, there appears to be two regimes of importance. When gasoline prices are greater than about $1.75 per gallon, presidential approval is strongly correlated to gasoline prices. Once gasoline prices fall below $1.75 per gallon, gas prices become less of a concern and are less associated with presidential approval ratings.

When considering all the data for the Bush presidency, the square of the correlation coefficient relating gasoline price and presidential approval is 0.54. This implies that about 54% of the variations in presidential approval can be linearly related to the price of gasoline. However, in the next graph, we only include gasoline prices larger than $1.75 per gallon. For these higher prices, the correlation coefficient is significantly larger, about 0.76. Thus, 76% of the presidential approval can be explained by gasoline prices.

What does it imply for our current situation? If we believe the linear relationship, for every 10 cent decrease in the price of gasoline, Bushs approval percentage will increase by 1.2%. Since August 12, 2006, the price of gas has fallen nearly 50 cents. We could expect roughly a 6% improvement in presidential approval, roughly consistent Rasmussens daily tracking data. If gasoline prices decrease to near $2.00 per gallon, Bushs approval could cross the 50% point, close to what it was when elected to second term.

It is unclear how much gasoline prices will drop in the near future or even if they will turn around and increase. It is also unclear whether any improvement in Bushs approval rating will measurably improve prospects for Congressional Republicans. It is clear that from the perspective of an incumbent, falling gasoline prices are preferable to rising gasoline prices.