Friday, September 13, 2013

Sluggish Growth: What's Going On?

As we're approaching the 5th anniversary of the Lehman collapse, and everyone seems to want to write about the fallout from the financial crisis, now seems a good time to take stock of the macroeconomic consequences of that event. The first chart shows the log of U.S. real GDP, and the linear trend that best fits the whole post-1947 time series.
That's quite stunning. In terms of our standard measure of aggregate economic activity, the U.S. economy did not "recover" from the previous recession in the usual sense. Real GDP is far below the post-World War II trend, and is growing at a rate less than the average growth rate over the period 1947-2013. If you have read Reinhart and Rogoff's book, maybe this doesn't surprise you. Reinhart and Rogoff tell us that recoveries from financial crises tend to take a long time - financial crisis shocks have very persistent effects. So, if we think that Reinhart and Rogoff know what a financial crisis is, and all those previous crises really have something in common with our experience in 2008-09, then that gives us something to go on. Maybe what we're seeing in the above chart is related to financial factors.

But what exactly are the financial shocks associated with a financial crisis, and how do those shocks propagate themselves? Unfortunately, Reinhart and Rogoff aren't going to tell us much about that. What's going on? I think most economists recognize that policy played some role in causing the financial crisis - if only through poor regulation - but does our current sluggish growth path represent some dramatic failure of economic policy post-crisis? Are policymakers such dummies that they have failed to implement obvious solutions? Is there some unusual market failure at work that we have failed to understand? Or is what we're seeing actually the best possible outcome we could have hoped for, given what happened in 2008-09?

Old-Keynesian IS-LM AD-AS macroeconomists, who still exist, much to the surprise of many of us, think they have an answer. However, I've found the arguments based on that paradigm turn into a web of contradictions, so there's not much point in taking those arguments seriously. As George Bernard Shaw said
Never wrestle with pigs. You both get dirty and the pig likes it.
New Keynesians, on the other hand, are serious people. They have models, which incorporate bits and pieces of well-accepted economic theory, and they fit those models to data. The New Keynesian story about the recent recession is that there is a shorthand for a financial crisis shock - it's a preference shock. Consumers became much more patient during the financial crisis, and that patience has persisted. Unfortunately, given that type of shock, conventional monetary policy cannot correct the ensuing sticky price problem, because of the zero lower bound on the nominal interest rate. With no movement in anticipated inflation, the adjustment that has to occur is a reduction in consumption and output.

But prices will adjust over time of their own accord, and the sticky-price inefficiency (relative prices out of whack) will go away. The question is, how quickly does this happen? Is this a case of "in the long run, we're all dead," or what? Well, considerable effort has gone into studying price adjustment at the micro level, of late, so we know something about this. The best summary of evidence, to my knowledge, is in Klenow and Malin's chapter from the recent edition of the Handbook of Monetary Economics. Klenow/Malin conclude, among other things, that the average length of time between price changes is about one year. That's a longer period than what Bils/Klenow reports, which was a period of four months. The difference seems to be due to the fact that Klenow/Malin choose to exclude price changes from sales, for example. Toothpaste could be priced at $2.00, go on sale for a week at $1.50, then go back to $2.00, and Klenow/Malin would not count that as a price change. So they seem to be bending over backward to make prices look more sticky. But even if all prices change once every year, if we suppose that there was one financial-crisis shock that occurred in the fall of 2008, then five years later we won't have much residual inefficiency due to price stickiness. Further, Klenow/Malin report that prices are more flexible for goods that are cyclically sensitive, i.e. sectors of the economy for which employment and output are more variable relative to trend. Since those are the sectors where the friction would have to matter in delivering the large drop in output we're seeing in the first chart, it's hard to see why a New Keynesian sticky price mechanism is explaining what we're seeing.

Further, what we should be seeing, according to New Keynesian models, is real GDP returning to the post-war trend, and a gradual increase in the inflation rate as that happens, as we should be climbing the New Keynesian Phillips curve. But that's not what's happening. People at the Fed and elsewhere are using New Keynesian models as forecasting tools - that's in part why they've been over-optimistic. In a New Keynesian model, a shock hits, and there is some direct effect from the shock (what the frictionless model delivers) and the indirect effect from the sticky price friction. Ultimately the indirect effect dies out, and the economy returns to it's TFP-growth-driven trend. Whether the nominal interest rate is at the zero lower bound will make a difference, as away from the zero lower bound monetary policy can give quicker adjustment of relative prices. But sellers are not going to change their prices any more or less quickly because the nominal interest rate is zero than otherwise, everything else held constant.

