People are honest about how they spend their money

In 2008, early in the financial crisis that caused the U.S. housing market to collapse, the federal government distributed more than $100 billion to roughly 130 million eligible taxpayers through the Economic Stimulus Act. Individual tax filers received $600 and joint filers received $1,200, with an additional $300 for each child who qualified for the child credit.

The distribution provided MIT Sloan professor Jonathan Parker and University of Pennsylvania professor Nicholas S. Souleles with a unique opportunity to compare two methods for measuring consumer behavior: revealed-preference and reported-preference.

In a new paper [PDF], the researchers found the two methods largely returned the same results — with one important caveat: Economists should not take it literally when consumers report spending a specific dollar amount.

That being said, if someone reports mostly saving their money, it is safe to assume that is what they are actually doing. Ultimately, the researchers believe that economists can use reported-preference for predictive reasons, but using the two methods in conjunction would work best.

Revealed versus reported-preference.

Economists have long relied on the revealed-preference approach to learn about behavior. They find situations in which people are treated differently, either by an employer or government policies, and compare the differences in their behavior. Economists can then infer that different behavior was caused by the different treatment.

Increasingly more, though, researchers, policymakers, and businesses are forgoing that method for something easier and faster: asking people how they behaved in certain situations, or how they would behave in hypothetical situations. This is the reported-preference approach.

Using reported-preference raises some questions, though. Is what people report accurate? Is it reliable?

“There are a lot of ways in which those questions might not reveal the truth,” Parker said. But in some situations, it is the only option. Using the revealed-preference methodology requires that the situation researchers are studying to have already happened, and that different people were subjected to different treatments.

Additionally, with an uptick in its use — such as the New York Federal Reserve conducting a survey [PDF] on how much people would save when facing different hypothetical interest rates — reported-preference “might even be used to guide policy,” Parker said. That is why economists needed to know if they can trust it.

To test this, the researchers added survey questions to two existing consumer surveys that collect information about how people spend: the Consumer Expenditure and the Nielsen Consumer Panel. They asked about when people received the stimulus, how much they received, and whether they used it to mostly increase spending, increase savings, or pay off debt. In the Nielsen Consumer Panel survey, they also asked how much people spent from the stimulus. They then compared these results to revealed-preference behavior.

Where it works and where it doesn’t.

Parker and Souleles found that the reported-preference data was consistent with the revealed-preference data in two ways: households that reported mostly spending did increase their spending after receiving the stimulus, and both methods produced similar estimates of average spending across the population. But, Parker said, these findings were right, but maybe not for all the right reasons.

“The main measurable difference across people is completely missed by this reported-preference,” Parker said.

The problem is that the data did not differentiate between how people with more or less liquid saving and disposable income spent. For instance, people with lots of liquid wealth reported mostly saving or mostly spending at similar rates to people with little in liquid assets, counter to both revealed-preference data and economists’ understanding of basic consumption models.

There are two possible reasons for this. First, people may be reporting their behavior relative to how they typically spend. This means that someone who typically does not spend money as it arrives — but who did spend slightly more after the stimulus payment — may have reported they “mostly spent,” even though they still may have saved most of the stimulus.

Second, people with lots of liquidity or income who did not actually change their spending behavior when the payment arrived may have reported mostly spending their stimulus payments when in fact they spent it on something they would have bought even without the additional money. “They mentally allocated this spending to something that they would have spent on anyway,” Parker said.

Going forward, the researchers concluded reported-preference responses are informative for predictive exercises. But if someone reports spending a specific dollar amount, that should not be taken literally without further evidence.

Parker thinks that using both revealed and reported preference together — or “conjoined analysis” — may prove the most beneficial. “It would be extremely useful to be able to use the information from the reported behavior to extrapolate confidently beyond what you can see in the revealed-preference results,” he said.