Friday, September 26, 2014

Macrocomplaining vs. macrosplaining



I don't blog much about macroeconomics methodology debates that much anymore, but every once in a while it's still fun to wade back in.

Mark Thoma wrote a column in the Fiscal Times in which he explained where he thought macroeconomics went wrong before the 2008 crisis. Some key excerpts:
There are good reasons to be critical of the rational expectations, dynamically optimizing, representative agent approach that underlies modern macroeconomic models. For example, the representative agent approach makes it difficult to study financial markets. At least two agents with different views about the future price of a financial asset are needed before we can even begin to model markets for financial assets, financial intermediation, and other key elements of the financial sector... 
[M]acroeconomists, for the most part, did not think questions about financial meltdowns were worth asking, so why bother with those theoretical complications? The financial collapse problem had been overcome, or so some macroeconomists thought. 
Nobel prize winning economist Robert Lucas, for example, in his 2003 presidential address to the American Economic Association famously claimed that the “central problem of depression-prevention has been solved.”
Josh Hendrickson, whose blog is called "The Everyday Economist" but whose Twitter handle is @RebelEconProf, decided not to be a rebel every day, and came to the defense of pre-crisis macro, Lucas, etc. Josh makes some good points and some unconvincing points.

Here is a really good point:
Suppose there is some exogenous shock to the economy. There are two possible scenarios. In Scenario 1, macroeconomists have models that describe how the shock will affect the economy and the proper policy response. In Scenario 2, macroeconomists have no such models. 
In Scenario 1, we avoid a severe recession. In Scenario 2, we could possibly have a severe recession...Given [macro critics'] logic, the only time that we would have a severe recession is when macroeconomists are ill-prepared to explain what is likely to happen and to provide a policy response.
In other words, our perceptions of the failures of macroeconomics are hugely influenced by selection bias. True! How many more crises would we have had without the models we have developed? Maybe none, or maybe some. How useless macro has or hasn't been depends on the crises we avoided, not just the ones we failed to avoid.

Here is a point that is somewhat less convincing:
Thoma argues that the reason that we lacked a proper policy response to severe recessions was because people like Robert Lucas thought we didn’t need to study such things. However, this is a very uncharitable view of what Lucas stated in his lecture (read it here)...when Lucas says that the depression-preventing policy problem has been solved, he actually provides examples of what he means by depression-prevention policies. According to Lucas preventing severe recessions occurs when policymakers stabilize the monetary aggregates and nominal spending. This is essentially the same depression-prevention policies advocated by Friedman and Schwartz. Given that view, he doesn’t think that trying to mitigate cyclical fluctuations will have as large of an effect on welfare as supply-side policies.
Lucas' speech, in other words, is a sort of old-monetarist variant of the Great Moderation hypothesis, which was very common among macroeconomists at the time. But the Great Moderation turned out to be illusionary, and that's Thoma's whole point. It may be unfair to single out Lucas for an idea that most macroeconomists shared at the time, but he is a very famous and influential guy, after all.

Josh then makes the odd point that since we didn't actually adopt the policy that Lucas claims we did adopt (stabilization of monetary aggregates), Lucas wasn't wrong. That's just too weird of a point, so I'll skip it.

Josh then makes another good point:
Thoma argues...that economists spent far too little time trying to explain the Great Moderation. This simply isn’t true. John Taylor, Richard Clarida, Mark Gertler, Jordi Gali, Ben Bernanke, and myself all argued that it was a change in monetary policy that caused the Great Moderation.
This is true. I think Mark probably misspoke; what he probably meant was not that economists didn't try to explain the Great Moderation, but that they didn't question the Moderation's stability as suspiciously as they might have.

Josh then makes another unconvincing point:
Similarly, his criticism that economists simply didn’t care enough about financial markets is unfounded. Townsend’s work on costly state verification and the follow-up work by Steve Williamson, Tim Fuerst, Charles Carlstrom, Ben Bernanke, Mark Gertler, Simon Gilchrist, and others represents a long line of research on the role of financial markets. Carlstrom and Fuerst and well as Bernanke, Gertler, and Gilchrist found that financial markets can serve as a propagation mechanism for other exogenous shocks. These frameworks were so important in the profession that if you pick up Carl Walsh’s textbook on monetary economics there is an entire chapter dedicated to this sort of thing. It is therefore hard to argue the profession didn’t take financial markets seriously.
This is the idea that "there exist papers on X" means "the profession took X seriously". But that doesn't seem right to me. After all, there is nothing limiting the topics on which macroeconomists write papers, and there is every incentive for them to write papers imagining any and every conceivable phenomenon. So there will, in general, be a macro paper on any topic. But it is impossible for the profession to simultaneously take every topic seriously, so we need a better criterion than the existence of papers.

