A central idea in modern macroeconomics is the permanent income hypothesis. The basic idea is as follows. Suppose that you could dichotomize your income into a permanent component and a temporary component. The permanent income hypothesis suggests that you would base your consumption decisions on the permanent component.
Why would people behave this way?
Well, individuals want to smooth the marginal utility of their consumption over time. To understand this, consider the following example. Suppose that you varied your consumption proportionately with your current income and that your income fluctuated significantly from year to year. This would imply that your consumption would be high in years when your income was high and low in years when your income was low. However, if consumption is subject to diminishing marginal utility, this would mean that the marginal utility of consumption in high income years is less than the marginal utility of consumption in low income years. So wouldn’t it be nice to take some consumption from your high income years (when the marginal utility is low) and transfer it to the low income years (when the marginal utility is high)? Yes, because your lifetime utility would be higher. Fortunately, you can do this by adjusting your savings behavior in response to temporary fluctuations in your income over time (note that this includes borrowing behavior, which is just negative savings).
So this sounds reasonable, but it is important to think about why the permanent income hypothesis might not hold.
Given our discussion, one obvious reason pops up. What if some fraction of consumers are subject to credit constraints (i.e. either limits or lack of access to borrowing). Individuals who face borrowing constraints might find themselves unable to borrow following a reduction in their income. If this is true, individuals might not always be able to smooth the marginal utility of their consumption over time.
Some have posited other, “behavioral” reasons why the permanent income hypothesis might not hold. For example, maybe people are myopic and don’t plan adequately for the future.
So how do we know if the permanent income hypothesis is a good guide in thinking about consumption decisions? Well, we have to go to the data.
Now suppose that you wanted to test the implications of the permanent income hypothesis. There are a number of ways you might do this. You might test the cross-sectional predictions of the model outlined by Milton Friedman. You might try to estimate consumption Euler equations with aggregate data. Or you might find identify periods of time in which people experience a significant decline in income and see what happens to their consumption behavior.
The evidence testing Friedman’s predictions with cross-sectional data seem to support the permanent income hypothesis. Estimates of the consumption Euler equation do not. (However, as John Seater points out, this is likely due to problems with aggregation and not the theory itself since aggregation imposes pretty strong implicit assumptions about households.) Finally, the work focusing on periods of significant declines in income seems to show a corresponding significant decline in consumption. It is this last bit of evidence that I want to discuss in more detail.
If you notice that consumption declines significantly after a person becomes unemployed or after they retire, this would seem to provide evidence against the permanent income hypothesis. The unemployed worker should be spending out of his savings or borrowing until he finds a new job. The retired person should have planned better for the future.
The problem with this assessment is that whether or not the permanent income hypothesis holds depends on consumption behavior. In reality, most of our data on consumption is typically consumption expenditures. It is important to understand the difference.
To understand why it is important to distinguish between consumption and expenditures, consider the following example. Suppose that I am interested in food consumption. How should I measure food consumption? I could measure food consumption by how much I spend on food. I could also measure food consumption by the number of calories I eat. This might not seem like an important difference, but it can be quite important.
Imagine that I can eat the same exact meal at home as I can at a restaurant. If I eat it at the restaurant, then my expenditures are equal to the market price of the meal. If I eat it at home, my expenditures are the cost of the ingredients. The latter should be less than the former. In addition, the degree to which the latter are less than the former will depend on how much time I spend shopping for the lowest prices of those ingredients. Nonetheless, despite the difference in expenditures, my food consumption is the same (by definition, it’s the same meal).
So why is this distinction important?
Think about people who become unemployed or people who retire. What they have in common is that they have more time than they had when they were working. The opportunity cost of their time has fallen. As a result, those who are unemployed and those who are retired are likely to spend more of their time cooking than they would if they were working. They are also likely to spend more time searching for better prices on the ingredients to make their meal than they would if they were working. The result is that the individual will spend more time on what economists call “home production” (and therefore home consumption) while reducing market expenditures.
This is important for the following reason. There is a difference between expenditures and consumption. Expenditures are simply the subset of consumption that occurs in a market setting.
So how significant is this distinction?
It turns out that this distinction is quite important. Mark Aguiar and Erik Hurst have a paper in the Journal of Political Economy that uses a cool data set that consists of food diaries of U.S. households. What the paper shows is that neither the quality nor quantity of food intake by retired households decline after retirement. In addition, they find that the food intake of unemployed workers does decline, but only as much as one would predict from the decline in permanent income typical of being displaced for some period of time from one’s job. In other words, if one considers the role of home production, then the evidence of a significant decline in expenditures following retirement or job displacement should not be interpreted as evidence against the permanent income hypothesis. Relying on expenditure data to measure consumption might cause one to incorrectly reject the permanent income hypothesis.
What does this mean for economists and their models?
First, as more and more micro-level data becomes available, it is important to consider whether one has the correct measure of the variable of interest before embarking on hypothesis testing. Second, this result seems to imply that if you are going to take a model to the data and you use standard measures of consumption expenditures, the model should include home production in the household decision. Otherwise, what the researcher is calling consumption and how consumption is calculated are not consistent.
The Everyday Economist 