Robert Gordon on reasoning from a price change

One common theme that runs through both my blogging and my academic research is that economists need to avoid “reasoning from a price change”.

Matt Yglesias directed me to an old Robert Gordon paper that criticizes modern New Keynesian DSGE models. Gordon also emphasizes the need to avoid reasoning from a price change:

An explanation was needed to reconcile the dominant role of demand shocks as the explanation of the Great Contraction of 1929‐33 in the same model as would explain the positive correlation of inflation and unemployment in 1974‐75. Once recognized, that explanation became obvious. Just as the output and price of corn or wheat could be positively or negatively correlated depending on the importance of micro demand or supply shocks, so aggregate output and the rate of inflation could be positively or negatively correlated, depending on the relative importance of aggregate demand or supply shocks.

Gordon worries that a failure to distinguish between different types of inflation has handicapped modern DSGE models:

The supply side of the toy model and of most DSGE models is likewise handicapped by including in the inflation equation only expectations of future inflation and the current value of the output or employment gap (some versions substitute marginal cost, but this does not help, see Gordon 2009 and the references cited there). The key elements of the 1978‐era demand‐supply synthesis missing from modern macro are the basic distinction between expectation formation and the sources of backward‐looking behavior through staggered overlapping contracts, and the absence of any treatment of supply shocks. The supply side of modern DSGE models contains no element that explains why inflation and unemployment are sometimes negatively and sometimes positively correlated, which is not surprising since efforts to simulate dynamic responses in such models have been limited thus far to monetary policy shocks, not supply shocks.

If you have an inflation model that relies on output gaps, then you are implicitly treating all inflation as demand-side. Note that Gordon wants to go back to 1978-era AS/AD models, and so do I.

Here he provides a bit more detailed critique:

Likewise there are three problems with the aggregate supply equation based in the NKPC. First, the reliance on purely forward‐looking expectations is contradicted both by facts and by the underlying logic of the 1978‐era Phillips Curve with the intrinsic backward‐looking behavior based on staggered wage and price contracts and a web of producer‐supplier relations extending back into the input‐output table. Second, the NKPC equation contains no terms to represent the impact of supply shocks directly on the inflation rate and indirectly on real output, depending on the nature of the policy response. Thus the NKPC approach has no explanation of why the inflation rate is sometimes negatively and sometimes positively correlated with the unemployment rate, a fact evident in Figures 1 and 2 above. Third, the NKPC does not recognize that prices are both flexible and sticky at the same time; the basis of the Gordon‐Phelps supply‐shock models discussed above is to combine flexible auction‐market prices in the shocked oil market with sticky prices in the remaining non‐oil sector of the economy. The implied macroeconomic externality when adverse supply shocks cause recessions is particularly missing from the NKPC, from the toy model, and apparently from the rest of modern business‐cycle macroeconomics.

I’m no expert on NK models, so please tell me if I’m missing something.



23 Responses to “Robert Gordon on reasoning from a price change”

  1. Gravatar of BC BC
    20. May 2021 at 23:09

    Possibly related question. For most of the period following the financial crisis, stocks and treasuries were negatively correlated. Hence, stocks and inflation were positively correlated. Over the last few months, stocks and treasuries were positively correlated. (Stocks and inflation negatively correlated.) Does that indicate that the biggest macroeconomic risks right now are on the AS side rather than AD side?

    During normal AD recessions and recoveries, monetary stimulus increases inflation and NGDP expectations. Stock prices rise with expectations of a recovery but treasury prices fall with higher yields. If supply side shocks are dominant, however, then a negative supply shock increases inflation and decreases output. Stock prices fall along with treasury prices.

  2. Gravatar of Asher Asher
    21. May 2021 at 01:30

    There are two distinct issues involved. 1. How important are expectations, compared to “backward looking behavior” (the post gives the example of staggered contracts)? 2. How much are expectations themselves based on modeling the future, compared to extrapolating from the past? The old models gave a lot of weight to the past and present in both ways. They considered current prices and agreements to be an anchor, and they considered them to be a central factor in forming expectations. Often they considered some naive extrapolation from past experience to determine expectations, and that was a mistake. But the “modeling the future” paradigm, the extreme version of the rational expectations approach, goes much to far in the opposite direction.