There's an additional puzzle associated with that idea. The next chart shows my crudely constructed measure of TFP (total factor productivity), using annual capital stock data (available only up to 2011), and household-survey employment.
There's no particular problem with TFP growth. Indeed, TFP grew at a relatively high rate in 2010 and 2011. The average rate of growth from 2000-2011 is about 0.9%, which is lower than the post-WWII average of about 1.2%, but that's not bad, given the effects of two recessions over that period. Thus, the U.S. economy is very productive, but the quantity of labor input is really low, as we all know.

Speaking of employment, it's useful to look at what is happening across sectors of the U.S. economy. The next chart shows some selected components of employment, for the period 2000-2013.
I've normalized each time series to 100 at the first observation. The key message here is dispersion - employment growth has differed markedly across sectors over this period. Employment in durables and nondurables manufacturing fell by about 30% over the thirteen-year period while, at the other end of the spectrum, employment in education and health care grew by almost 40%, and in leisure and hospitality by more than 20%. Construction, which was an employment-growth sector until 2006, suffered the largest employment decline during the recession. Thus, some of the dispersion in employment growth is the result of the recession (construction), but a good part is the result of secular changes in the composition of employment. This is important, as we want to understand more about these long-run changes in sectoral activity - what is due to technology, what is demand shifts, what might be due to the financial factors at work during the recession.

I've written before about how it is difficult to square the experience we've had with sectoral shifts with sticky price theory. Relative stickiness in prices has implications for secotoral employment over the business cycle, but those implications are not consistent with the above picture.

What about wage stickiness? That's certainly important in the Keynesian narrative, though New Keynesians tend to think more about sticky prices. Suppose, for the sake of argument that a particular worker's wage had been stuck at its pre-financial crisis nominal level until now. How much would that worker's wage have declined in real terms by mid-2013? The answer depends on when the worker's wage became stuck. If it was in January 2007, the decline would be 12% (using pce inflation); if in December 2007, 9%; and if when Lehman went down, 6%. That's large in any case, and that's with zero adjustment over 5 or 6 years. You think wage stickiness matters over that length of time, or that wage stickiness somehow explains the drop in employment we saw in the construction sector? I don't think so.

Another useful exercise is to look at the time series for the components of GDP since pre-recession times. The next chart shows these, leaving out net exports.
Then, compare that to what we see for the 1981-82 recession. In the next chart, I've made the sample period the same length as in the first chart, and the first observation is the same number of quarters before the beginning of the recession, the way the NBER dates it.
First, focus on government expenditure. At the turning point in the recession, that's actually higher, relatively, in the recent recession than in 1981-82. You can see the effects of the stimulus and, if you believe in stabilization policy, the timing is really good - government expenditure is high at the turning point, and until mid-2010, before declining. That decline is interesting, as you might think - if you listen to people who argue that the multiplier is really large - that this would have been a disaster. But I don't see the multiplier at work when I stare at that chart. Consumption just keeps growing at a steady rate, though that rate is lower than what we usually see coming out of a recession - for example, look at what happened in the 1981-82 recession.

Probably the most important feature of the data in the two charts is the difference in the behavior of investment. In 1981-82, investment declines by about 12% from the first observation, then rebounds significantly, to the extent that it grew more than consumption and output by mid-1983. In the last recession, investment declined by more than 30% from the beginning of 2007 to mid-2009, and in second quarter 2013 was still about 5% lower than in first quarter 2007. Thus, if there is something we should be focusing on, it's not multipliers and consumption, but why investment is so low. That low level of investment, over a five year period, has now had a significant cumulative effect on the capital stock. Thus, we've got OK growth in TFP, but growth in factor inputs is low. That's the key story, from a growth accounting point of view.

So, what to make of this?

1. There are many frictions to think about. (i) There are financial frictions. Limited commitment and private information, coupled with a low supply of safe assets, can lead to the misallocation of capital and labor in the economy, for example. (ii) There are distortions caused by inefficient taxation and subsidies - for example in the housing sector. (iii) There is inefficient regulation at work, particularly in the financial sector, as we know well. (iv) Heterogeneity and private information in the labor market creates matching frictions, which can be exacerbated by sectoral shifts, and by financial frictions. With all that to think about, why focus on sticky wages and prices? Some people want to take the persistent level decline in GDP, and the persistently low GDP growth rate, as indications that sticky wages and prices are really important, when they should be thinking the opposite. As sluggishness persists, this casts further doubt on sticky wage and price frictions as being the source of the problem. People who cling to those sticky ideas, particularly the Old Keynesians, are either being lazy, or are simply unimaginative.

2. What should policy be doing? This is not a problem for monetary policy. The real effects of Fed actions, outside of crisis periods, are small and transitory at best. Whether by accident or design, the Fed has been doing a good job of controlling inflation, and should focus on that. If there are remedies to the poor real-side performance of the U.S. economy, those are fiscal remedies. There are plenty of things to think about: (i) further financial reform; (ii) housing and mortgage market reform - get rid of housing subsidies, Fannie Mae, and Freddie Mac, for example; (iii) health care: Congress should stop fussing about Obamacare - which is imperfect, but nevertheless an improvement - and do something useful for a change; (iv) education: If you believe the cross-country test results, the quality of American education in K-12 is declining, and we know there are huge disparities in access to good education. Some work points to a big payoff from early childhood education; (v) the problems in the north side of St. Louis, and in many other American cities, look very similar to those in the poor countries of the world. The problems may seem intractable, but we should be putting more effort and resources into solving them.

