Thursday, January 16, 2014

Pissing Contests

Take a look at the cartoon to your right. It's funny, no? But humor aside, the cartoon reflects a serious idea. It's the idea that wealth transfers from the rich to poor can create wealth--at least, in depressed economic conditions. This basic idea is a staple of undergraduate level "Keynesian" economics -- the so-called "spending multiplier."

In terms of the cartoon, this is how the spending multiplier works. Imagine that the rich dude drops a dollar in the poor guy's tin cup. Or, equivalently, imagine that the state takes the rich dude's shoes and gives them to the poor guy. (The multiplier story does not depend on whether transfers are voluntarily or not, or whether they are made with cash or in kind.) What happens next?

Well, let's imagine that there's an unemployed cobbler in the neighborhood and that the poor guy really wants shoes. Then the poor guy takes his dollar and uses it to order shoes from the (previously) unemployed cobbler. Shoe production (GDP) goes up. The same thing happens if the rich dude wants to replace his stolen shoes. The spending multiplier at this stage is one: a one unit transfer of shoes results in a one unit increase in the production of shoes.

But a multiplier greater than one is possible if there are also unemployed butchers, seamstresses, etc. That is, after receiving his dollar, the cobbler himself goes on a shopping spree, the effect of which is to increase employment in other depressed sectors.

What is the empirical support for this seductive idea? In a recent WSJ piece, Robert Barro argues that there is no "meaningful" support for the idea (either theoretical or empirical). Barro's dismissal of the "wealth transfers create wealth" idea was quickly mocked by Paul Krugman, who called it "anti-scientific." Antonio Fatas also weighs in here with his own characterization of what he calls the "anti-demand coalition."

Goodness, where to begin? Well, first off, I think that Barro goes too far in suggesting that there is no "meaningful" theoretical support for the idea (it depends on what one considers "meaningful," I suppose). We can certainly write down coherent "Keynesian" models where the result--or something like it--holds. The result can also hold in models with perfectly flexible prices, like the OLG model, or neoclassical models with credit constraints (see also the work of Roger Farmer). Of course, whether one buys into all the assumptions that drive these results is another matter. But this is where the need for empirical support comes in. And of course, that's one of the most problematic parts of our discipline.

I'm not sure what the evidence says about the wealth-creating propensity of wealth transfers in depressed economies. Let's just say that I, like many others, are skeptical that the mechanism is quantitatively important. As Barro says, "I am awaiting more empirical evidence." Note that Krugman offers no direct empirical support for the proposition in question. For that matter, nor does Barro provide any for his favored hypothesis. So far, all we have is a pissing contest.

I've already explained why I don't like the way Barro pissed on "Keynesian" theory. Now let me explain why I don't like the way Krugman pissed on "regular" economics. Here is Krugman:
If you read Barro’s piece, what you see is a blithe dismissal of the whole notion that economies can ever suffer from am inadequate level of “aggregate demand” — the scare quotes are his, not mine, meant to suggest that this is a silly, bizarre notion, in conflict with “regular economics.”
While macroeconomists frequently use terms like aggregate supply and demand, one should keep in mind that these objects are not so straightforward to identify in general equilibrium theory (let alone in reality) where everything seems to depend on everything else simultaneously. In any case, "regular economics" offers plenty of examples where equilibrium allocations are socially inefficient. While a well-designed policy intervention may be desirable in such circumstances, it does not necessarily imply that wealth is created through wealth transfers, although this may certainly be the case. So we don't need theories of "deficient demand" to motivate policy interventions.
You’d never know, either from the WSJ or from people like Barro, why anyone ever felt that regular economics — the economics of supply and demand and all that — was inadequate. But you see, there are these things we call recessions. And if you believe regular economics is all there is, you should find them very upsetting.
The suggestion that Barro -- or anyone who does not subscribe to the "deficient demand hypothesis"--somehow missed the recession is quite funny. I know this is too much to bear for the true believer, but alternative interpretations are possible.
Think, for example, about the Great Recession and its aftermath. Regular economics says that economies should normally get richer each year, as their work force and capital stock grow, and technology advances. But after 2007 the United States and other advanced countries suddenly went into reverse, becoming poorer instead of richer, and for an extended period too.
Regular economics says no such thing of course. Whether an economy grows or not depends on a host of factors, including the parameters of the institutional environment (property rights, tax regimes, regulatory regime, etc.). Here is Antonio Fatas using the same rhetorical device to make the other side seem silly and anti-scientific:
Their models are only driven by changes in the productive capacity of an economy which means that the Great Recession (or the Great Depression) must have been the result of some destruction in our capital stock or our inability to remember how to work or produce or somehow our technology got worst than in previous years.
First of all, technology does not not have to "get worse" to generate a temporary recession. It is easy to build equilibrium models with Schumpeterian "gales of creative destruction" that generate recessions. Although I have never seen it, one could conceivably combine this impulse mechanism with a Fisherian "debt-deflation" dynamic to produce prolonged economic slumps from a positive technology shock.

