Subjective outcomes in economics.

• Author: Hamermesh, Daniel S.
• Position: Association Lecture

1. Introduction

   There has been an upsurge in the use and analysis in economics of subjective outcomes. A long survey article (Frey and Stutzer 2002) discussed some of the literature, but even that was restricted to only one part of the interest in this type of outcome. A relevant question is why economists are looking at these expressions of behavior now. I think it is partly due to the fact that we are seeking things to do with the large amounts of data that we have suddenly discovered--the Mt. Everest phenomenon: If it's there, we must climb it. Partly too, it may be that we believe that our skills enable us to do better jobs of analyzing outcomes that have previously been the sole purview of other social scientists.

   In the mid-1970s, I did some research on a subjective outcome--job satisfaction. The comment from the editor of a journal to which I submitted the article was, "This is not economics." Was the editor correct, and would he still be correct today? There is a lot of work being done on these topics, but what kind of work should we be doing? Here I go through a variety of examples of my own and others' research in trying to distinguish between the kinds of topics where we have something to offer and others where we do not. I neglect many examples with which I am not familiar that may be at least as good as those that I present. This brief discussion indicates that there is a disjuncture between what we economists are doing in these areas and what we should be doing.

   This discussion raises a larger question, which I am not going to answer: "Is there anything beyond the scope of economics?" I have some bona fides in asking this question, having written articles on such bizarre topics as suicide, beauty, and sleep that some people argue are not subjects that we should be working on. If those topics are off limits for us, discussing the economics of subjective outcomes is even further away from what we should be considering. Some of it is, but I will try to demonstrate that some of these things are very much within our purview, and that we should be thinking about them. While I cover a fair number of topics, there are other subjective measures (e.g., preferences toward risk, rates of time preference) that I do not discuss but that have been studied by economists and used in empirical analyses of economic outcomes.

2. Life Satisfaction

   The first research area that I deal with is life satisfaction: How happy one is with one's life. Economists' work on this topic comprises, if not the majority of our research on subjective outcomes, at least surely a plurality. The general idea here is to obtain data on how satisfied the subjects are, based on their subjective responses, and to look at the effects of economic variables on their satisfaction. The typical economic variable is income. The general research question is, "How happy does your income make you?" Many of the people who work in this area try to equate this happiness measure with utility (with Levy-Garboua and Montmarquette [1997] being the best example so far). On the surface, this is very appealing. Utility is one of the two foundation ideas that we deal with in our intro micro classes, and it is a topic we should be able to say something about. How happy are people over time within a country? How happy are people across countries? What are the effects of various demographic correlates on differences in happiness?

   I find most of the analysis by economists on this issue to be rather silly. The first difficult issue is the scaling problem--the problem of converting what must be some continuous ordinal measure into a measure described by a few categories. How people scale it in their own minds is a question that we economists have barely begun to think about, although cognitive psychologists have paid this issue substantial attention.

   The larger problem, however, is not the scaling problem. I am willing to admit that we might not be able to solve the scaling problem, but at least we can wave our hands and say that somehow there is an underlying ordinal index of people's behavior that could be captured if we were able to spy on them sufficiently carefully. But there is another problem here: There is no apparent link of these subjective responses to any underlying measure of utility, at least measures that maintain the properties that we desire for utility functions. On a theoretical basis, Homans (1961) pointed out that there is no reason to believe that long-run improvements in objective circumstances will generate long-run increases in subjective responses. While we generally believe that utility rises with additional inputs of goods and the time to consume them, many psychologists believe that satisfaction adjusts to expectations about living standards. If we believe that utility equals expressed satisfaction, there is little justification for any policy designed to raise living standards.

   On the empirical side, there are books and articles surveying hundreds of empirical studies of life satisfaction by psychologists (e.g., the survey by Diener and Biswas-Diener [2002]). If you examine this literature, you find that every specification that economists have used has also been included in at least several psychological studies. For example, I discussed an article at the 2002 American Economic Association meetings (Gardner and Oswald 2001), which received immense amounts of press, in which the focus was on lottery winners. If I won the Texas Lotto, I would be very happy for a while. But after I was used to the lottery win, I would think, "The lottery's my due, I deserve to have won it." My satisfaction would revert to its mean. This is exactly what was observed in this study. It is also exactly what was shown in two very similar studies of lottery winners done by psychologists in the 1970s.

   Our ability to push buttons in STATA, SAS, TSP, or whatever, is not unique: Psychologists and sociologists are perfectly capable of doing that. Our strength, I would argue, and the thing that should underlie whether we do anything in this area, is the extent to which we can bring economic theory to bear on the specifications that we try to formulate. On this topic in particular, I have seen nothing where any economic theory has been brought to bear. Thus, in terms of a hierarchy that I will summarize at the end of this article, it is clear that economists' work on this topic has not progressed very far. (1) Now perhaps somebody will generate more ideas that use economics here, but simply running fancy regressions on this subject is not new (except to economists). Although I would like to see more people working in this area, thus far we have not done much to justify our incursions.

3. Job Satisfaction

   I have published two articles on job satisfaction, one each quarter century (Hamermesh 1977, 2001). To show that I am not completely self-congratulatory, I rather doubt that we have offered much in this area either. We have only one notion here that other fields do not have, and that is the idea of equilibrium. If people choose an occupation based on the returns to working in it, then a permanent shock to those returns will in the long run have no...