Reply to Smith and Rognlie

Karl Smith recently quoted me and then responded:

Scott writes

“The new GDP figures offer a reminder that one can’t analyze movements in GDP by looking at components of GDP.  We saw huge increases in spending on cars (pushing consumer durables up by 15.3%) and houses (up 19.1%.)  And yet overall RGDP growth fell to only 2.2%.  Why didn’t the spending on cars and houses help?  Because in the short run it’s NGDP growth that drives RGDP.  And the Fed continued its tight money policy by allowing only 3.8% NGDP growth, same as the previous quarter.”

If the Fed had a strict NGDP target and everyone knew what it was the Scott would be correct. However, if either the Fed doesn’t have a strict target or we don’t know what it is then Cars and Houses are causal or informative, respectively.

This is because Cars and Houses pull harder on NGDP more than most sectors of the economy. If we think of the Fed trying to influence NGDP indirectly – through the interest rate or some other mechanism – then this would imply that an exogenous negative shock to the purchase of cars and houses would tend to lower the demand side pressure on NGDP and so the Fed would need to loosen its indirect target to keep NGDP growth constant. Conversely and exogenous positive shock would raise demand side pressure on NGDP and the Fed would need to tighten its indirect target to keep NGDP growth constant.

In the comment section Matt Rognlie agreed:

Indeed. I love Scott Sumner, but sometimes the fixation on NGDP goes overboard. It’s true that an NGDP level target might be a very good rule for the Fed to follow, but unfortunately we’re not operating under an NGDP level target or anything close to it. In practice, we’re close to a barely-flexible inflation targeting framework.* And in that environment, the dynamics of consumer spending matter a great deal.

I love Matt and Karl too, but I don’t quite agree with this.  I actually wasn’t assuming the Fed was targeting NGDP, although I can see how they would have made that assumption.  Rather I assumed that Fed policy determined NGDP, if only through errors of omission.

Now I suppose you could make an argument that car and house output would have an impact on NGDP, given the Fed’s operating procedure.  But I think that very unlikely.  For instance suppose they target inflation, which is probably not all that far from current policy.  In that case the Fed shifts AD to offset any change in AS, and hence car and house output only boosts NGDP if it shifts AS to the right.  That could happen, but it seems unlikely to me.

Suppose we make an even weaker assumption, such as money supply targeting.  Now it’s much easier to tell a story where car and house output boosts NGDP.  These products are usually bought on credit, as they are essentially investment goods.  So more demand for these products tends to raise interest rates, and also velocity.  That would boost NGDP under money supply targeting (or interest rate pegging–via a bigger money supply.)

But even then I’d defend the broader point I was trying to make here.  In Q1 we saw big increases in cars and houses, and yet very little growth.  So either the effect didn’t occur at all, or these sectors prevented what would have otherwise been a big fall in NGDP.  In either case, I don’t see where those sectors help us to predict where we go next.

On this point I’m quite prepared to backtrack, if the evidence cuts the other way.  We have a thriving industry in “leading indicators.”  I’m pretty sure that stocks and the yield spread have been shown to have significant predictive power.  I doubt the various sectors of GDP have much power, once you account for overall growth in GDP.  Thus I doubt that a prediction of future growth based on current growth in GDP can be substantially improved by looking at output by sector, with the possible exception of inventories.  But the evidence is out there.  If sectors are predictive, then I’m sure we would have discovered it, and they’d be in leading indicator models.

This link suggests that housing doesn’t predict GDP, but building permits do.  That makes sense to me, as I’d expect growth in building permits to be correlated with future growth in construction.

PS.  Yes, I have an unhealthy obsession with NGDP.  But our profession has the exact opposite problem, and my goal is to fix that oversight.



12 Responses to “Reply to Smith and Rognlie”

  1. Gravatar of Benjamin Cole Benjamin Cole
    4. May 2012 at 07:47

    “PS. Yes, I have an unhealthy obsession with NGDP. But our profession has the exact opposite problem, and my goal is to fix that oversight.”—-Sumner.

    Verily, I too am fixated on generating real growth and jobs by the best means.

    Add on: We have “economists” who say we are not like Japan as they had deflation (of less than one percent a year) and the USA has inflation (running now at record law rates, and well under two percent).