25 comments:

  1. Steve, nice analysis until your proposals. You argue rightly, that sticky prices should have adjusted by now, but clearly health, education, city problems whatever have adjusted, to tHe extend that nothing has changed and they are at pre-crisis level. Why would a change in education policy lead to a a short run increase and strong recovery in investment? I dont think so either, unless you come up with a story that the shock, i.e. the magnitude, interacts strongly with the level of the given institutIon. Thats of course conceivable, and I can make up a model, but is that your point? A nonlinear interaction effect between the level of school education, the financial shock and the recovery in investment?Frankly, you argue it seems outside a quantified model.

    ReplyDelete
    Replies
    1. No, I'm not saying these things are the causes of what is going on in the first chart. I'm saying that if you want to mitigate or totally offset those things, this is what you should be thinking about. Of course it helps a lot to understand exactly what the causes are - maybe it's possible to just do some things in reverse.

      Delete
  2. More pix: http://econsnapshot.com/2013/08/31/upward-revision-to-real-gdp-signs-of-improvement/

    ReplyDelete
    Replies
    1. Dave,

      Thanks for pointing that out. Cooley and Rupert's "U.S. Economic Snapshot" provides a nice summary of business cycle data, particularly related to the recovery, or non-recovery.

      Delete
  3. Steve: since investment is so pathetic, how about a cut in the corporate tax rate? A man can dream can't he?

    ReplyDelete
  4. Very interesting thoughts. About the sticky prices, I find it hard to disagree...what's interesting is that even some of the monetarists who are really in favor of monetary policy approaches seem not to care that much about whether prices really are sticky or not. They seem to me not to really believe in sticky prices, but to view it as a useful modeling proxy for some friction that they believe is real but don't understand.

    Have you gotten that impression?

    ReplyDelete
    Replies
    1. We can't read minds, so it's hard to know what is going on in there. By "monetarists," I take it you mean the people who are promoting NGDP targeting. That's about as close as anyone gets to old monetarism these days, but it has the same flavor. Some of Friedman's ideas about business cycles were like that. He had a strong view of the policy prescription, but was vague about what the short-run non-neutrality of money is.

      Delete
    2. Some of Friedman's ideas about business cycles were like that. He had a strong view of the policy prescription, but was vague about what the short-run non-neutrality of money is.

      Huh. Interesting.

      So now we have massive QE-type things and a more loose-money stance in Japan, and it seems to be having some effects. Even given that we don't yet understand the source of monetary non-neutralities, does it make sense to conclude that "This is not a problem for monetary policy"?

      Delete
    3. 1. No, it's not that "we" don't understand. I was guessing you were referring to the NGDP targeters. I don't think they understand much. However, if you take all the monetary theory we have, and all the empirical evidence, put that in someone's brain, and then, provided the brain is a good one, ask that person what we should do now in the United States, I think they would say: Well, you have pretty much thrown everything you can, in terms of monetary intervention, at this short-run problem you think is there, so you're done. There seem to be some monetary non-neutralities out there, we have some ideas about what those are, and any of those theories tell you that you have nothing left to exploit.

      2. On Japan: I don't pretend to understand what is going on there. At best the jury is out on that - we have to wait and see how it unfolds. What we conclude from that episode may or may not be important.

      Delete
  5. "Thus, if there is something we should be focusing on, it's not multipliers and consumption, but why investment is so low." The NIPA fixed investment measure is broken down into nonresidential FI and residential FI. The two components of nonresidential FI are structures and equipment & software. The Cooley-Rupert snapshots show that both residential and nonresidential structures have not yet reached their pre-recession peaks. As of 2013q1, growth in equipment & software FI relative to 2007q1 was about 8% (similar to what's seen for consumption & GDP in the above graphs). So, it's growth in FI in structures (both residential & nonresidential) that's been particularly low, and presumably that's related to all of the real estate fun we've had.

    ReplyDelete
    Replies
    1. Yes, exactly. I think we can think of that as the same phenomenon. The shopping malls go with the housing developments.

      Delete
  6. There's no particular problem with TFP growth. Indeed, TFP grew at a relatively high rate in 2010 and 2011.

    Steve: I seriously wonder how much of this apparent TFP growth is due to composition bias (the fact that less productive workers and firms are let go or remain idle in a downturn).

    ReplyDelete
    Replies
    1. Yes, I was thinking the same thing. Possibly more important, if that is what is going on, is the sectoral composition of output. I'd guess that housing construction is a low-productivity activity, for example.

