It’s the NGDP, stupid

Housing starts February 2008 = 1.10 million.  Unemployment = 4.9%

Housing starts November 2013 = 1.09 million.  Unemployment  = 7.0%

It’s not about housing, and never was.

PS.  Alex Tabarrok has a very good post on Robert Shiller’s proposal for “trills,” which would be Treasury bonds with interest payments indexed to (one-trillionth of) nominal GDP.  He sort of skims over the most important use, however.  The market price of the interest payments (sold as a derivative) would be a market forecast of NGDP, available in real time, and hence could be pegged by the Fed.

I proposed the creation of a nominal GDP futures market in a paper published way back in 1989. Tabarrok concludes with the following observation:

I featured Shiller’s work on macro markets in my book Entrepreneurial Economics: Bright Ideas from the Dismal Science. I think of this body of work as his most visionary and deserving of the Nobel.

Naturally I agree.



27 Responses to “It’s the NGDP, stupid”

  1. Gravatar of foosion foosion
    19. December 2013 at 10:12

    I see two major claims about housing, neither of which are addressed by showing the same number of starts in 2008 and today.

    1) A construction bubble which peaked around 2005 which resulted in a large decrease in spending when it burst.

    2) An approximately $8 trillion increase in housing equity resulted in a large decrease in spending when it burst.

    Together, around 6% of GDP. At today’s GDP, that’s around a trillion dollars.

    For example,–should-we-be-worried–by-dean-baker

  2. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. December 2013 at 10:27

    Off Topic.

    John Carney has a reply to your recent conversation:

    December 18, 2013

    Teaching the market monetarists about money
    By John Carney

    “In a recent post I proposed that Scott Sumner, the premier market monetarist, expects too much of an inflationary effect from quantitative easing because his definition of money is too narrow.

    Sumner appears””I say “appears” because, like a lot of people, I’m never quite sure what Sumner’s saying””to think that when thinking about inflation and QE, “base money” provides us with the most useful metric. The term “base money” was coined by Karl Brunner in his 1961 article, “A Schema for the Supply Theory of Money.” Exactly what should count as “base money” was immediately subject to all the usual two-handed economist revisions and challenges. It also got called a lot of different names: “outside money,” “high-powered money,” “non-interest bearing government debt.” But, roughly speaking, the concept is that base money is the total of currency in circulation plus reserve balances of banks.

    I don’t doubt that at certain times and in certain circumstances, this is a very useful definition of money. Or was. It certainly seems to have a lot of explanatory power when looking at the monetary policy of the Great Depression.

    But times have changed. Government bonds now do a lot of the work that currency and reserves once did. They are used to collateralize loans that fund the securities portfolios of broker-dealers, for instance. One recent study by the Fed showed the Treasury and Agency debt accounted for 85 percent of the collateral used in the Tri-Party Repo market, one of the least understood and perhaps most important components of the U.S. financial system. (The Fed study linked here provides a good explainer.)…”–ric/Archivio-25/Copeland_et-al_prox_FRBNY-Econ-Policy-Rev.pdf

    I’ve been doing some research on the tri-party repo market and will say some more on this in a follow up comment.

    P.S. John Carney’s post was linked to at Pragmatic Capitalism (“Government Bonds in a World of Moneyness”) and Mike Norman Economics (“John Carney “” Teaching the market monetarists about money”).

  3. Gravatar of Doug M Doug M
    19. December 2013 at 10:32

    Argentina sold GDP linked warrents when it came out of bankruptcy a several years ago. I don’t remember the exact structure, but I think it payed a fixed payment every quarter that GDP growth was above some level.

    I don’t know of any other GDP linked securities currently out there.

  4. Gravatar of TravisV TravisV
    19. December 2013 at 10:32

    John Taylor video below:

    Question: Does Taylor actually believe what he’s saying?

    Couldn’t he be intentionally spouting nonsense in order to appeal to elite Republicans such as Paul Ryan and Scott Walker?

    Put yourself in Taylor’s shoes. He sees Mankiw praise Yellen and Taylor thinks “THIS IS MY OPPORTUNITY to outflank Mankiw (influence among elite Republicans)!!!”

  5. Gravatar of MJuliusKing MJuliusKing
    19. December 2013 at 10:53

    In fairness, one could make the argument that “housing starts” is not the same metric as “construction jobs”:

    Indeed, roughly 2 million jobs in construction were lost from peak to trough. However, not all were related to housing, and many were only after the decline in NGDP growth as well. And the total job loss from peak to trough seems to be around 8.8 million:

    I guess then a “housing theorist” could say that one would expect construction job numbers to lag housing starts (due to early maintenance, upgrades, time when a new house is counted as a new house, etc.)

