Reply to John Taylor

John Taylor has a new post criticizing NGDP targeting:

One change is that, in comparison with earlier proposals, the recent proposals tend to focus more on the level of NGDP rather than its growth rate. This removes some of the instability of NGDP growth rate targeting caused by the fact that NGDP growth should be higher than its long run target during the catch up period following a recession. But it introduces another problem: if an inflation shock takes the price level and thus NGDP above the target NGDP path, then the Fed will have to take sharp tightening action which would cause real GDP to fall much more than with inflation targeting and most likely result in abandoning the NGDP target.

I see lots of problems here, but in fairness this may reflect my particular vision of NGDP targeting, which is the “target the forecast” approach.  Under my plan, the Fed would constantly adjust policy so that expected future NGDP (12 or 24 months forward) remained right on target.  Ideally this would involve NGDP futures markets, more likely it would involve an internal Fed NGDP forecast, which also incorporated the consensus private NGDP forecast as well as various asset prices such as TIPS spreads.

Taylor is right that there might be inflation shocks under NGDP targeting, just like there are under inflation targeting.  For instance, the rise in oil prices in 2007-08 caused CPI inflation to rise far above the Fed’s implicit target.  But he’s wrong in assuming these inflation shocks would raise NGDP, indeed NGDP growth slowed to well under 5% during the 2007-08 oil shock.

The deeper problem with this criticism is that wages are not set on the basis of expected inflation, but rather the expected rate of NGDP growth.  That’s why wages in a country like China have been rising at double digit rates for years, despite much lower inflation rates.  It is why wages in the US remained well behaved in 2007-08, even as headline inflation rose to over 5% (as NGDP growth was slowing.)  As long as the Fed keeps targeting NGDP expectations, wage growth will remain anchored.  Workers and employers understand that wages cannot compensate for every spurt in prices at the gas pump.  If actual NGDP does change suddenly, it will be easy to reverse as long as expected future NGDP (and hence wages) remain on track.  (Note; this argument applies best if the Fed targets NGDP per capita, or per working age adult.)

If the Fed had been targeting inflation in 2008 the crisis would have been far worse, as monetary policy in mid-2008 would have been much tighter.  The Fed actually takes both inflation and real growth into account.  But an NGDP target would allow them to do so in a much more explicit fashion, and would have allowed them to ease much more aggressively in late 2008.

Taylor continues:

A more fundamental problem is that, as I said in 1985, “The actual instrument adjustments necessary to make a nominal GNP rule operational are not usually specified in the various proposals for nominal GNP targeting. This lack of specification makes the policies difficult to evaluate because the instrument adjustments affect the dynamics and thereby the influence of a nominal GNP rule on business-cycle fluctuations.” The same lack of specificity is found in recent proposals.

That may be true of Romer and Krugman, but they were basically endorsing the proposals of others.  And certainly no one can claim that my proposal lacks specificity—it is just as rule-based as Taylor’s famous policy rule.  It also has the advantage of being forward-looking, which is a huge plus in a fast moving financial crisis like 2008.  The Fed used Taylor Rule-like reasoning in deciding not to cut rates below 2% in their September 16, meeting, which occurred right after Lehman failed.  They cited a roughly equal risk of recession and inflation.  Incredibly, the risk they saw was excessively high inflation, not excessively low inflation.  How could the Fed have made such a bone-headed mistake?  They were looking in the rear view mirror, at nearly 5% headline inflation over the previous 12 months.  They should have looked down the road as Svensson suggests, as the TIPS spreads that day showed 1.23% inflation over the next 5 years.  Taylor Rule-type thinking caused the Fed to unintentionally leave money way too tight to hit their implicit inflation and employment targets.

I’m sure that today even Ben Bernanke would agree that they erred in not sharply cutting rates at the September 2008 meeting.  During the fall of 2008 the Fed needed to do enough stimulus so that forecasts of 2010 NGDP remained roughly 10% above actual 2008 levels.  They didn’t even come close, which is why the recession was so much worse than it needed to be.  The sub-prime fiasco made a mild recession almost inevitable, but the fall in NGDP (the biggest since the 1930s), made it far worse than it needed to be.  Sharply falling NGDP expectations in late 2008 led to sharply lower asset prices, which dramatically worsened bank balance sheets.  IMF estimates of expected US banking system losses nearly tripled in late 2008 and early 2009, even though the sub-prime fiasco was already well-understood by mid-2008.  What wasn’t predicted in mid-2008 was the catastrophic fall in NGDP over the next 12 months.