Second, it is easy to generate what look like "aggregate demand events" in even RBC models. One way to do this is to hypothesize the existence of a "news shock;" see, for example, here. Most dynamic models have something like this equation living inside them:

Investment = Increasing function of ( Expected After-tax Return to Investment )

So investment demand (the most volatile component of GDP) depends on "news" (information) concerning the likely future return to investment. Note that investment could be interpreted broadly here to include human capital investment and (in labor market search theories) investment in recruiting activities.

It is not implausible to imagine that expectations concerning future returns might fluctuate a lot, or remain depressed for long periods of time. (This was certainly an important theme in business cycle theory well before Keynes ever coined the term "animal spirits.") Changes in these expectations are likely to be interpreted by econometricians as "aggregate demand shocks" because they induce movements in output and employment when contemporaneous measures of "supply" (technology and capital) remain fixed.

An important question here, I think, concerns the source of these expectation shocks. Do expectations move around passively in accordance with movements in the perceived reality? In other words, do people become rationally optimistic/pessimistic ex ante (even if they may be wrong, ex post)? Or do expectations move about in ways that are inconsistent with the surrounding reality--irrational mood swings? Or can expectations themselves move about for no good reason, shaping reality in a manner that results in a self-fulfilling prophecy? (Here is a paper on the question.)

But surely, one might say, is it not obvious that the current problem is insufficient demand? For example, many firms cite as their main problem a "lack of sales." In the same vein, Fatas reports:
In fact, when most people are asked to describe the dynamics of economic crisis, they immediately refer to some notion that shortages of demand cause recessions.
But as I have discussed before (here), we have to be careful how we map these individual (micro) impressions to what is actually happening at the macro level. Consider, for example, a market-clearing model with intersectoral linkages (like the original RBC model). A real shock in one sector is going to affect the derived demand for a product in another sector. An individual supplier in this model may very well report a "lack of sales" to be his main problem--but this does not necessarily have anything to do with the deficient demand hypothesis. Similarly, we can't say something like "Hey, if the problem really is deficient demand, then wouldn't we expect all beggars to have signs like the guy in the cartoon above?"  No, I'm afraid not.

I am not even sure I agree with Fatas on the point of why so many people appear to resist the idea of countercyclical policy. A majority of people likely do buy into countercyclical policy--there certainly seem to be a lot of "automatic stabilizers" built into the U.S. economy, and both the fiscal and monetary authorities have taken considerable discretionary measures (some would have liked more, and some would have liked less, of course). But maybe Fatas is correct in saying many people just do not trust the government to do the right thing with sequestered resources in current economic circumstances and given the current political climate. Most economic models assume away inconvenient political-economy issues. Optimal interventions can be easily identified in theory, but whether the political landscape is likely to permit implementation of an optimal policy is a different question altogether.

This is already getting way to long for a blog post, but I'd like to conclude with one final thought. At the end of the day, the real issue at stake was how to rationalize an extension to UI benefits. It should be noted that Barro was not arguing against the UI extension -- indeed, his column was subtitled: Food stamps and other transfers aren't necessarily bad ideas, but there's no evidence that they spur growth.

O.K., let's suppose transfer doesn't spur growth (create wealth). There are still other ways to rationalize the extension using "regular" economics. There is a consumption-insurance aspect that might be relevant in a world of incomplete insurance markets. There is the idea of helping the unemployed finance an extended spell of job search to increase the likelihood of a good match. And so on. The same holds true for how we can rationalize public infrastructure investment based on standard cost-benefit analysis--a lot of unemployed construction workers and interest rates at very low levels. Let's stop the pissing contest and start looking for some common ground here.

24 comments:

  1. Poor people have higher marginal propensity to consume, no?

    http://www.econ2.jhu.edu/people/ccarroll/concavity.pdf

    ReplyDelete
    Replies
    1. Yes, and standard growth models tell us that, if the aggregate savings rate is lower, GDP per capita falls in the long run.

      Delete
    2. Hi John. Yes, that is almost surely true, but what is your point? I model heterogeneous agents all the time (with different propensities to consume and invest). In some models, a transfer of wealth from high MPC to low MPC is what creates a boom, if the latter are credit constrained, for example.

      Delete
  2. My macro pre-dates 1985, so I'm curious about your apparent reduction of stimulus or activist fiscal policy to 'wealth transfer.' Of course there is the Keynesian idea that differing MPCs by class make a wealth transfer effect plausible, but I always thought the dominant activist policy option was deficit-financed public spending (w/monetary accommodation). To me that is very different from a wealth transfer. Borrowing from somebody is not transferring wealth away from him. Feel free to enlighten me. I know you will.