  6. Gravatar of MJuliusKing MJuliusKing
    19. December 2013 at 11:00

    Actually, it does seem that construction job numbers lag housing starts by about the amount one would expect from the previous graph:

    So now I’m not sure. It looks like the decline in employment related to the housing market issues may be more important than suggested by a comparison of housing starts and unemployment.

  7. Gravatar of jknarr jknarr
    19. December 2013 at 11:02

    Teaching?!! Wow, quite the converse. Accuse others of what you do yourself. These guys work hard to obfuscate, promote, and blur the definition of money as far as I can tell: not to clarify the topic. Such an open definition of money that everything falls out.

    As far as I can tell, the MM position is that there is only one legal tender, which is base money, and the means for all MOA. It’s the Fed’s entire superpower! All other financial contracts (that they call money) are liquidity-demanding, for heaven’s sake. Treasuries are liquid, but they are still liquidity demanding.

    They are collateral for leveraged positions, you say? Just because a hugely-liquidity-demanding risk asset positions require less-liquidity-demanding collateral does not make either any less liquidity demanding! The ultimate liquidity is base money, because it is legal tender. If banks need collateral, get hold of base money reserves! (Oh, yeah, they did and still do….)

    When you demand liquidity, you are putting demands on other entities’ balance sheets. When leverage is abundant and balance sheets are kind of crappy, then guess what? You can’t demand quite so much liquidity out of your fixed income asset/liabilities, and this includes “safe assets”.

    Bonds are leverage, which, however liquid, are simply promises to pay base money in the future: debt now, base money later.

  8. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. December 2013 at 11:56

    Off Topic.

    I accidentally left out a chunk of the article where Carney explains how monetarists view QE above. No matter, read the article.

    Here’s my thoughts on the Tri-Party Repo Market.

    First of all, a useful link is this one:

    From there you can link to historical data on the types and amounts of securities involved all the way back to May 2010.

    John Carney’s central point seems to be that QE removes eligible repo collateral and replaces it with reserves. Since he seems to be arguing reserves and bonds are close substitutes QE does little to increase the money supply.

    First of all let’s look at the collateral side of this issue. As Carney points out Agencies and Treasuries make up about 85% of the securities used in tri-party repos.

    How large is the Agency and Treasury component of the collateral side of the tri-party repo market?

    The data since May 2010 shows that it has varies between $1.33 trillion (September 2013) and $1.7 trillion (November and December 2012). To put this in perspective, according to the level tables of the flow of funds (L.209 and L.210) there were $19.7 trillion in Agencies and Treasuries outstanding at the end of 2013Q3, of which the Fed held $3.5 trillion leaving $16.2 trillion avilable for the tri-party repo market. So over the period since May 2010 the proportion of such securities used in tri-party repos has varied between 8.3% (September 2013) and 10.4% (November and December 2012). Moreover the peak in Agency and Treasury security use in tri-party repos in late 2012 is an anomaly. On the whole, the size of the market has been roughly stable at $1.4 trillion while the amount of outstanding Agencies and Treasuries not held from the Fed has grown from $14.2 trillion in 2010Q2 to $16.2 trillion in 2013Q3.

    How large is the cash side of the tri-party repo market?

    From what I can figure out “cash” in the tri-party repo market generally refers to zero-maturity money (i.e. MZM).(If anyone thinks this is totally or partially wrong, please correct me.) MZM has grown from $9.4 trillion in May 2010 to $12.2 trillion in November 2013. The proportion used in tri-party repos has varied between 11.0% (September 2013) and 15.4% (September 2010) since May 2010. Aside from the anomalous blip in late 2012 the trend is downward as the amount of cash has been more or less steady at about $1.4 trillion and MZM is growing at nearly 8% per year.

    Before moving on, also note that the Agency and Treasury component of the tri-party repo market is larger in proportion to MZM than it is to the available collateral. So MZM is more of a constraint than the amount of available Agencies and Treasuries.

    We know precisely the effect of QE on the amount of available Agencies and Treasuries, but what has its effect on the amount of available cash been?