At first glance Taylor’s piece looks like a critique of NGDP targeting.  But on close inspection it is something different.  It is a discussion of tactics; level versus growth rate targeting.  Rules versus discretion.  I’ve added the issue of forward-looking versus backward-looking rules.  These are all interesting issues, and I actually agree with Taylor on the importance of policy rules.  He is well aware that some of the most distinguished proponents of NGDP targeting (such as Bennett McCallum) have proposed explicit NGDP policy rules.  He also knows that the dual mandate embedded in NGDP targeting is not that different from the dual mandate embedded in the Taylor Rule.  Readers of critiques by Taylor and Shlaes need to keep in mind that their real target isn’t NGDP targeting, it’s discretion.  I hope John Taylor will consider jumping on board and writing an explicit “Taylor Rule” for NGDP targeting, so that if the Fed does move in that direction they do so in a responsible way.

BTW,  What’s the non-discretionary Taylor Rule suggestion for Fed policy if rates fall to zero and further stimulus is needed?

HT:  Marcus Nunes



30 Responses to “Reply to John Taylor”

  1. Gravatar of John Taylor “strikes back” at NGDPT | Historinhas John Taylor “strikes back” at NGDPT | Historinhas
    19. November 2011 at 07:52

    […] Scott obliged! GA_googleAddAttr("AdOpt", "1"); GA_googleAddAttr("Origin", "other"); […]

  2. Gravatar of bill woolsey bill woolsey
    19. November 2011 at 09:28

    By specific, he means a specific rule.

    r = 1 + 1.5 (inflation -2) -.5 gdp gap.

    (and he wants to drop that troublesome second term.)

    So, you would need to use r = 1 + 1.5 NGDP gap

    or something like that.

    r = 1 + 1.5 * expected NGDP gap as should by futures markets.

    Yes, it is true that McCallum used a feedback rule using NGDP growth.

    B growth = …. (I don’t really know what he did exactly) growth NGDP – 3

    That is what he wants.

    A rule for the target instrument. The overnight rate is OK with him, or base money.

    Goldman Sachs uses r = .6 (NGDP – 4.5) and if r , 0, 1 trillion in asset purchases for each percentage point to bring the effective target rate back to zero. So -5 = 5 trillion in quantitative easing and r = 0.

    Taylor is specifically rejecting the “as much quantititive easing as needed to get expected NGDP to target.” The rule has to be, NGDP is this far from target, so the Fed knows it must purchase so many bonds.

  3. Gravatar of John hall John hall
    19. November 2011 at 09:40

    Due to the strong relation between both expected unemployment and expected real GDP and between changes in unemployment and changes in log real GDP growth, I operate on the assumption that a Taylor rule can be approximated by something of the form:
    ENGDP+earlierNGDP=ERGDP+earlierRGDP+EPGDP. If you subtract out earlier NGDP growth, then a Taylor rule is like a NGDP growth rule where you don’t care about past prices.

  4. Gravatar of John hall John hall
    19. November 2011 at 09:41

    oopps sorry, NGDP level rule without caring about past prices, an NGDP growth rule where you adjust in the opposite direction for past prices

  5. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    19. November 2011 at 09:48

    I suppose this would be the time to report that I tried to get David Friedman, on his blog, to tell me what his father would have thought about this debate, but got a disappointing refusal;

    ‘Patrick asks what my father would have said about something.

    ‘Macro isn’t my field, and I can’t speak for what my father would have said. I suggest asking Anna Schwartz, as the nearest equivalent available.’

    My response to the above didn’t even merit a reply;

    ‘You may not be a macro specialist, David, but you wrote quite ably about monetary theory in your Price Theory textbook. In fact, I once pointed Sumner to your explanation of ‘the price of money’ when it’s bought and sold (rather than lent and borrowed) on his blog, The Money Illusion. He liked it.