      Delete
    2. New firm formation is depressed. New firms have smaller scale and therefore lower productivity. Taking this one step further, in an economy with less entry an exit, firms would enjoy more market power and investment would be lower.

      This is anecdotal but here goes anyway. What I am struck with as an entrepreneur is the wall of capital directed at expensive legacy assets (i.e. REITs) while new firms are starved of start-up capital. To me, Fed policy seems partly to blame for the focus on safe, subsidized "reach for yield".

      Delete
    3. David,

      Why don't we observe this pattern in previous recessions? I know that Prescott argues we are measuring TFP incorrectly, but I'd be interested to get your thoughts.

      Delete
  7. I like David Andolfatto's Japan paper from 2003:

    What is likely needed are economy-wide reforms that enhance the
    willingness and ability of individuals to adopt potentially disruptive technolog-ical advancements and work practices

    Technology disruption is hamstrung by patent law and government subsidy to too-big-too-fail orgs.
    Espinosa' comment on new firms seems a propos

    ReplyDelete
  8. Krugman and the old Keynesians think the problem is that the invisible hand isn't working (sticky prices).

    But actually it is the invisible foot that isn't working - the invisible foot being the kick to all-too-comfy incumbents from new upstarts with disruptive technology.

    ReplyDelete
    Replies
    1. I had never thought of it that way. The usual story about "creative destruction" and the business cycle downturn is that you lose the inefficient firms, and there is a cleansing effect. But in boom times, there is another kind of cleansing effect, which comes from new startups taking over the turf of the lazy incumbents. Interesting.

      Delete
    2. Steve,

      Aghion and Howitt (1994) deal with this issue in "Growth and unemployment". Innovation has a destructive effect (shortens the duration of employment matches) and a capitalization effect (increases vacancies by creating new profitable opportunities). In their model exogenous parameter changes that speed-up innovation have no impact on unemployment because the two effects cancel each other out. However, under different assumptions Sener (2000) finds a negative effect.

      Of course, as I have mentioned before, my research points to an explanation that also has Schumpeterian flavor but works somewhat differently. The idea based on Acemoglu (1999) is that the diffusion of a general purpose technology leads to greater technical heterogeneity that reduces the sectoral mobility of workers and therefore matching efficiency, as in Barnichon and Figura (2011). By now I have tried several different estimation techniques to appease various referees, and I keep getting the same results: Efficiency declines when the sectoral dispersion of the stock of software rises.

      So it seems to me, and this is something that needs further investigation, that much of what we are seeing is due to the third wave of the IT revolution (1st is mainframes, 2nd PCs, and 3rd the internet). The over-investment in housing and consumption durables (refrigerators, stoves, etc.) brought about by tax distortions, financial frictions, etc. would have caused a recession regardless. However, the recovery takes longer because the resources that were released are less able to move to alternative uses than in the past, because doings so now requires technology-specific skills. This may explain why in the last two years vacancies have increased but unemployment has not declined accordingly.

      Delete
  9. I think you are right to single out the slow investment spending as a driving factor.

    My own take on that development is in a paper I did in 2010 (basically there's a capacity overhang dating back to the second half of the 1990s). It can be found here http://www.boeckler.de/pdf/v_2010_10_29_milleker.pdf

    The analysis also concluded that a US unemployment rate in the 7 to 8% range would be normal in the following years (not too bad, given that the conventional view at the time was for unemployment to normalise quickly to something like 5.5% at the time.

    ReplyDelete
  10. An article which describes the potential causes for the worst recession since the Great Depression and does not use the word demand (as in aggregate demand shortfall) or debt (as in households, firms and banks gotta deleverage) is worthless

    ReplyDelete
    Replies
    1. a blog reply whch uses the word demand is worthless. Gosh how easy it is to play this game ! Fun !

      Delete
    2. John D, we know it's you.

      Delete
  11. About associating Old Keynesians with sticky prices and wages, well, last time I checked this school of thought did not care about why there can be persistent demand shortages on the labour and product market but rather analyzed the consequences of them.

    About associating the name of Keynes with stick wages and prices, well, somebody direly has to read Ch.19 of the General Theory. Keynes correctly pointed out that sticky wages are not the source of the problem; falling wages would reduce the cost of labour and thus lead to nice supply side effects but the entire problem is a shortage of demand which is aggravated by a reduction of wages.

    I am all for serious criticism and taking apart Keynes or Hicks/Samuelson/Tobin. But you first gotta familiarize yourself with what you wanna take apart.

    ReplyDelete
    Replies
    1. The 1960s called...they want their textbook back. We have moved on pal, so stick to commenting on analog blogs like Thoma where you'd be more at home. You are creating a negative externality to an interesting discussion, such as CA's.

      Delete