Pulse Points Fall 2022: Responses to Record-High Gas Prices

The national average price for a gallon of gas increased over 40% from March to July 2022. Economic theory predicts consumers will respond to price increases in at least one straightforward way: As gas prices increase, people will spend less on gas as a function of their preferences and budget constraints. (How drastically they make this adjustment on average is an empirical question.) However, it’s possible that consumers respond to price increases in other ways as well.

Take this clip from the sports talk show “NBA on TNT” as one lighthearted example. In it, one of the hosts, Kenny Smith, says he doesn’t want to buy an inefficient vehicle because it would cost him $80 to fill the tank. His co-host, Shaquille O’Neal, argues that Kenny can avoid spending that much money by not letting the tank get less than three-quarters full.

One charitable interpretation of Shaq’s argument is that he seems to be claiming that purchasing gas four times for $20 feels less bad or is easier to manage than purchasing gas once for $80. Shaq is not the only person known to express this kind of sentiment. An article published in The New York Times in fall 2021, includes an interview with a man at a gas station who “said he had taken to getting smaller amounts of gas twice a week to soften the blow to his bank account.”

These accounts resonated with some of my own experiences as a driver. For example, I remember buying exactly $20 worth of gas at a time in high school. Purchasing smaller amounts of gas per transaction in this way gave me a sense of control over my cash flow.

Taken together, these anecdotes represent far less straightforward responses to price increases than simply cutting back on driving overall or switching to lower-grade gasoline. Do rising gas prices influence people to purchase gas more often? If so, what motivates this behavior?

One explanation could be a cognitive bias like the “pain of paying” that results from loss aversion.¹ This could influence people to act as if dollars spent on gas above a certain threshold (say, $20) are more costly than dollars spent below that threshold. Consumers may also engage in a form of “mental accounting,” budgeting a modest amount to spend at the pump on a visit-by-visit basis, regardless of their actual budget constraints or willingness to pay for gasoline.² It could simply be that some consumers are “anchored” around paying a specific amount to fill up their tanks, and are averse to spending more than what they are used to spending at each visit to the pump.³

It is also possible that consumers will use this type of budgeting strategy if they face liquidity constraints (i.e., the inability to complete a transaction above a certain dollar amount). Those who are not Financially Healthy are more likely to face these challenges, as indicated in their FinHealth Scores® by negative cash flow, difficulty paying bills on time, insufficient short-term savings, and poor credit. If people who are not Financially Healthy respond to price increases by splitting their gas purchases into more frequent, smaller transactions, it could suggest that liquidity constraints play a role in how people manage their gas consumption.

How did Financially Healthy and Unhealthy People Respond to Record-High Gas Prices?

There are two questions we can directly answer with Financial Health Pulse® data to lend some insight:

1. 1. Did the frequency of gas station transactions increase as prices increased in our sample overall, after accounting for changes in total gas consumption?
   2. Did Financially Unhealthy people increase their number of gas purchases faster than Financially Healthy people, after accounting for changes in total gas consumption between the two groups?

We answer these questions with the following data:

• • Participants’ credit card and checking account purchases that occurred at gas stations, which we are able to identify using Financial Health Pulse transactional data.
  • The FinHealth Scores of these participants (calculated from their Financial Health Pulse survey responses), which we use to identify participants as either “Financially Healthy” or “Financially Unhealthy.”⁴
  • National average gas retail prices, which are publicly available on a week-by-week basis from the Energy Information Administration.
  • Rough estimates of how much gas was purchased per transaction, calculated by dividing dollars spent on that transaction by the average gas price in the nation that week.
    

Total gas consumption in our sample decreased steadily overall as prices rose from February 2021 through July 2022 (Figure 1). The downward trend was similar for both Financially Healthy people and Financially Unhealthy people, and each group’s sensitivity to price over this period was statistically indistinguishable.⁵ This shows that both Financially Healthy people and Financially Unhealthy people in our sample reduced their gas consumption at similar rates as prices increased.

Figure 1. Gas consumption decreased overall as prices rose.
Daily average of monthly gas purchases (30-day rolling total, in gallons) and 30-day rolling national average retail gas price.

Notes: Shaded areas are 95% confidence intervals plotted around the mean activity estimated on each day. Lines have LOESS smoothing applied as a visual aid. n = 353 people observed on each day (125 Healthy, 228 Not Healthy).

Despite a steady decline in the amount of gas purchased, there were no dramatic changes in frequency of gas station visits in our sample overall (Figure 2). However, each group’s trend in number of trips to the pump per month after March 2022 suggests that when gas prices began their steepest climb, people who are not Financially Healthy may have begun increasing their number of monthly gas refills relative to Financially Healthy people.⁶

Figure 2. Frequency of gas station purchases by financial health tier may have started to diverge in March 2022.
Daily average of monthly gas station transactions (30-day rolling total) and 30-day rolling national average retail gas price.

Notes: Shaded areas are 95% confidence intervals plotted around the mean activity estimated on each day. Lines have LOESS smoothing applied as a visual aid. n = 353 people observed on each day (125 Healthy, 228 Not Healthy).

The estimated number of gallons of gas purchased per transaction (i.e., the average transaction size) may have also diverged between Financially Healthy and Unhealthy people. Figure 3 shows that during the period with the steepest price increases (March 1-July 31, 2022), the amount of gas purchased per transaction stayed roughly steady for Financially Healthy people and decreased slightly for Financially Unhealthy people in our sample. This is what we would expect to see if Financially Unhealthy people responded to price increases by purchasing gas in smaller increments. Together, these two visual trends suggest that people who are not Financially Healthy might have responded to increases in the price of gas not only by cutting back on the total amount of gas they purchased per month, but also by increasing the frequency and reducing the size of their purchases.

Figure 3. Financially Unhealthy people purchased smaller amounts of gas per transaction when gas prices were highest.
Daily average of estimated gallons of gas purchased per transaction and 30-day rolling national average retail gas price.

Notes: Shaded areas are 95% confidence intervals plotted around the mean activity estimated on each day. Lines have LOESS smoothing applied as a visual aid. n = 353 people observed on each day (125 Healthy, 228 Not Healthy).

Historical trends in the number of trips to the gas station and the amount purchased per trip offer clues to how gas purchases varied for each group over time, but other methods are needed to estimate the relationship between price and consumer behavior directly. Regression analysis allows us to model the relationship between price and transaction frequency for Financially Healthy and Unhealthy people to test whether each group responded to price increases differently at different price points.⁷ 

Our results show that each group’s predicted relationship between price and transaction frequency is distinct.⁸ This is true even after taking into account differences in overall gas consumption between the two groups and seasonal differences in demand for gas. Predictions from our regression model with these controls taken into account are plotted in Figure 4.

Overall, people in our sample increased the frequency of their gas station visits by about 0.18 transactions per month for every $1 price increase. Those who were not Financially Healthy increased their trips to gas stations when gas prices were very high, while those who were Financially Healthy maintained roughly the same number of trips to the pump per month once prices reached $4 per gallon. In other words, Financially Unhealthy people increased the frequency of their gas purchases relative to Financially Healthy people when gas prices were near $5 per gallon, holding total gas consumption constant.

Comparisons of the slopes of each model at different price points confirm that when gas prices were highest (between $4.50 and $5 per gallon), Financially Unhealthy people in our sample were increasing their number of purchases per month at a faster rate than Financially Healthy people (between 0.24 and 0.45 purchases per month faster for each $1.00 increase).⁹

Figure 4. When gas prices were very high, Financially Unhealthy people increased the frequency of their gas purchases relative to Financially Healthy people.
Estimates of transaction frequency by price, after controlling for total gas consumption and seasonality.

Notes: Plot shows predicted transactions per month for each group, produced by an OLS regression model with a quadratic term for price and controls for gallons purchased per month and monthly indicators to control for seasonality. n = 192,738 person-day observations. Standard errors clustered at the person level (n = 353). Shaded areas are 95% confidence intervals. Though confidence intervals overlap slightly, the estimated number of transactions per month are statistically significantly different at lower price points (~$2.50 to ~$3.25) and higher price points (~$4.50 to ~$5). Contrasts of marginal effects show that the slopes of each line are statistically significantly different between ~$4.50 and ~$5.