    ‘But, if you’re not following that debate (such as Christina Romer’s open letter to Ben Bernanke, in which she essentially pumps for Sumner’s prescription of targeting the right side of the Equation of Exchange), so be it.

    ‘Also, and at risk of being ungallant, I don’t think Anna Schwartz even remembers, these days, what she wrote in the 1960s.’

  6. Gravatar of Brian Ritz Brian Ritz
    19. November 2011 at 11:42

    First of all, thanks for putting all the time you do into the blog. I’m a senior at Indiana University and have been following your blog (and many other econ blogs) since I was a freshman. I owe much of what I have learned about economic thinking to blogs like yours.

    I realize that John Taylor was bringing up questions about level targeting vs. rate targeting, and rules vs. discretion, etc., but I can’t quite get my head around what he is trying to say in the first paragraph you quoted.

    His statement, “if an inflation shock takes the price level and thus NGDP above the target NGDP path, then the Fed will have to take sharp tightening action which would cause real GDP to fall much more than with inflation targeting” is very puzzling to me. What does he mean when he says “inflation shock?” Increased velocity? Negative real supply-side shock? Under NDGP level targeting, I think the answer makes a huge difference when deciding if tightening (i.e. lowering the expected NGDP growth rate) is warranted.

    If “inflation shock” means something that suddenly increases money velocity, then under NGDP level targeting the Fed should tighten (just like they would with inflation targeting) to bring in NGDP on target. However, if “inflation shock” means something like an oil shock that lowers real incomes, I’m not sure the Fed would really need to tighten to bring NGDP down to target. The real shock wouldn’t increase expected NGDP, it would just affect the RGDP/Inflation split at a given NGDP level. In this case, it is under INFLATION TARGETING that the Fed unduly tightens and lowers RGDP lower than it would be under NDGP level targeting.

    Do I have all this straight or am I missing something?

  7. Gravatar of flow5 flow5
    19. November 2011 at 12:00

    He should talk. The “Taylor Rule” is backward looking, not forward looking. I want to know how to adjust monetary policy to hit nominal gDp in the current & subsequent qtr. Taylor’s equation won’t do that.

    Bloomberg “Economists at JPMorgan Chase & Co. in New York now see gross domestic product rising 3 percent in the final quarter, up from a previous prediction of 2.5 percent”

    Targeting nominal gDp is fool proof. You use monetary flows (our means-of-payment money X’s its rate-of-turnover). MVt’s rate-of-change has slowly risen the entire years as I posted.

  8. Gravatar of Benjamin Cole Benjamin Cole
    19. November 2011 at 12:08

    Boy, if even a John Taylor can mount only such feeble criticism of Market Monetarism….

    We might have inflationary shocks? We might anyway! And some inflationary shocks–such as oil prices—-are better accommodated rather than cranked down (unless you want deep recession with every upward and possibly speculative spike in oil prices).

    Not rules-based, says Taylor? Huh? Sumner has rules up the wazoo!

    In the end, the anti-NGDP crowd seems to come back to the failed principle that the need for very stable prices trumps all other policy concerns. This flies in the face of recent US economic history, as we prospered nicely from 1980-2008, with mild and varying inflation. And let’s see how Japan is doing.

    Many economists have developed an unhealthy and peevish fixation on inflation, which they conflate with lazy and corrupt federal spending. Or they detest Obama so much, they want the Fed to cut off the effects of his stimulus spending.

  9. Gravatar of ssumner ssumner
    19. November 2011 at 15:27

    Bill, I understand that, but I have a specific rule as well, and it doesn’t have the zero bound problem.

    John, Don’t you have that backward? With level targeting you adjust for previous overshoots.

    Patrick, Thanks for trying.

    Brian, I’m as confused as you are. I would think “velocity shock” would be more descriptive than inflation shock.

    flow5, Yes, backward-looking is a big problem.

    Ben, Yes, I am not seeing good arguments against NGDP targeting. Most people are grasping for straws.

  10. Gravatar of John hall John hall
    19. November 2011 at 17:15

    Perhpas the confusion might be I was just doing a quick post. By earlier NGDP, I meant the sum of the past n quarters NGDP. So ENGDP + earlier NGDP would be the level target.