    So, if the USA sags just like Japan, but our prices expand by one or two percent more than Japan’s, then we must fight inflation, and worship price stability before real economic growth?

    This is where the bulk of the economics profession is today? How has such a peevish fixation on price stability become so deeply ingrained in otherwise reasonable people?

  2. Gravatar of Benjamin Cole Benjamin Cole
    4. May 2012 at 07:55

    All-time low inflation expectations, reports the Cleveland Fed.

    If you put a monetary noose around the neck of an economy, yes, you can asphyxiate inflation expectations. Of course, the economy will turn blue along the way….

  3. Gravatar of John hall John hall
    4. May 2012 at 08:40

    This is a very interesting statement: “I doubt the various sectors of GDP have much power, once you account for overall growth in GDP. Thus I doubt that a prediction of future growth based on current growth in GDP can be substantially improved by looking at output by sector, with the possible exception of inventories.”

    A few points:
    1) The problem with inventories is that it is not expressed in the same terms as the other GDP variables. It is expressed as a change, rather than a level. When I try to predict inventories, I convert it to levels first so that it is more comparable with the other components of GDP.
    2) So let’s say you have the levels for each of the components of final sales + the levels for inventories + headline GDP. Perform a PCA on the covariance matrix of the log changes changes of all of these. The first factor will likely be strongly correlated with GDP or productivity growth. In this sense, you would be right that broad growth in GDP explains a large amount of the variation within the sectors.
    3) The question, however, is how much of the variation is explained by this first factor. Is it so much that all the others can be thought of as noise? It could be that the first PC is a random walk, but others are more mean-reverting? Or it could be that the second PC is even more driven by a regime-switching process than the first PC. However, I think that ignoring the other top PCs would still ignore important information.

  4. Gravatar of Benjamin Cole Benjamin Cole
    4. May 2012 at 09:06


    So, if the USA sags just like Japan.

    Japan isn’t sagging.

    Japan has consistently outperformed many countries that have had “looser” monetary regimes.

    You’re looking at the wrong data.

  5. Gravatar of Luis Pedro Coelho Luis Pedro Coelho
    4. May 2012 at 10:17

    Certain components of (R|N)GDP might be better forward indicators even without any causal relationship to overall (R|N)GDP.

    Just as “consumer confidence” surveys aggregate overall predictions about the economy, maybe car purchases (taking on a new or increased monthly payment) equally correlate with brighter future prospects.

  6. Gravatar of Luis Pedro Coelho Luis Pedro Coelho
    4. May 2012 at 10:18

    (I’m not making an argument that this is empirically so, just that it is logically possible).

  7. Gravatar of Bill Woolsey Bill Woolsey
    4. May 2012 at 10:44

    Suppose that we have interest rate targeting.

    The quantity of money varies to keep interest rates fixed.

    Perhaps the best way to find out what nominal GDP will be is to add up the spending on each and every thing.

    Suppose you have inflation targeting, but no one spending or producing pays any attention.

    You are the Fed. How do you figure out what nominal GDP will be given the current level of interest rates?

    Could it be that adding up all the different types of spending is the way to go?

    Then, if it is too big or too small, you adjust interest rates.

    If you don’t fix the quantity of money, but adjust it, then what is happening to velocity? Do you figure that out by adding up spending on all the different things? If it is too high or too low, you adjust the quantity of money.

    Now suppose those buying and selling (including you and me,) are paying attention. If total spending is adding up to something that is too high, then we know the Fed will contract, and so, we better reign in our spending. But how do I figure out that total spending is adding up to something too high? Don’t I have to add it up?

    And most importantly, suppose you are speculationg in index futures contracts. Is nominal GDP going to be above or below target in one year? How do you know? Isn’t adding up the parts something that you would need to be doing?

    What is the perspective that we are taking when we say that it will all add up to the targeted amount? Or to a level such that the inflation rate is on target?

    It seems to me someone is going to have to observe that spending in this sector is growing extra fast. Where are the slow growing or shrinking sectors? Oh, there. Or nowhere, we are going to be above target. Buy futures now!!!

    Housing is skyrocketing. Well, the domestic toy and underwear producers are in the dumps. Total spending will grow on target. Stay out of the NGDP index futures….

  8. Gravatar of Peter N Peter N
    4. May 2012 at 12:40

    It’s not surprising that making predictions from our economic statistics is difficult. They’re a mess. Congress has been starving the agencies that produce them for years. Large segments of the economy are effectively invisible. Corrections can be enormous (which had serious consequences in 2008).

    Now the House is proposing cutting their funding. This is crazy. Economists have to speak up on this issue, or they’ll have no current economic data to use.

    There’s also a somewhat related issue – GDP. It’s not clear how useful a target it is, considering it was never designed to be such a thing. Some problems are

    The high level of imputations some of them quite dubious

    The valuation of imports

    Tracking outsourcing

    Definition and calculation of intermediate production

    Omission and inconsistent treatment of non-production expenditure

    Poor tracking of certain sectors (like health care)

    Of course calculations of second order statistics, like productivity are even worse.

    It would be better to decide what you want to track based on whatever you feel are the appropriate principles and calculate an Economic Activity Indicator. You’d start (as many economists do now in publications) with a corrected GDP and GDI and then try to acquire the missing data.

    If enough economists could agree on the appropriate principles, this could be a very worthwhile program.

  9. Gravatar of Major_Freedom Major_Freedom
    4. May 2012 at 13:17

    You are the Fed. How do you figure out what nominal GDP will be given the current level of interest rates?

    First, I’d gaze into my crystal ball to find out what NGDP should be. I mean that’s how it’s done now anyway.

    Oh sorry, what I meant to say is that a bright and educated group of economists sat together, and spoke gobbledygook with each other, masking it behind a series of sophisticated and complex mathematical models give the gobbledygook a veneer of legitimacy, and then they did a show of hands for various numbers that sound nice, like 5%. The number with the most hands wins.

    This is how the intricate laws of the universe are revealed to the non-economists parading around as economists due to their intellectual financing coming from a violence backed institution. It is much better than doing real science based on private investment.

  10. Gravatar of ssumner ssumner
    4. May 2012 at 14:23

    Ben, I’ve never understood their technique, but the results are certainly sobering.

    John Hall, I don’t know much about this issue–I’ll be interested in what others have to say.

    Luis, That’s possible, but I’m skeptical.

    Bill, With index futures we certainly wouldn’t need to look at anything else.

    Peter N, I’d be thrilled if Congress cut off the funds for measuring inflation.

  11. Gravatar of Peter N Peter N
    4. May 2012 at 19:03

    “I’d be thrilled if Congress cut off the funds for measuring inflation.”

    That would mean cost of living adjustments would be a complete political football. Do we really need to borrow trouble, when we already have so much?

    We already don’t know whether the right number for past productivity is %2.8 or %0.8. How close GDP is to any measure of actual production is anybody’s guess, but an error of 20% or so wouldn’t surprise me a bit.

    Yes, the GDP deflator is a mess, but so are our import and export numbers. Then there’s over $3 trillion of imputed economic activity, problems of classification…

    And you want to target this mess? What real economic variable do you think you’re actually targeting?

    Oh yes. There’s also the little matter of corrections. Do you want to target something where they come back a year later and say “You know that %4 figure we gave you? Well we’re revising it to %1.2”.

    At what revision point are your futures going to pay off. Are the proceeds going to be held in escrow 2 years until there are final numbers.

    And, of course, there are the seasonal corrections.

    Yes inflation numbers are too high, but there’s no budget to fix them, and Congress can’t even decide to use the best current ones, even though it would save $100s of billions.

    It’s the political will that’s the real problem.

  12. Gravatar of ssumner ssumner
    6. May 2012 at 08:12

    Peter N, Just index them to average wages, makes much more sense.

    I’m not concerned about measurement errors with NGDP. Yes they exist, but if we stabilized expected NGDP growth then we’d get a much more stable rate of actual growth. The errors in growth rates are mostly unforecastable, and hence wouldn’t distort NGDP futures targeting.

    There isn’t even any theoretical way to measure inflation, as economists have never even defined the term. That’s because they’ve never defined a “good.” So spending more money won’t “fix” the problem.

Leave a Reply