    Well, by definition QE increases the monetary base, or the amount of currency and reserve balances. What is less well known perhaps is that the vast majority, say about 95%, of security sellers are non-banks meaning the vast majority of QE open market operations end up as increased MZM. Since the Fed’s holdings of Agency and Treasury securities has increased by $1.4 trillion between 2010Q2 and 2013Q3 that means perhaps $1.33 trillion of the $2.62 trillion increase in MZM during that time is directly attributable to QE, or over half. That also means thanks to this effect alone QE has increased MZM by about 14% during that period.

    In contrast, if QE had not occured, there would be $17.6 trillion in available Agencies and Treasuries instead of $16.2 trillion which is about 8% less.

    So based on this analysis alone it seems that QE is a net plus to the tri-party repo market. QE has directly increased the available cash by nearly double the proportion than it has reduced the available Agencies and Treasuries.

    I haven’t even considered the fact that QE has probably catalyzed the creation of MZM by raising NGDP. (My own econometric results show that the monetary base Granger causes commercial bank loans and leases over the period since December 2008 at the 5% significance level.)

    And, perhaps more importantly, I haven’t even brought up the fact that a tri-party repo is a profit-making opportunity for the cash side of the transaction.

    So it sounds to me like Agencies and Treasuries are useful in tri-party repos predominantly for their ability to get loans of cash, which in turn means their degree of “moneyness” is a lot lower than just plain old cash.

  9. Gravatar of The Market Fiscalist The Market Fiscalist
    19. December 2013 at 11:57

    Off Topic question:

    Assume that via monetary policy/fiscal policy (or a combination of the 2) its is always possible to hit an NGDP target.

    How do we know that the targeted level of NGDP will generate the optimal level of RGDP ? At any given level of NGDP increase there will be many factors that will determine the RGDP/inflation outcome If you’re going to use futures tracking why not ask people what level of RGDP they expect in 12 months time and then adjust the NGDP target to keep RGDP expectations at optimal levels ?

  10. Gravatar of ssumner ssumner
    19. December 2013 at 11:58

    foosion. Note that GDP kept rising even as housing construction fell sharply (and btw, housing completions tell roughly the same story. So the housing collapse did not sharply impact GDP.)

    Mark, Thanks, I’ll do a post.

    MJuliusKing, Bob Murphy and I did this debate a while back. The construction jobs series is not the right one, as nonresidential construction crashed much later than housing construction. That makes the delay look much bigger than it really is, and makes the job losses in housing construction seem much bigger than they really are.

    Not saying your job loss lag point has no validity, but with the proper data you;d see that only a small share of total jobs lost were due to the decline in residential construction up through early 2008, even accounting for lags. Most were due to later drops in construction, and nonresidential construction, and manufacturing and services.

    Of course when NGDP fell that pushed housing construction even lower, creating even more job losses. But that’s NGDP, housing was endogenous after mid-2008.

  11. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. December 2013 at 12:05

    One last point on the tri-party repo market.

    We know empirically that QE raises longer term interest rates. (My own econometric anlysis shows that the monetary base Granger causes 10-year T-Note yields over the period since December 2008.) This effectively means it reduces the scarcity of Agency and Treasury collateral, which is the opposite of what John Carney is argueing.

  12. Gravatar of ssumner ssumner
    19. December 2013 at 12:05

    Thanks Mark.

    Fiscalist. We don’t know.

  13. Gravatar of benjamin cole benjamin cole
    19. December 2013 at 12:11

    Excellent blogging.
    Note to Travis: John Taylor advised the Bank of Japan to go with QE in 2002 and then gushed about the positive results–he published paper on this in 2006.
    It is inexplicable–as economics. As social norms change…
    Today PhD economists rhapsodize about zero inflation and a “dove” is a FOMC board member who would “tolerate” a 2 percent inflation rate. The profession has become peevishly fixated on inflation. Forgotten today is that when inflation-fighting king Volcker left office inflation was 4 to 5 percent and people were happy with that.
    To obsess with a nominal price index of dubious accuracy is not good macroeconomic policymaking.
    I attribute this change in thinking to the rise of powerful independent central banks in curious alliance with zealous partisans in the right-wing who often suffer from auriferous fever.
    I would prefer the profession become obsessed with real growth.

  14. Gravatar of MJuliusKing MJuliusKing
    19. December 2013 at 12:23

    “The construction jobs series is not the right one, as nonresidential construction crashed much later than housing construction. That makes the delay look much bigger than it really is, and makes the job losses in housing construction seem much bigger than they really are.”

    Ah, I see now. Weird that construction employment went up so much before, but that explanation makes sense.