    Anyway, it is possible to solve for the expected nominal GDP growth required (under any RGDP, PGDP, NGDP growth or level target) with the following formula:
    where g1 and g2 are weights on growth rates and l1 and l2 are weights on levels and C is a constant.

    So earlier NGDP and earlier RGDP are known so if you put a positive weight on g2, you have to have an opinion on what it should be as well.

    The Taylor rule, converted to a forward-looking RGDP level target (since that is like forward-looking unemployment), would thus be like NGDP growth, but without any weight on the history of prices. To do this in the above formula, you would have effectively set weights on g1=1,g2=0,l1=0,l2=1. You would set K at like 2% inflation plus 3%+n*3%/4 to account for the real GDP growth at the horizon and then solve for expected NGDP, which is what the central bank can control.

  11. Gravatar of Benjamin Cole Benjamin Cole
    19. November 2011 at 17:19

    BTW, when did 2 percent inflation become chiseled into granite? What if 3 percent inflation results in more robust real growth (as I suspect)?

    3 percent strikes me as precariously close to zero, and bad things start to happen at zero, as we are finding out.

  12. Gravatar of StatsGuy StatsGuy
    20. November 2011 at 06:24

    “But it introduces another problem: if an inflation shock takes the price level and thus NGDP above the target NGDP path, then the Fed will have to take sharp tightening action which would cause real GDP to fall much more than with inflation targeting and most likely result in abandoning the NGDP target.”

    The utterly ludicrous thing about this critique is that it applies even more so to inflation targeting. He’s literally getting his supply and demand curves backwards.

    When NON-LABOR INPUT COSTS drive inflation (oil), an inflation target will impose massively tight money to keep input costs under control by hammering AD, and the price of keeping input costs under control by necessity will drive NGDP very low (well below the target).

    An NGDP target is just slightly more flexible – it will still impose somewhat tighter money, but this will be balanced in so far as inflation will be allowed to rise slightly if the AD required AD hit to keep inflation on target is too heavy.


    John Taylor – “I’m the living embodiment of Not Invented Here”

  13. Gravatar of ssumner ssumner
    20. November 2011 at 07:20

    John, Thanks for working that out. I agree that NGDP can be made into a Taylor-type rule.

    Ben, Yes, any number is arbitrary.

    Statsguy, Great point.

  14. Gravatar of dwb dwb
    20. November 2011 at 08:02

    ahem, you could have quoted from page 82 of the paper:
    But whether a nominal GNP rule, another rule, or even no rule at all is instituted, focussing policy discussions on nominal GNP would in my view greatly improve policy performance. The aims of policymakers would be much easier to interpret if their goals for nominal GNP were clearly stated. The Fed in conjunction with COngress and the administration should state realistic forecasts of nominal GNP growth conditional on their intended plan for monetary and fiscal policy.

    so i guess he was for it before he was against it?

  15. Gravatar of Integral Integral
    20. November 2011 at 08:10

    For the record,

    “Inflation shock” can be interpreted as “supply shock.” An inflation shock happens to the Phillips Curve which in the NK framework is basically just an aggregate supply schedule. Any of the usual supply shock stories (high oil prices, bad crop, whatever are directly applicable to an inflation shock.

    Though strictly speaking it makes no sense to talk about a “shock” to an endogenous variable.

    Right. Taylor is looking for you to take the target rule:

    E[(NGDP(t+2) – NGDP*)^2] = 0

    where time is measured in years and NGDP* is the target, and turn it into an instrument rule:

    r = a + b*(y + pi)

    where y is real growth and p is inflation. Indeed Andy Harless does precisely this in his June 20, 2011 post! Though it’s only a heuristic argument, it is very close to what Taylor “wants”.

    Given a model of the private sector formulating an interest-rate rule is easy. You have an equation for how NGDP is determined that depends on the interest rate. Simply back out the appropriate rule and you’re done. See: Bean 83, Asako and Wagner 92, or Ball 99 for examples.

    In the real world things are more complicated and I don’t think you can commit to a hard interest-rate rule. But that is entirely the point. Instead of managing an interest-rate rule, you replace E(NGDP) with either the central bank’s forecast or the market’s forecast and adjust the interest rate / money supply / balance sheet / announcement of future policy until expected NGDP two years out equals the desired target level.

