Monday, May 17, 2010

Searching for Truth in Economics

Do you ever sense God has a giant foam arrow pointing you in a direction of discovery? My paths on blogs and in my personal endeavors have led me to ask some tough questions to my beloved field of economics.

1. When economists say they "know" something, what do they mean?

2. How do we know it's true?

These are two questions that I've been exploring more deeply by looking into the Philosophy of Science. People in this camp ask really interesting questions about what qualifies as "good science".  My opinion since thumbing through these thoughts? In order for good science to happen you need to have a testable claim (a theory AND you might even have multiple theories). Then, you carefully manipulate variables relevant to the testable to claims to see if they are influencing the outcome you try to test. When I tell my students that economists know "stuff" through experimental tests and econometric analysis I feel quite justified in saying, "Economists know X or Y" and I feel very scientific because I feel like we take seriously this kind of scientific method. But, I've noticed something for sometime . . . . and a confluence of events as caused me to confront this head on: econometric analysis can be misleading.

By the way, for those not familiar with econometrics . . . The central idea of econometrics is to hold some variables constant while seeing what the impact is of another variable. For example, we know there is a wage differential between women and men. But, what we really want to know is whether holding all the other relevant variables (such as experience, educational certification, marital status, race, etc.) constant whether being a man or woman has a significant impact on wage. In general this seems like a great way to analyze data, but, there are problems. Here is where my confessions of confusion and skepticism need to be brought to the light so you can respond.  

Last night on the Greyhound Bus I listened to Russ Roberts talk with Robin Hanson about Truth in Economics where the really interesting question was asked,

"When is the last time an empirical study ever changed your mind about a topic?" 

Can you think of a study you read and then said, "Wow! This just blows my initial thoughts on Policy X out of the water!" There are competing empirical studies for just about everything in economics and while consensus is large on quite a few issues it is quite slim on a number of other issues. This is captured by varied opinions on the impacts/magnitudes of impacts that minimum wage, universal health care, immigration, foreign aid, etc. have on the goals of employment, quality of care, low skilled labor market, and growth respectively. For more information about how economists have different views check out Dan Klein's work at Econ Journal Watch.

What are the basis for these different outcomes in the studies? Different data? Different econometric techniques? Biased political and academic processes that gives incentives to find "the right answer"? (For example, if you stake out a desirable position about a major policy question you can open up future possibilities for funding and careers in policy think tanks) I hope the latter is not true but suspect that it might be. Economists follow incentives like everyone else and if these incentives are driving the statistical outcomes is the integrity of the profession is seriously in question? Others have noticed this problem as well, wondering why pundits of these various policies overlook facts in favor of entertaining lies.

This brings me to a final point. The new Journal of Economic Perspectives has a symposium in their new issue to discuss some of these points but is more cheerful in tone heralding the the brand new "credibility revolution in econometrics". Perhaps this should make me feel at ease about the current status of econometrics, BUT,  while minor improvements can be made to the econometric techniques around the edges how much of econometric technique has the flavor of being a fad? At the heart of the debate is an oldie but a goodie, "Taking the Con out of Econometrics" by Ed Leamer. My guess is that this symposium will cause loads chatter throughout the economics profession about what is true, but I could be overly optimistic. The folks at Permutations blog write provide some nice links to some good summaries here and Russ Roberts interviewed Leamer on EconTalk here to talk about the state of econometrics.

All is not lost in my couch confession! Healthy doses of skepticism are probably a good thing. Admittedly I am a novice at these kinds of advanced econometric techniques and look forward to getting up close and personal in the future so I have a better idea about the validity of different tests. But, here are some items in the midst of this crisis toned post that I feel I can say with some confidence:

1. People operate with limited resources (scarcity) and they make tradeoffs with those limited resources.

2. Human beings make tradeoffs many times based on a cost-benefit calculation (however quick it might be)

3. The costs and benefits of an action are often determined by incentives via social and legal rules

4. Humans have limited knowledge and cannot know everything which is why there are so often unintended consequences to policies

No comments: