Customers seem to pay more attention to the leftmost digits on price tags, which is why stores sell many of their items at prices ending in 99 cents. A candy shop owner may decide to price a chocolate bar at $2.99 instead of $3 because changing the leftmost number from a three to a two seems to affect a customer’s decision to buy that candy, even though the actual price difference is just one cent.
People may feel that stopping to tally up cents may not be worth their time and effort. Instead, they prefer to rely on simple rules of thumb, such as focusing on numbers to the left of a decimal point. When small amounts of money are involved, people may be forgiven for not paying attention to every digit on a price tag.
But a study by Chicago Booth associate professor Devin Pope, with Nicola Lacetera of the University of Toronto and Justin Sydnor of the University of Wisconsin, finds that people exhibit a similar left-digit bias when shopping for cars, expensive items that people typically spend a lot of time thinking about before purchasing. The stakes are much higher than they are for small candy purchases, and inattention to detail could cost each car buyer much more than 99 cents.
In the study “Heuristic Thinking and Limited Attention in the Car Market,” Pope, Lacetera, and Sydnor analyzed 22 million sales transactions at one of the largest operators of wholesale used-car auctions in the United States. There were two main types of sellers at the auctions: dealers who brought in cars that they did not want to sell in their own lots, and companies such as rental-car and leasing agencies that owned fleets of cars. Licensed used-car dealers were the only type of buyer. The highest bidders received the cars, which they took back to used-car lots and resold.
The number of miles registered on a car’s odometer is an important factor in determining the amount that a car buyer is willing to pay, and buyers generally pay more for cars with fewer miles. But Pope, Lacetera, and Sydnor found surprisingly high prices for cars with odometer values just under a 10,000-mile threshold, evidence that buyers at auction tend to focus on the left digits of the odometer.
For example, a car with an odometer reading of 79,900 miles sold on average for about $200 more than a similar car with 80,000 miles, controlling for observable characteristics such as make and model. Used car dealers were willing to pay more for a car that had a 79 rather than an 80 on the left side of the odometer. This large premium vanished between thresholds—a car that had racked up 79,800 miles sold on average for only $10 more than a car that had 79,900 miles. There were also irregular price drops at the 1,000-mile thresholds, indicating that buyers were overlooking right-hand digits at other points, as well.
Astute car owners who are aware of buyers’ left-digit bias may want to sell their cars to dealers shortly before their odometers hit 10,000-mile thresholds. Indeed, the researchers found large spikes in the quantity of cars brought to auction with odometer values just under each 10,000-mile mark. This was especially true for cars sold by dealers.
The Inattentive Consumer
Could left-digit bias be overstated? Consider that vehicle owners who were careful enough to sell their cars right before their odometers hit a 10,000-mile threshold might also be people who were especially careful about maintaining their cars. If so, then the price premiums that Pope, Lacetera, and Sydnor observed near the 10,000-mile thresholds could have been partly due to the fact that these cars were in better condition than others.
To make sure they were not overestimating their findings, the researchers ran their analysis separately for those cars sold by dealers and those that were brought to auction by so-called fleet companies. Dealers sold cars that came from individual car owners and were likely to have differing amounts of wear and tear. By contrast, there was reason to believe that cars brought to auction by fleet companies were of similar quality. Rental-car companies, for instance, probably cared for all of their cars in the same way.
If the price-drops at the 10,000-mile marks were much larger for dealer cars than for fleet cars, then one would suspect that the price differences could have been partly driven by unobservable quality effects. While the researchers found slightly bigger price changes at the thresholds for dealer cars, the changes were largely similar for both groups: dealer cars had an average price drop of approximately $173 at each 10,000-mile threshold, while fleet cars had a price drop of $157.
Another concern: did expiring car warranties produce the observed price-drops at the 10,000-mile thresholds? A car buyer concerned about buying a lemon could value a warranty, even one with only a few hundred miles left on it. A car that had a warranty through 10,000 miles, say, would have a higher price than one without a warranty. However, the authors found that most of the cars that had odometer readings just under 10,000-mile marks actually did not have warranties, which suggests that expiring warranties did not affect the results.
Yet another possibility is that odometer tampering may have affected the authors’ results. A seller could have manipulated an odometer to set it just below a 10,000- mile threshold in order to ask a higher price. This could explain the large spike in the volume of cars brought to auction at each threshold. However, the authors found no evidence of cheating—cars sold before a 10,000- mile mark were not older than expected. Moreover, if tampering had occurred, that could have brought prices down—not up—because buyers would have been suspicious of odometer readings.
The unit of measure in which the numbers were reported didn’t matter either. Using a smaller dataset for Canadian car auctions, where odometer readings were in kilometers rather than miles, the authors found significant price jumps at the 10,000-kilometer marks, but not at the 10,000-mile marks. Whereas in the US car dataset, a car’s value naturally declined as it racked up miles on the odometer, and almost one-third of the depreciation a car experienced as a result of mileage increases occurred at the 10,000-mile thresholds, where a buyer’s left-digit bias came into play.
The results suggest that less savvy customers are paying a high price for this bias. Based on the study’s dataset alone, the difference between the cars’ expected price tags and their actual selling prices, inflated by left-digit bias, was $2.4 billion.
Consumers, not dealers, ultimately paid this higher price. The study found that experienced dealers paid more at auction for cars with odometer readings near 10,000- mile thresholds, and they passed on the higher prices to their customers, who willingly paid those premiums.
Past research has examined consumers who fail to process all available information about products, but in many cases, consumers fail because certain information about those products is somehow hidden, or if the relevant information is not hidden, people don’t know how to use it. For example, in a previous study, Pope found that many people pay close attention to rankings of universities and hospitals, even though the information used to create those rankings is readily available.
In this study, by contrast, each car’s odometer reading was fully disclosed. Moreover, consumers were clearly looking at the odometers when determining how much to pay. And yet car buyers still exhibited a strong tendency to ignore numbers to the right—valuable information that was right in front of them.
The tendency to focus on leftmost digits may be rational assuming that consumers are willing to pay more than $100 to save themselves time and effort. A consumer has to expend both of those to look at all digits and to do additional computations. But whether rational or irrational, the left-digit bias has implications that go beyond candy bars or the used-car market. Hiring and admissions decisions are typically based on grade point averages and test scores. Investors and other parties evaluate companies based on revenue figures and other numbers presented in financial reports. And consider how the public reacts to government programs that involve additional spending. A bill that legislators say will cost taxpayers $497 billion may be better received than one predicted to cost $503 billion.
Devin Pope is Associate Professor of Behavioral Science and Robert King Steel Faculty Fellow.