  15. Gravatar of Sina Motamedi Sina Motamedi
    19. December 2013 at 12:58

    Patience, Scotty.

  16. Gravatar of Andy Harless Andy Harless
    19. December 2013 at 13:55

    Nick Rowe also wrote a post about trills last year (around the same time as Shiller, it appears, but AFAIK neither references the other).

  17. Gravatar of TravisV TravisV
    19. December 2013 at 15:33

    benjamin cole,

    Thanks for the reply. To be fair, a lot of the right-wing inflation hatred is due to the fact that Obama is president. Things would look a lot different if the President were a Republican.

  18. Gravatar of Geoff Geoff
    19. December 2013 at 15:37

    There are more than one reason for the rise in unemployment. Housing is one of them:

    Another, more important reason for the rise in unemployment is the fall in saving and investment, which was not caused by falling NGDP, but rather was the proximate cause for falling NGDP.

  19. Gravatar of ssumner ssumner
    19. December 2013 at 19:49

    Ben, Yes, 2% is very dovish today—almost radical.

    Sina, Maybe next year.

    Thanks Andy, I’ll take a look.

  20. Gravatar of ssumner ssumner
    19. December 2013 at 19:52

    Andy, If we compared the yield on trills and regular bonds, we could figure out if you are right about expected NGDP growth, right?

  21. Gravatar of Geoff Geoff
    19. December 2013 at 21:06

    “I proposed the creation of a nominal GDP futures market”

    NGDP futures would only be worth the arbitrary discretionary rate of interest the Fed will pay. An investor cannot take hold of NGDP, so any derivative based on NGDP would be worthless.

  22. Gravatar of Andy Harless Andy Harless
    20. December 2013 at 09:05


    We could certainly make a more educated guess as to whether I’m right, but there’s still an issue of differences in risk premia (and also differences in maturity, but maybe 30 years is close enough to perpetual). I guess we would expect trills to have a higher risk premium than long duration nominal bonds (though I’m not sure, because it depends on the expected correlation between inflation and real growth). But given a reasonable range of risk premia, trill pricing would presumably allow one to rule out some ranges of expected NGDP growth.

  23. Gravatar of Andy Harless Andy Harless
    20. December 2013 at 09:31

    Scott (and Nick, if you happen to read this),

    Another issue that occurs to me is heterogeneity. I’ve been willing to assume that we could make the NGDP futures market deep enough that the marginal participant will have close to typical expectations. But thinking about Nick’s trill thought experiment, he effectively assumes homogenous expectations (or equivalently, that the market would be deep enough to get marginal participants with typical expectations). But then he talks about issuing a single trill. I doubt offhand that a single trill would produce a deep enough market. Rather, the trill would likely be purchased by investors with atypically high growth expectations and priced accordingly. Obviously there is some number of trills that would produce a deep enough market, but it’s not obvious how deep the market would have to be.

  24. Gravatar of Andy Harless Andy Harless
    20. December 2013 at 09:36

    OK, wait. Now that I think about it, the issue of depth is not so much of a problem for NGDP futures, anyhow, because the private sector would typically be taking both sides of the trade. (Obviously depth matters for liquidity, but heterogeneity is not a big problem.) But with trills, the government is, by design, taking one side of the trade, and doing so (by Nick’s construction, anyhow) without even the intent of maximizing profits, so you need a much deeper market to get an accurate signal of market expectations.

  25. Gravatar of Nick Rowe Nick Rowe
    20. December 2013 at 12:39

    Andy: a quick thought: you might only need one trill perpetuity, but broken up into a thousand pieces, to get the liquidity. Near-infinity divided by one thousand is still a big number.

    Shiller was way ahead of me.

  26. Gravatar of Andy Harless Andy Harless
    20. December 2013 at 14:51


    But liquidity isn’t the issue I’m concerned about with trills. I’m concerned about heterogeneity. Even if the market is perfectly liquid, the marginal participant may be far from typical. A single trill, even if it’s broken up into thousands of pieces, is only going to be owned by people who are unusually optimistic about the growth rate. They’ll outbid everyone else, and the market price won’t accurately reflect the expectations of typical investors. I still have to think through the implications for the argument in your old blog post. But at least to Scott’s point, this does mean that the trill — unless there are a whole lot of them — won’t work as a gauge of typical growth expectations.

  27. Gravatar of ssumner ssumner
    20. December 2013 at 15:27

    Andy, That’s one reason I prefer NGDP futures to NGDP indexed bonds.

Leave a Reply