    ReplyDelete
    Replies
    1. . . . chirp . . . chirp . . .

      Delete
    2. EconoCat, while your macro may predate 1985, I assure you that your bad manner appears very modern.

      I did not attempt any reduction of stimulus to wealth transfer. I specifically targeted this idea because it was the idea Barro was criticizing and the idea that Krugman was defending.

      Delete
    3. I know you're Canadian, but you still seem to be a little thin-skinned. The crickets alluded to your comrade S. Williamson's disinclination to respond to my question (which was pretty modest & civil, all in all, wouldn't you say?).

      Explanation for my comment: your post began in more general terms, with reference to "*the* [emphasis added] spending multiplier" and "undergraduate 'Keynesian' economics" incl spooky quotes around Keynesian. FWIW, I have always thought the idea of exploiting differing MPCs by income class was weak tea.

      But thank you for the informative clarification. Cheers.

      Delete
    4. EconoCat,

      First, your question seemed directed at me, not SW. To address the SW comment, you should place yours directly below his. No big deal.

      Second, why do you follow Krugman in interpreting " " as spooky quotes? I use them to alert the reader that the definition of the quoted term may differ across people.

      Third, yes, I suppose that exploiting heterogeneous MPCs is weak tea. I suppose one way to interpret my point is that one can do so in both "neoclassical" and "Keynesian" type models. The basic idea does not depend on any notion of "deficient demand."

      Delete
  3. Good post.

    FYI: The differing MPCs argument is now being used to argue that raising the minimum wage is likely a stimulus.

    http://www.epi.org/minimum-wage-statement/
    http://www.epi.org/publication/raising-federal-minimum-wage-to-1010/

    ReplyDelete
    Replies
    1. Thank you, James. It's a silly way to motivate a price floor, imho.

      Delete
  4. Article completely misses the other side. If the rich guy kept the money in the bank, the money would get lent out. It would eventually make its way into the hands of a business owner, who will pay an employee, and the employee might want to buy shoes. That or someone would borrow it to buy a house, where that money would have been paid to a laborer, who would buy shoes. If he invested the money, it could go into in a mutual fund, eventually make its way to a business via an IPO, which would then pay an employee, who could also buy a pair of shoes. The only real difference between the scenarios is that the money ends up in the hands of someone who worked for it vs in the hands of someone who didn't. Had that work been something like farming, you have farmed crops + shoes vs just shoes.

    Everything else, schools of economic thought, tendencies of certain characters to spend, all have no meaning in here once you consider the other side. All of that is just a biased viewpoint from a political agenda.

    ReplyDelete
    Replies
    1. You are assuming that we are in a world of full employment and not at the ZLB. How does savings become investment? The rich person saves another dollar, which puts downward pressure on interest rates, which increases investment demand and decreases the desire to save until, once again, savings equals investment. If interest rates cannot fall, then perhaps this mechanism does not work.

      Similarly, usually, the poor person buying shoes will not increase production one-for-one. If we are at the full-employment level of output, then the increase in consumption may simply lead to higher prices or the fall in saving may lead to more workers making shoes and fewer workers building capital (and so future consumption falls).

      Delete
    2. "...we are in a world of full employment and not at the ZLB."

      I think you mean, suppose there are sticky prices, for example. Then we have some distortions to worry about. Of course you can fix those by way of tax policy - you need to get the intertermporal prices right. The ZLB is a red herring. Again, there are tax policies that work to correct the distortions. Nothing to do with multipliers and MPCs.

      Delete
    3. I agree. But, I was addressing RI's argument, not discussing optimal policy. My point was simply that classic market clearing arguments do not refute Keynesian arguments that wealth transfers create wealth. The Keynesian arguments are obviously more subtle because they do not apply in all circumstances. They (usually) rest on the assumption that we are below full-employment and at the ZLB (and other factors such as sticky prices).

      Delete
  5. "An important question here, I think, concerns the source of these expectation shocks."

    One aspect of this has to do with the difference between risk and (Knightian) uncertainty. "Expectations" as economists model them are (I think) a representative agent's perception of the probability distribution of outcomes. Shocks are hard to explain because: 1) their perceptions are correct, i.e. they are rational; 2) the possible outcomes fit this distribution. This view of the information set regarding future outcomes is misguided. In reality, the economy is a complex system, which means it is non-linear, has interdependence, is path dependent, etc. So expectation "shocks" have to do with the fact that predicting the outcome of this complex dynamic is difficult because it is irreducible. We have adapted to this problem by developing heuristics that attempt to capture the essence of the complexity without understanding each part. These heuristics are open to influence, which in turn results in herding behavior.