In particular, the claim that macroeconomists thought enough about financial shocks and frictions before the crisis seems to conflict with the huge outpouring of work on financial shocks and frictions since the crisis. If macroeconomists were clued in to the dangers of financial stuff before the crisis, why all the new models?

Josh then misunderstands one of Thoma's criticisms:
The same thing can be said about representative agent models. Like Thoma, I share the opinion that progress means moving away from representative agents. However, the profession began this process long ago. While the basic real business cycle model and the New Keynesian model still have representative agents, there has been considerable attention paid to heterogeneous agent models.
But Thoma was talking about heterogeneous beliefs. The models Josh is talking about incorporate heterogeneous wealth, productivity, etc., but not heterogeneous beliefs. There are lots of heterogeneous-belief models in the finance literature (see here for some of them). But macro models mostly continue to adhere to the rational-expectations framework advocated by Lucas, or occasionally a representative-agent version of a learning model, neither of which incorporates heterogeneous beliefs. I say "mostly" because I haven't seen any recent macro models that include heterogeneous beliefs, but Rule 1 of macro is "There is a macro paper on any topic."

Anyway, Josh makes some good points, but he also goes pretty soft on the profession and its leading thinkers regarding the pre-crisis consensus, which really did downplay the importance of finance, and really did avoid thinking about heterogeneous beliefs.

26 comments:

  1. Noah,

    A couple points re a couple of your points. Not everyone who's studied the Great Moderation concludes it was monetary policy wot done it. See (shameless self-promotion) http://kansascityfed.org/publications/research/er/er05q3.cfm, or Stock & Watson.

    Also, I'm a macro guy so hadn't heard of Xiong's paper (but will read it), this seems like the sort of thing that Roger Farmer's been talking about with his "belief function," (maybe?). And while I read a lot of the hetero agent stuff several years ago, and there's lot of it, it definitely seemed to fall in your category of "there exist papers on X..."
    FWIW,

    PS

    ReplyDelete
    Replies
    1. Thanks for the link! I actually haven't read Stock & Watson (I know, I know).

      Farmer's thing is a bit different. The Xiong paper is a survey paper, so it's pretty easy to skim.

      Delete
  2. Mike Toreno9:54 PM

    No, the reason macroeconomics didn't predict the crisis is because macroeconomics is all lies. It was easy to predict that the crisis would happen, and it was predicted. All you have to do is give up some of the observation-free orthodoxy on which macroeconomics is built. Look at this: The easiest way to get rich is to borrow money and not pay it back. People are motivated to take actions that benefit themselves. Now, you just take these two observations and apply them to banks. What were people who ran banks motivated to do? Benefit themselves. What was the easiest way to do that? To cause the banks to borrow money and transfer a large portion of it to the people who ran the banks. That, and no other reason, was why banks entered into economically useless transactions that had a volume greater than that of the combined world gdp. That is the only reason banks entered into credit default swaps. To inflate the volume of transactions so the portion available to be skimmed off would be greater.

    You don't need macroeconomic models. You need honesty. This lack of honesty is why those macroeconomists who failed to predict the crisis, failed to predict it.

    ReplyDelete
  3. Anonymous6:57 AM

    central problem of depression-prevention has been solved and is known to economics. yes some time ago. But not known by Politicians.

    ReplyDelete
  4. Noah, have any macro models been set up that have multiple agents? I seem to recall that the reason economists use single agent representative agent models is because multiple agent models are intractable and have multiple equilibria. The latter doesn't sound like a problem to me. It sounds like the real world. After all, my constraints (and everyone else's) are determined by aggregate choices. We can't ex ante know what the constraints are, but everyone makes choices believing that they know what the constraints are. If the macro economy is in a steady state, no problem. If not, people revise beliefs and choices - and the constraints change again. Why wouldn't there be multiple equilibria?

    ReplyDelete
    Replies
    1. This gets into Noah's Rule 1 territory, but there are some, and they do indeed frequently involve multiple equilibria. A few people have advanced multiple equilibria as an explanation of recessions (recessions are a bad equilibrium, and government action can push us into a good equilibrium).

      The intractability comes less from multiple agents and more from what's known as "incomplete markets". If there are no missing financial markets, then the models can be phrased as optimization problems, which can generally be solved. If there are missing financial markets, then you can't do this anymore, and no one is really sure whether these models can be solved or not.

      Delete
    2. If the model can be expressed in mathematical terms then it can be solved numerically in the sense of moving from an initial condition to a future state. If you try to find the optimum of some quantity then at the least you should be able to find a local optimum, global optimums may be impossible to find.

      Delete
    3. I'm talking about multiple equilibria in the sense that an infinite number of solutions may exist. The whole problem is poorly defined because my spending becomes your income and vice versa. Thus my budget constraint is determined after all other actors commit to a course of action and my actions affect the budget constraints of other people. The whole structure could crash if we all develop a lack of confidence in the system.

      Delete
  5. Anonymous10:58 AM

    I like your last point.

    Thank you for cluing us non-economists in on the state of representative agent models. I didn't know that existing, heterogeneous representative agent models didn't account for heterogeneous beliefs. Clearly that is where the weakness lies.

    It now seems abundantly clear why attempts at models using heterogeneous agents have failed. The most prominent model that I've seen, which attempted to group the world into three groups, grouped the people wrong because it grouped them by their function (banker, investor, consumer) without differentiating by beliefs.

    Beliefs are the most important aspect of financial markets, and they are what causes failure. The proper macro model incorporating finance should divide people into groups based upon information set and likely beliefs. You just can't make the model too fine, it has to be course and macro-based. This is why I think the existing line of agent-based models will fail to account for macro effects. We need a macro model using heterogeneous, representative agents based upon both belief and function.

    It isn't hard to do this. It is actually easy, and that is perhaps what causes economists to reject this idea. The fact is, in my belief, that most people who have experienced life could create a pretty good grouping of the people based in common sense. The analysis takes some heavy lifting in mathematics, but the basic framework is not hard to create.

    ReplyDelete
    Replies
    1. You probably need at least a hundred different agents with different skills, endowments, knowledge and beliefs (and each representing thousands or millions of individuals) to make a model useful in representing the dynamics of a market economy.

      Delete
    2. Anonymous6:17 PM

      I disagree. I have a model in mind that only contains around 5 behavioral agent groups. If you get the groups correct, you don't need that many of them.

      I believe that there could be more agents in a pool that are predefined, and a computer could crunch different models and groupings to look for the predominant beliefs and behaviors that fit certain trends. For instance, In a depression/recession there are different beliefs/behaviors that are important vs. in a healthy economy. Leading up to a financial crisis, there are a very small number of behavioral agents that provide the most important insights.

      Delete
  6. Anonymous2:55 PM

    Isnt Angeletos at MIT doing macro models with heterogeneous beliefs ?

    ReplyDelete
    Replies
    1. I know about his paper with dispersed information, but I don't know about one with financial markets. Got a link?

      Delete
    2. Anonymous3:08 PM

      no, its business cycle model. As he says "whether our approach can be fruitfully applied to macro models with finance is an open question"

      Delete
  7. It is a cop-out to attribute the housing bubble and financial crisis as being the result of exogenous shocks to the economy. Both the bubble and the crisis were the result of intrinsic instabilities. Economists should be researching the mechanisms that gave rise to those instabilities and what institutional changes could reduce or eliminate those instabilities (I personally favor executing bankers who publish misleading financial statements as a useful first institutional change but I suspect I am for the time being in the minority on that).

    ReplyDelete
  8. Hendrickson just channeled Ken Boulding:

    Attention macro-mockers:

    Faced with an exogenous shock, there are clearly only two possible scenarios:

    Scenario 1: Assume economists have a can opener
    Scenario 2: Assume economists do not have a can opener

    In Scenario 1, the can can be opened. In Scenario 2, the can can or cannot be opened. Therefore, using the macro-mocker's very own logic, the only time we need to be concerned with recessions is when economists do not have a can opener. AM I RITE?

    ReplyDelete
  9. You are (as often) too kind. In 2008-9 the Fed did stabilize monetary aggregates. Thus they tested Friedman Schwartz, Lucas (and Bernanke)'s claim that this would have prevented the great Depression. We now know that Friedman was wrong as were Schwartz, Lucas and Bernanke. The phrase "stabilize the monetary aggregates and nominal spending. " suggests that if monetary aggreates are stabilized then so is nominal spending. This is exactly what Friedman et al asserted. Asserting it now is like asserting that the world is flat.

    It is unreasonable to be too hard on Lucas. As I mention above, Bernanke claimed that Friedman had explained how to prevent depressions (it was at a birthday party for Friedman but I'm sure it wasn't insincere flattery).

    Hendrickson assumes that nominal demand can be stabilized using monetary policy alone (following Lucas, Friedman etc). This was the general view of macroeconomists (including new Keynesians). The evidence from 2008 proves that they were wrong (it might be that some monetary policy would have worked but they one they advocated and Bernanke implemented didn't).

    This should all be well known to the general public, but in case proof is needed, I show growth rates over the preceding year of of M1, M2, nominal GDP and GDP/M1 .

    https://research.stlouisfed.org/fred2/graph/?graph_id=197117

    This time the Fed made sure M1 went up. GDP fell sharply. Asserting in this day and age that macroeconomists and in particular Friedman had discovered that the way to prevent depressions is to stabilize monetary aggregates is just like asserting in this day and age that physicians know the way to cure disease and it is applying leeches and bleeding. It was attempted (by the book the book being "A Monetary History of the United States" and it didn't work.

    Again everyone should know this. It is shocking that an economist dares assert that the question is still open.

    ReplyDelete
    Replies
    1. Anonymous5:02 PM

      Well since there was no depression, I guess Friedman based on that isolated case, was right.

      Delete
  10. To put it briefly. Monetary aggregates were stabilized in 2007-8. In fact M1 increased sharply. Friedman's hypothesis (which was accepted by Bernanke as well as by Lucas) has been tested. It is known to be false.

    I don't see how anyone who asserts that the Friedman etc proposal wasn't followed in 2008 can dare to assert that he is an informed citizen. That someone who wrote that perceives himself to be and is accepted as a macroeconomist in good standing is shocking.

    This is a debate about what happened to the very most closely watched macroeconomic time series. There is no room in any science or most pseudosciences for such a debate.

    I expect higher scientific standards from astrologers.

    ReplyDelete
    Replies
    1. Anonymous3:15 PM

      Are drugs legal in Italy?

      Delete
  11. "True! How many more crises would we have had without the models we have developed?"

    This is an excellent point, Noah! Every morning when I get up, I follow my rational model of elephant-prevention and scratch my buttocks 12 times (no less, and no more). For the last 40 years, I have not had to face an enraged elephant in my bathroom first thing in the morning. How many would I have had to face without the rational model I had developed? We may never know.

    ReplyDelete
  12. Anonymous6:16 AM

    Lack of inflation pressure in the US since 1980 has been a result of weakness in the labor market: No wage inflation, no wage-price spiral, no inflation pressure other than commodity inflation. Monetary policy since 1980 has acted to suppress wages and produced decades of stagnation that finally led to insufficient demand.

    Maybe the Great Moderation was a bad sign that over concern about inflation was hurting labor and wages which led to over leverage and collapse? Perhaps, rather than a victory over inflation, the Great Moderation was a pathology?
    jonny bakho

    ReplyDelete
    Replies
    1. Anonymous4:52 PM

      You mean lack of inflation since 1997.

      There was indeed inflation since 1980.

      Delete
  13. Anonymous4:55 PM

    (Whispering). I am going to tell you a hint, the Great Moderation never ended. It is still going on. The fact that you don't see it is quite telling how lost the econ-blogosphere has become.

    ReplyDelete
  14. The central academic macro theory failure was not representative agent models but ignorance of the credit cycle. The main mollifying belief of the great moderationists was that moderate consumer inflation implied moderate credit expansion. There is still widespread confusion and/or ignorance of how credit expansion relates to GDP. Still very few phds know that the net credit expansion rate is practically a component of final demand; many assume it's related to the growth rate of final demand or just don't know or don't think about it.

    ReplyDelete
  15. Thomas Palley has written a nice paper as a kind of critique of the mainstream macro view as applied to the financial crisis. The interesting thing is that he certainly is a macro economist (He's a traditional Keynesian), and believes in the power of macro, but is not happy about the restricted views that are currently accepted by the mainstream macro economists. The problem as he sees it is not with the field of macro, but with the preconceived notions that have become popular in the mainstream.

    See "The theory of global imbalances: mainstream economics vs. structural Keynesianism", August 7th, 2014 at:
    http://www.thomaspalley.com/

    ReplyDelete