Monday, May 12, 2014

Public Service Announcement: There is an inequality literature in macro

Here is a claim by Atif Mian and Amir Sufi:
Macroeconomists have traditionally ignored distributional issues, grouping all households into a a single “representative agent.”
This is misleading (unless Mian and Sufi define "traditional" as "representative agent," in which case it's a tautology). Here's an excerpt from Big Ideas in Macroeconomics (review here), which is already proving useful as a reference:
Starting in the late 1980s, a clutch of papers arrived that would open the floodgates for macroeconomic research into inequality. . . . A few of these papers deserve special mention. These are Ayse Imrohoroglu's 1989 paper on the pain inflicted by business cycles on households, John Laitner's 1992 paper on how luck in earnings and inequality are related in general equilibrium, Mark Huggett's 1993 paper on the "Risk-Free Rate in Heterogenous-Agent Incomplete-Market Economies," and the late Rao Aiyagari's seminal paper on "Uninsured Idiosyncratic Risk and Aggregate Saving" (1994). These papers taught an entire generation of macroeconomists like me how to frame questions in which uninsurable risks were likely to be important for decision making and for the implications of policy. . . . Twenty years on, we are still learning from this model. For instance, Guerrieri and Lorenzoni (2011), a quite standard [incomplete markets] setting, is at present the leading model helping macroeconomists understand the implications of credit crunches for real interest rates and aggregate consumption.
Of all the advances made by macroeconomists over the past two decades, these models have been at the top. . . . For thorough reviews of the models I refer readers to Heathcote, Storesletten, and Violante (2009) and Guvenen (2012). (291)
There is much more in following pages. I was taught these models in multiple classes in the macro sequence of my PhD program. If I had to define "traditional" macroeconomics, I would probably define it as the stuff they teach in grad school. Here is another compilation of relevant literature, yet some people seem to be unaware of it or refuse to acknowledge its existence. Inequality is an old topic in macroeconomics. The literature has considered policy, and it has shown that a variety of causal forces can produce observationally equivalent patterns in wealth and income distributions--a result that should inform the policy discussion! [Update: Note that I'm not claiming this literature is totally complete or conclusive; just that it exists and is active.]

Suppose, instead, Profs. Mian and Sufi said:
Many macroeconomists work primarily with representative agent models that abstract from distributional issues. We believe this focus to be misguided and hope that the data we have assembled will lead to greater emphasis on heterogeneous agent models.

Broadly speaking, I would agree with that. I've complained about rep agent before (1, 23, 4), and the data I regularly discuss on this blog suggest a need to focus less on rep agent. Some macroeconomists might disagree, pointing out ways that representative agent models have provided other useful insights and noting the high costs of doing policy analysis with het agent models (which are often very computationally intensive). But the point is that asking for a change in focus at the margin is reasonable, while suggesting that macroeconomists have been ignoring distributions this whole time is not.

Economists do their readers a disservice by saying that the field has until now totally ignored hugely important economic concepts (Piketty tried this too). To insiders, it may seem like the usual attempt to distinguish new research from the literature, but outsiders may take it seriously and be misled. Mian and Sufi have a new book; I hope it does not ignore the work others have done.

Update, 5/13: Kartik Athreya emails me the following helpful addendum:

Thanks for drawing attention to this huge branch of macro.  One point I'd stress to outsiders to the literature is that the models you note can be used to assess the distributional effects of policy when wealth inequality arises specifically from uninsurable fortune and misfortune to labor earnings. Discussions often had in the public sphere reveal that inequality is bothersome to many, and in these models, inequality is primarily reflective of insurance and/or labor market dysfunction, rather than being due to character "defects" or differences in future-orientedness or preferences for leisure. It is not the only way to generate inequality, but it seems a good starting place.


  1. It's quite possible to define 'traditional' in a non-tautological way. A good textbook should include references to traditional material. Benassy's Macroeconomic Theory (2011) is a good textbook. As the glowing blurb by Bob Lucas says, it reviews every important development in macroeconomic theory since the 1950s. "There are 590 references, from Abel to Zabel."

    Nowhere in the 590 references do I find mention of Laitner, Huggett, Aiyagari, Guerrieri or Heathcote. Guvenen gets two mentions but not for this work. There is no entry for 'inequality' or 'distribution' in the index.

    Economists do their readers a disservice by saying that the profession is paying serious attention to an issue when what they mean is that there are papers dealing with it if you search for them.

    1. Check Ljungqvist & Sargent, which is pretty standard in grad programs I think. Chapter 17.

    2. ...though I agree that you can define "traditional" in non-tautological ways. I provide one definition above--stuff they teach in grad school. I encountered het agent models in multiple classes in grad school.