The cost of filling up your car with gas has a huge impact on how American voters view their country’s commander-in-chief. Recent research published in American political research The researchers found that gas prices act as one of the strongest predictors of presidential approval ratings, operating in an uneven pattern where initial price increases cause the most political damage. The survey found that voters judge presidents primarily based on the direct economic pain felt when filling up the tank, rather than viewing fuel costs as a broader economic warning sign.
Political commentators and election strategists frequently discuss the electoral impact of energy prices. Opinion polls are expected to fluctuate as economic conditions change, often leading to intense political messaging around the cost of living. Rangan Gupta, an economist at the University of Pretoria in South Africa, along with colleagues Christian Pierziok and Aviral Kumar Tiwari, wanted to formally test this relationship. They sought to determine exactly how fuel costs affect the American public’s support for the president.
Two main academic theories exist to explain why voters punish or reward presidents based on driving costs. The first is the pocketbook mechanism, which suggests that voters are responding to the immediate personal economic losses they experience when fuel becomes more expensive. According to this theory, voters will judge governments based on the shrinkage of their bank accounts.
The second theory is the sociotropic mechanism. This idea suggests that the public views energy costs as an easily visible source of information about the overall health of the national economy. Because people see gas prices advertised on large billboards during their daily commute, they may use those numbers to infer whether the large financial system is thriving or declining.
Previous studies of this public opinion movement have often yielded contradictory results. Many early mathematical models assumed linear cause and effect, where for every $1 increase in fuel costs, the president’s popularity would logically decrease as well. The research team suspected that human behavior and media coverage did not function in such a simple straight line. They wanted to test irregular thresholds at which voters’ attitudes suddenly change.
To analyze these changing public attitudes, the researchers used an advanced computational technique called random forests. Random Forest is a machine learning algorithm that builds hundreds of individual decision trees. These decision trees function like complex flowcharts that test different paths through the data, categorize information, and predict outcomes.
By averaging the results of hundreds of these individual decision trees, the algorithm avoids placing too much weight on a single piece of misleading information. This allows computers to find irregular patterns hidden in large data sets. The researchers chose this method because it does not require humans to guess the shape of the pattern in advance and allows them to identify non-uniform behavioral changes.
The researchers collected monthly data over a 50-year period from October 1973 to December 2023. They compiled figures on inflation-adjusted retail gasoline prices, global oil costs, and presidential approval ratings meticulously compiled by the polling organization Gallup. Gallup polls have asked the exact same questions about presidents’ job performance for decades, providing a very consistent measure of public opinion.
Researchers didn’t come up with these broad economic indicators out of thin air. They utilized a large economic database consisting of hundreds of financial and macroeconomic time series. This extensive collection included hard data on country employment hours, manufacturing output, real estate building permits, and international trade volumes. By condensing this vast amount of information into eight core elements, the researchers ensured that the algorithm had a complete picture of the U.S. economy.
Beyond standard economic performance, the algorithm also took into account broader social unrest. The researchers incorporated formal indicators of macroeconomic uncertainty and financial market volatility. These uncertainty indexes capture unexpected shocks to the economy and reflect the general anxiety that consumers and businesses feel about future financial conditions. The inclusion of these measures was critical in testing whether gasoline prices merely act as a proxy for general economic instability.
To fill remaining gaps in public opinion predictions, the team also included a database that tracks geopolitical actions and threats drawn from historical newspaper archives. These indicators aggregate the number of news articles that focus on international conflicts, nuclear threats, and acts of terrorism. Before running the model, the researchers specifically separated out the cost of global unrefined crude oil from the retail price of gasoline.
Although international crude oil serves as the main feedstock for vehicle fuel, retail prices also reflect domestic shocks. Local refining capacity, local taxes, and seasonal operating demands all change the final price consumers pay at their local station. By isolating retail costs, researchers were able to focus on the exact numbers that voters see and pay.
Machine learning algorithms reveal that the relationship between retail gasoline prices and public approval is highly nonlinear. When gas prices are low and begin to rise for the first time, the negative impact on the president’s popularity is significant. Researchers attribute this precipitous drop to an initial surge in news media coverage that suddenly alerts previously unattended voters to a decline in purchasing power.
If energy prices continue to rise beyond a certain threshold, the penalty for the president will be reduced. Public approval ratings have bottomed out and remain at a relatively low level despite additional costs. At this advanced stage, dissatisfied citizens are likely accustomed to high costs, and continued negative publicity has not provided any new shockwaves to their political opinions.
Setting aside past presidential approvals, inflation-adjusted gasoline prices have emerged as a very strong predictor of future approval ratings. The algorithm was able to use historical fuel prices to predict out-of-sample changes in presidential popularity. This shows that fuel prices contain actual forecasts and pass rigorous tests of forecast accuracy, rather than simply matching historical trends.
Because fuel costs were found to be a key predictor, independent of other broader indicators of economic health, the researchers argue that their results strongly favor the cash-in-pocket mechanism. If voters had simply used gas prices as a social proxy for inferring the overall state of the economy, the algorithm would have relied heavily on the broader financial data provided. Instead, the direct price of filling the tank retained its unique predictive power. This means that voters are holding political leaders accountable for direct personal economic costs.
Although the statistical link between gas pumps and the Oval Office is incredibly strong, US presidents typically have little power over global energy markets. Because the cost of energy is determined by the global balance of supply and demand, political leaders have limited options to quickly change prices.
Presidents may temporarily suspend domestic taxes or adjust import tariffs, but executive power generally cannot make large-scale systemic changes. Because of this dynamic, presidents are likely to continue to face political opposition to global energy changes that are beyond their direct control.
The researchers suggest that these political realities make a strong case for a transition to renewable energy sources. Reducing a nation’s dependence on fossil fuels could ultimately protect domestic consumers from sudden global supply shocks. Conversely, achieving this type of green energy transition may protect future political leaders from unpredictable voter backlash caused by international energy markets.
Future research should examine how energy costs predict longer-term political outcomes. Current analysis focuses on predicting changes just one month ahead. Exploring alternative forecasting models and extending the forecasting window can help political scientists understand exactly how long it takes for price shocks from the pump to register in the national election booth.
The study, “U.S. Gasoline Prices and Presidential Approval Ratings,” was authored by Rangan Gupta, Christian Pierdzioch, and Aviral Kumar Tiwari.