  16. Gravatar of Integral Integral
    20. November 2011 at 08:12

    Edit: “June 20, 2011” should read “May 13, 2011”. Pasted the wrong date.

  17. Gravatar of Taylor fires at NGDP targeting – the Market Monetarists fire back « The Market Monetarist Taylor fires at NGDP targeting – the Market Monetarists fire back « The Market Monetarist
    20. November 2011 at 10:22

    […] Scott Sumner Nick Rowe Marcus Nunes Bill Woolsey […]

  18. Gravatar of dtoh dtoh
    20. November 2011 at 18:42

    I think you have fundamentally missed the point that Taylor is trying to make. You say, “It [Taylor’s point] is a discussion of tactics; level versus growth rate targeting.” That is not his criticism. His criticism/question is not whether it is a level or growth rate target, but rather what are the instruments used to achieve that target? Elimination of IOR? Adjustment of the Fed Funds rate? Asset purchases? Manipulation of required asset/equity ratios (as I have repeatedly suggested)? Or simple manipulation of market expectations by mere Fed pronouncements.

    I agree with Taylor. The single biggest weakness in your argument is that you have not clearly specified this in an empirically convincing manner.

  19. Gravatar of Integral Integral
    20. November 2011 at 22:38

    This thread is rapidly being pushed off the first page — I hope Scott sees these comments.

    dtoh, I think Scott’s been pretty specific about his instruments. They are (1) announce a level target, (2) announce unlimited QE until you reach the target and (3) negative IOR.

    However, putting forward a convincing theoretical and empirical case still needs to be done. Empirically Scott’s strong suit is event studies and I think those can be used effectively to make the case for NGDP targeting. Theoretically his case is best made in an AD-AS framework but no-one’s actually made the theoretical case for NGDP targeting in a published paper for a while now. McCallum has but that was a decade ago. It’s a ripe area for research…

  20. Gravatar of dtoh dtoh
    21. November 2011 at 06:03

    Good comment, and I am aware that Scott has been pretty specific on the instruments…I just don’t think he has made an “empirically convincing” case.

    The naysayers will argue that when when rates are near zero, credit demand is slack, and the economy is not growing, that a change in the IOR rate will have no/or minimal impact on the amount of reserves held. Similarly, the naysayers will argue that QE will merely cause financial institutions to shift their holdings from Treasuries to reserves and have no impact on the money supply or velocity.

    Scott always responds to these arguments with a “a central bank can always inflate if it wants to” answer. He or someone needs to be much more specific in quantitatively demonstrating how the use of various policy instruments will translate into increased MV and how increased MV breaks down between increased RGDP and inflation.

  21. Gravatar of ssumner ssumner
    21. November 2011 at 12:54

    dwb, He doesn’t quite endorse a constant NGDP growth rate, but comes pretty close.

    Integral, I had thought that an inflation shock was a price shock, but when I read Taylor it seemed like that could not be what he meant, as he stated that an inflation shock would raise NGDP, and supply shocks don’t raise NGDP.

    So I assumed that the meaning of “inflation shock” must have changed since I was in school. I’d be very interested in knowing whether he actually meant a supply-shock, as that would make his argument somewhat incoherent.

    And don’t worry, I always come back to old comments. Your insights are very valuable, and highly appreciated here.

    dtoh, I’ve been very specific–use OMOs to peg the price of NGDP futures contracts. If you can’t do that then use OMOs to peg the Fed’s internal NGDP forecast. If you buy up the entire stock of world assets and are still below target, then start lowering IOR.

    As for “empirically convincing case”, I’d point to the fact that when NGDP grows at a stable 5% we do well, when it slows sharply we get bad recessions, and when it grows too fast we suffer from high inflation. Macro is a very imprecise science, I don’t see how we can do much better than that. I’d also point to many political arguments that I made in my National Affairs article. Political feasibility is also very important, as we are learning to our dismay in recent weeks.

  22. Gravatar of Dtoh Dtoh
    21. November 2011 at 14:09

    Making statements like “buy up the entire stock of world’s assets” is exactly the problem. Trust me on this. You need to say “if the fed buys 837.3 billion dollars of xyz assets, then NDGP will increase to 4.56%.”. It will frame the debate. Right now your detractors can just write you off by saying you’ve made an impractical suggestion. If you’re specific, they’ll come back and say “No it would require the Fed to buy 1.273 trillion in assets.” At that point you have hugely advanced the argument because you’re arguing about details not principle.

  23. Gravatar of Scott Sumner Scott Sumner
    22. November 2011 at 07:02

    Dtoh. I don’t think that’s a fair criticism, as it applies equally well to the Taylor Rule. The Taylor rule does not specify how many assets must be purchased. At near zero rates it could be a HUGE amount. So both policies have identical problems. But I don’t see people raising that objection regaridng the Taylor Rule.

  24. Gravatar of Dtoh Dtoh
    22. November 2011 at 08:41

    But nobody is trying to sell the Taylor Rule.

  25. Gravatar of ssumner ssumner
    24. November 2011 at 07:06

    Dtoh, But it’s the standard model! How can you criticize a proposal for having the same flaw as the standard model, if no one cares about that flaw in the standard model?

  26. Gravatar of Free Banking » Wide Off the Mark, or, Nonsense about NGDP Targeting Free Banking » Wide Off the Mark, or, Nonsense about NGDP Targeting
    1. December 2011 at 11:59

    […] is, I fear, one of which even some proponents of NGDP targeting occasionally appear guilty. Thus Scott Sumner himself, in responding to Taylor, observes that "the dual mandate embedded in NGDP targeting is not that […]

  27. Gravatar of TheMoneyIllusion » George Selgin on John Taylor and the dual mandate TheMoneyIllusion » George Selgin on John Taylor and the dual mandate
    1. December 2011 at 16:52

    […] I did a post a while back that responded to John Taylor’s criticism of NGDP targeting.  In the comment section Statsguy pointed out that Taylor’s argument was even more flawed than I assumed.  And now George Selgin shows that it was even more flawed than Statsguy assumed: Thus Professor Taylor complains that, “if an inflation shock takes the price level and thus NGDP above the target NGDP path, then the Fed will have to take sharp tightening action which would cause real GDP to fall much more than with inflation targetting.” Now, first of all, while it is apparently sound “Economics One” to begin a chain of reasoning by imagining an “inflation shock,” it is crappy Economics 101 (or pick your own preferred intro class number), because a (positive) P or inflation “shock” must itself be the consequence of an underlying “shock” to either the demand for or the supply of goods. The implications of the “inflation shock” will differ, moreover, according to its underlying cause. If an adverse supply shock is to blame, then the positive “inflation shock” has as its counterpart a negative output shock. If, on the other hand, the “inflation shock” is caused by an increase in aggregate demand, then it will tend to involve an increase in real output. Try it by sketching AS and AD schedules on a cocktail napkin, and you will see what I mean. […]

  28. Gravatar of Selgin Takes Down Taylor on NGDP Targeting « Uneasy Money Selgin Takes Down Taylor on NGDP Targeting « Uneasy Money
    3. December 2011 at 17:46

    […] reply Scott Sumner wrote a good defense of NGDP targeting, focusing mainly on the forward-looking orientation of NGDP targeting in contrast […]

  29. Gravatar of Wide Off the Mark, or, Nonsense about NGDP Targeting – Alt-M Wide Off the Mark, or, Nonsense about NGDP Targeting - Alt-M
    29. March 2015 at 13:29

    […] is, I fear, one of which even some proponents of NGDP targeting occasionally appear guilty. Thus Scott Sumner himself, in responding to Taylor, observes that “the dual mandate embedded in NGDP targeting is not […]

  30. Gravatar of Wide Off the Mark, or, Nonsense about NGDP Targeting – Alt-M Wide Off the Mark, or, Nonsense about NGDP Targeting - Alt-M
    22. December 2015 at 06:32

    […] is, I fear, one of which even some proponents of NGDP targeting occasionally appear guilty. Thus Scott Sumner himself, in responding to Taylor, observes that "the dual mandate embedded in NGDP targeting is not that […]

Leave a Reply