Sunday, March 21, 2010

Target the Cause Not the Symptom

Olivier Blanchard of the IMF recently made the case for monetary policy targeting a 4% inflation target instead of the standard 2% target. His main argument for doing so is that it would make the zero bound on the policy interest rate less of an issue. Here is how the Wall Street Journal summarized his view:
At a 4% inflation rate, Mr. Blanchard says, short-term interest rates in placid economies likely would be around 6% to 7%, giving central bankers far more room to cut rates before they get near zero, after which it is nearly impossible to cut short-term rates further.
There was a lot of push back on this argument in the blogosphere from folks like Ryan Avent, Mark Thoma, and David Altig who countered that (1) the zero bound isn't really a constraint for monetary policy, (2) higher inflation will lead to increased relative price distortions, and (3) there is mixed evidence and thus less certainty on the benefits of higher inflation. While all of these points are valid, I think there is a more fundamental problem with Blanchard's view: inflation targeting at any rate is merely responding to the symptom not the underlying causes--shocks to aggregate demand (AD) and shocks to aggregate supply (AS)--of macroeconomic volatility. This is problematic because monetary policy can only meaningfully counter AD shocks and therefore, it must be able to discern which shock is driving inflation. Inflation, however, can be hard to interpret. For example, if there is a sudden burst of inflation is it due to a positive aggregate demand shock (e.g. sudden, unsustainable increase in household spending) or a negative aggregate supply shock (e.g. temporary spike in oil prices)? In the former case inflation targeting would act appropriately by reigning in excess spending while in the latter case inflation targeting would only make matters worse by further restricting economic activity. Rather than targeting inflation, then, monetary policy should directly target the underlying source of macroeconomic volatility over which it has real influence, AD. Doing so would have gone a long way in making the U.S. economy during the 2000s more stable, a point I have made repeatedly during this crisis.

The importance of targeting AD can easily be illustrated using an AD-AS model. Here I use the AD-AS model developed by Tyler Cowen and Alex Tabarrok in their new macroeconomic textbook. Their version of the AD-AS model places growth rates on the two axis rather than levels. Below is the model in equilibrium with an AD growth rate of 5% that can be split up into an inflation rate of 2% and a real growth rate of 3%. (Click on figure to enlarge.)

Now consider four shocks to the economy when monetary policy is solely targeting an inflation rate of 2%. First, let see what happens when there is a positive AD shock driven by say expansionary fiscal policy (click on figure to enlarge):

The positive AD shock pushes the economy beyond full employment, increases inflation to 3%, and real growth jumps to 4%. AD is now growing at an accelerated rate of 7%. Fed officials seeing the higher inflation tighten monetary policy to get back to 2% inflation and in so doing push the economy back to full employment. Here the 2% inflation target worked just fine and effectively served to stabilize AD at 5% growth.

Now consider a negative AD shock caused by say a sudden collapse in economy certainty (Click on figure to enlarge).

The negative AD shock causes inflation to fall and turn into -2% deflation while the real growth rate falls to -3 real growth rate. AD is now at a -5% growth rate. Fed officials see the deflation and loosen monetary policy to get back to 2% inflation. In so doing they push the economy back to full employment. Here again, the 2% inflation target worked fine and effectively served to stabilize AD at 5% growth. Scott Sumner argues this type of shock was behind the Great Nominal Spending Crash of late 2008, early 2009. If so--I buy his argument--then inflation targeting would have done wonders during this time for the Fed.

So far inflation targeting is doing its job. But so far we have only encountered AD shocks. How well will inflation targeting work with AS shocks? Consider first a negative AS shock due to say a temporary disruption of the oil supply. Given the temporary nature of the shock, the short run AS (SRAS) curve would shift left.

The negative AS shock causes inflation to increase to 3% and slows down real growth to 2%. Under pure inflation targeting, Fed officials would see the pickup in inflation and respond by tightening monetary policy to bring inflation down to 2%. As the second figure above shows, though, such a response would slow AD to a 2.5% growth rate and only further weaken the economy. A far better response would have been to do no harm by keeping AD stable at 5% growth. AD would still be growing at 5% and given the temporary nature of the negative AS shock, the economy would eventually return to full employment. Inflation targeting fails to stabilize here.

Now consider our final scenario: a permanent positive AS shock due to say an major improvement in technology (i.e. positive productivity shock). This shock would cause LRAS to shift right.

The positive AS shock causes inflation to drop to 1%, but increases real growth to 4%. Fed officials see the drop in inflation and respond by loosening monetary policy to bring inflation back to its 2% target. Doing so, however, increases AD to a 7% growth rate which pushes the real economy beyond full employment to an unsustainable 5% growth rate. Here too, a far better response would have been to do no harm by keeping AD stable at 5% growth. Such a response would keep the economy at full employment instead of entering a boom-bust cycle. Once again, inflation targeting fails to stabilize in response to an AS shock. Note that this positive AS shock creates benign deflationary pressures which are different than the malign deflationary pressure created by the negative AD shock above. This distinction helps us understand the difference between the deflation scares of 2003 and 2009.

So what are the main takeaways from this analysis? First, inflation targeting is only effective when AD shocks are the main source of macroeconomic volatility. If AS shocks are also important,then inflation targeting can be destabilizing. Second, a far more effective approach to minimizing macroeconomic volatility is to stabilize AD. In the above scenarios, stabilizing AD growth around a 5% target was all that was needed.

An obvious critique of the above analysis is that the Fed does not only aim to stabilize inflation but also looks at the output gap given its dual mandate for price stability and full employment. And there is ample evidence that the Fed has done this to some extent by implicitly following a Taylor Rule. Josh Hendrickson even shows the Fed during the Great Moderation effectively targeted AD. While all of this is true, it is also true that the Fed does not explicitly follow the Taylor Rule and has used this flexibility to deviate from it at certain times in a manner that is consistent with pure inflation targeting. Stated differently, in such times the Fed has been more likely to err on the side of stabilizing inflation than stabilizing AD. For example, during the deflation scare of 2003 of the Fed acted more like a pure inflation targeter--it aimed to stabilize inflation even though there was evidence of rapid productivity gains pushing down the inflation rate rather than it coming from a weakening economy--and as a result AD growth at that time was anything but stabilized. A more fundamental problem with this critique is that the Fed has simply failed too many times to stabilize AD. Case in point is the Great Nominal Spending Crash mentioned above where AD had its largest decline since the Great Depression of the 1930s.

The bottom line is that monetary policy should target the cause not the symptom of macroeconomic volatility over which it has control. Olivier Blanchard's proposal does not do that. If anything, it would only serve to make the Fed even more focused on the symptom, inflation, at the expense of stabilizing the cause, AD.


  1. It seems to me that you could easily extend this Cowen\Tabarrok dynamic AD-AS model to labor as well. The vertical axis would represent wage inflation and the horizontal axis the annual growth rate of the available labor force. The LRAS for labor would be vertical while the SRAS would be positive sloping.

    What this could readily show is that attempts by the Fed to stabilize price inflation in the short run due to supply side shocks destabilizes the growth rate of the labor force.

  2. Richard,

    Interesting point. I believe George Selgin does something similar in his book Less Than Zero but does not use the growth rate approach you suggest.

    BTW, your earlier comment about the McCallum rule makes sense. If the velocity part of the rule could be made forward looking or (at least not so backward looking) I suspect it would have been capable of providing a much better response that what we got from the Fed.

  3. David
    If only McCallum´s NGDP rule had won the "rules debate" back in the 1980´s...
    Greenspan in practice targeted AD (or NGDP) because: (a) Without consciouly trying to do so he "froze" total reserves and (b) The M´s (M2, say) were strongly deregulated so that changes in V were closely compensated by changes in M2.
    In the 90´s Greenspan recognized the effect of rising productivity on RGDP growth and resisted raising FF until mid 1999 when RGDP growth climbed above 5% YoY.
    In 2002-04 maybe he was influenced by Bernanke (Remember the Nov 02 speech: "Making sure "it" doesn´t happen here"?)
    Even if all shocks are of the AD type, targeting inflation - especially given the steep fall in output that happened this time around - would still leave the income path far below the original one.
    And Bernanke has as far back as 1999 and again in 2003 reccommended that Japan target the price level!
    So what he says is good for others he doesn´t think is something that he should do (remember his answer in December to DeLong´s question: "Why don´t you adopt a 3% inflation target"?)

  4. And I forgot to mention that in 2007-08 the Fed was all towards targeting inflation. Only that for some strange reason they focused on headline CPI that was going up due to oil and commodities. Meanwhile Core and Median CPI were stable.
    Just check the Stetements over July/07 Sept/08. Inflation was the greater risk (on several occasions there was dissent).
    No wonder everyone came to expect that AD would continue to be restrained...

  5. 5 years ago as a well-trained New Keynesian, I would have thought this great stuff....but I think the crisis has taught us that the standard overly-aggregated macro model growing out of Keynes and Friedman is decidedly not the way to go....
    Given your clear influence from Austrian sources David, I am surprised to see you still pushing these antediluvian Keynesian concepts of "Aggregate demand" and "Aggregate supply". It is beyond the wit of any central planner to successfully target any "Macro" variable, be it nominal GDP, broad money supply or price level. Surely this is the lesson from almost 100 years of Federal Reserve history! It is neither technically feasible and even if it were, public choice tells us that it would not be administered technocratically.

  6. ECB,

    I am not big on central planning but given that we do have a central bank we should try to make it as effective as possible. I am not saying my approach is perfect or even easy to do (Though it has never really been explicity tried.) I am just trying to be pragmatic given our current monetary arrangements. Larry White has shown that Hayek himself thought stabilizing nominal spending was the best goal for monetary policy.

  7. Very clear discussion, and separation of AS vs. AD shocks. Other separations are expected/future vs. past trajectory, and level vs. rate (ssumner's issue). Thanks. At some point it would be nice to see the discussion move to specific mechanisms for targeting AD; I'm not quite convinced that merely announcing the target will by itself make the target occur, as some might hope.

  8. Statsguy:

    Thanks. You are correct that there is more to consider with this debate. Among others things, I wanted to speak to the level vs. growth rate issue, but avoided doing so given the length of the post. I hope to return to it later.

  9. StatsGuy and David,

    It is all well and good saying that you want to target NGDP, but all the Fed can ultimately do is play with the mix of risk free assets in the economy (i.e. reserves relative to other govt paper). The problem with Sumner’s proposal, quite apart from the difficulties in creating a working NGDP prediction market, is that it relies on the assumption that the Fed has the magical ability to determine total income. How is this to happen? This is not a trivial question.

  10. Fair enough, if you want to be heard in the mainstream, you have to couch things in the language they will deem acceptable. If one starts campaigning for alternative monetary regimes, you will no doubt be marginalized. Pragmatism versus radicalism.

  11. this was a great blog, very helpful..i have been reading roche,summer,mosler and you for the past 6 weekends..never read economic blogs or any blogs before...can you tell me what is the gap right now(in trillions of $'s) between what our economy is producing and what it could produce at full employment/full capacity utilization and would it be producing at this level if nominal GDP had been growing at 5% per year