    So maybe the problem isn't that it is hard to explain expectation shocks, but instead that the view of reality (that randomness exhibits a particular type of order) is fundamentally incorrect.

    ReplyDelete
    Replies
    1. Diego, that is a possibility. Would love to see something like that formalized (note that there are many papers attempting to formalize notions of Knightian uncertainty.)

      Delete
  6. Relatedly, I got involved in a long debate concerning the Underconsumption Hypothesis in December at Interfluidity that became the subject of three other blogposts the first of which (by Winterspeak) is here (you can find links to the other two posts by Ramanan and Nick Edmonds there):

    http://www.winterspeak.com/2013/12/the-underconsumption-theory-and.html

    Much later, two other bloggers wrote about my comments (Steve Roth of Angry Bear and troll blogger Mike Sax of Diary of a Republican Hater).

    Needless to say this is a contentious topic.

    Maki and Palumbo (2001) estimated savings rates by quintile in a manner consistent with the NIPA accounts (in my opinion the only way that matters in Underconsumption Theory), and they showed the following for the year 2000 (Table 2):

    Bottom quintile : +7.1%
    Second quintile : +7.4%
    Middle quintile : +2.9%
    Fourth quintile : +2.6%
    Top quintile : -2.1%

    http://www.federalreserve.gov/pubs/feds/2001/200121/200121pap.pdf

    That's right, the higher your income, the less you save.

    This should not be surprising. First, there has been no increase in savings over time as average (not median) incomes have gone up. Second, the increase in concentration of income towards the top in the US has been accompanied by a decrease in the average savings rate, not an increase.

    Furthermore, if there was a concentration of savings at the top income levels we should see a relationship between income inequality and aggregate savings rates in cross-sectional studies. But in the largest, and by most accounts, best study to date on this question, by Schmidt-Hebbel and Serven (2000), involving World Bank data on 82 countries over 1965-1994, they generally found no significant correlation between measures of income inequality and aggregate savings.

    The lone exception was in a regression that used the income share ratio of top 20% to bottom 40%. The correlation was significant at the 5% level, but was negative. In other words, it suggested that greater inequality resulted in lower aggregate savings.

    http://www.sciencedirect.com/science/article/pii/S0304387800000638

    Now in the debate I mentioned that I saw no reason why MPC could not decrease with income or wealth even if savings rates do not, after all, one is a marginal and the other is an average. In fact the Carroll and Kimball paper that John Aziz linked to above was thrust under my face early on, and I pointed out it was not at all inconsistent with what I was saying. But this got me thinking, if the conventional wisdom on savings rates is so wrong, could the conventional wisdom on MPCs also be wrong?

    Note there are a boatload of theoretical models such as Carroll and Kimball that show MPC goes down with wealth, but is that how people actually behave? What we need are empirical studies, and lo and behold the following one by Sahm, Shapiro and Slemrod (2009):

    http://www.nber.org/papers/w15421.pdf?new_window=1

    In particular look at Table 7. They find no significant difference in plans to spend the 2008 Tax Rebate by income level, although those in the highest income category were more likely than any other group to spend the rebate. They do find a significant difference (p-value = 0.078) by wealth class (stocks), but again, they found those that were in the highest wealth category were by far the most likely to spend the rebate.

    In short there is a lot of ingrained assumptions, proof by assertions and empty theorizing in the econblogoshere. But when you check to see how people actually behave the empirical evidence often paints a very different picture.

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
    2. Using the Permanent Income Hypothesis framework, I can see some potential explanations for a rising MPC with income. Maybe those with higher incomes tend to work later in life (less need to save for retirement) and maybe those with higher incomes tend to expect a larger increase in income in the future (borrowing is simply smoothing). Finally, maybe the stats are skewed by a few people with high incomes but even higher wealth (from family members), who will dissave over time.

      Also, I'm not sure how we compute "saving" but perhaps it is not the best-defined concept. If a young person with a large income borrows to buy a house, then she is dissaving. Yet, that house functions as an instrument for saving.

      Delete
    3. Very interesting, Mark -- as usual.

      Delete
  7. Good points. As the recent series of dueling papers by Auerbach/Gorodnichenko and Ramey showed, the evidence on spending multipliers is inconclusive, and highly sensitive to assumptions. Transfer multipliers are likely lower than spending multipliers.

    ReplyDelete
  8. This comment has been removed by the author.

    ReplyDelete
  9. The funny part of this is that this is a "pissing contest" with very long lags. Krugman's post links to a Barro WSJ article from more than 2 years ago. More like a cricket match, I think.

    ReplyDelete
    Replies
    1. Ha, I didn't even notice that Steve. I'd say it's more like a pissing cricket match.

      Delete