  3. Gravatar of Michael Rulle Michael Rulle
    21. May 2021 at 04:53

    While there seems to be a general consensus on bad Fed Policy in the 30s, it never occurred to me there is no general consensus on the cause of inflation in the 70s. Not being an economist, I do find it interesting that something as important—even “basic” (says non-economist) as different types of inflation has not been resolved. It is obviously more difficult in real time, than 30-90 years after the fact. Which, of course, makes one concerned about today———where the dominant headlines seem to be that the Fed is underestimating the likelihood of much higher inflation. I still do not think the current Fed has been shown to be wrong——if anything his predictions and goals still seem consistent——-but I do guess the next 3-6 months will be critical.

  4. Gravatar of Howard W. Campbell Howard W. Campbell
    21. May 2021 at 05:11

    Did an obscure TV show from decades ago predict China’s plan against America?

    My cousin sent me the link, and the resemblance to our days is uncannily eery to say the least.

  5. Gravatar of Todd Ramsey Todd Ramsey
    21. May 2021 at 06:16

    Off-topic: would love to read Scott’s commentary on the May 19 release of the April Fed minutes and the subsequent decline in TIPS spreads.

    Seems to be as Scott predicted, that the Fed could apply the brakes easily when it wanted to.

  6. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    21. May 2021 at 07:49

    The evidence of inflation, contrary to the conventional wisdom, cannot be conclusively deduced from the monthly changes in the various specialized price indices. The price indices are passive indicators: for the average change; of a group of prices. They do not reveal why prices rise or fall.

    I.e., inflation targeting depends on how inflation’s defined (which somehow neglects to encompass the vast proportion of all past and present asset-bubbles).

    Only price increases generated by demand, irrespective of changes in supply, provide evidence of monetary inflation. There must be an increase in aggregate monetary purchasing power, AD, which can come about only as a consequence of an increase in the volume and/or transactions’ velocity of money.

    The volume of domestic money flows must expand sufficiently to push prices up, irrespective of the volume of financial transactions consummated, the exchange value of the U.S. dollar (per the Nattering Naybob: “reflected in FX indices and currency pairs”), and the flow of goods and services into the market economy.

    Money and money flows may be net robust, net neutral, or net harmful, depending upon the distributed lag effect of money flows, volume times transaction’s velocity. That saturation point is determined by the rate of inflation, the monetary fulcrum, the lag’s pivot. This is perfectly clear. It is aptly demonstrated by the distributed lag effect of money flows being mathematical constants, for > 100 years.

  7. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    21. May 2021 at 07:54

    Inflation peaked with the May release of April’s #s. It won’t substantially subside until Feb. 2022.

  8. Gravatar of ssumner ssumner
    21. May 2021 at 08:11

    BC, I think it shows that the balance of risks has shifted to overshooting.

    Asher, Even in 1978 there were ratex models. I think the real problem is reasoning from a price change. You have papers where negative supply shocks are expansionary because they create inflation, which reduces real interest rates. That’s wildly implausible. NK economics would be better off if inflation were entirely dropped from the model.

    Howard, China has a plan against America?

    Todd, I’m taking a wait and see attitude on the Fed’s policy. Time will tell whether they are serious about AIT. But at a minimum they’ve convinced markets they don’t view 2% as a ceiling.

  9. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    21. May 2021 at 09:17

    re: “Time will tell whether they are serious about AIT”

    That implies that they can accurately forecast gdp. I have my doubts.

  10. Gravatar of CCP kills people CCP kills people
    21. May 2021 at 09:41

    So is Scott Sumner still claiming the Russians stole the election and shrieking about tapes and interference or has he become a full apologist of the CCP totalitarian reign of terror?

    People who shill for the CCP now will have the same historical legacy of people who supported good relationships with Nazi Germany in the 30s.

    At least when some of those people in the 30s tried to find excuses for the Nazis, they weren’t aware the Nazis were going to kill millions. But Sumner already knows the CCP has killed millions and keeps killing – he just doesn’t care.

  11. Gravatar of Michael Sandifer Michael Sandifer
    21. May 2021 at 11:33

    I think there’s a non-trivial possiblity that, quite contrary to overshooting for any meaningful interval, we will be disappointed due to the weakness US economic recovery. The 10 year Treasury rate, for example, is still below the pre-pandemic level, while the S&P 500 is above the pre-pandemic trend. Hence seems likely that much of the rise in the stock market, just as is largely true in the housing market, is likely due to lower discount rates. As has been the case since the end of the tech boom, interest rates have been too low, because monetary policy’s been too tight.

  12. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    21. May 2021 at 12:05

    re: “because monetary policy’s been too tight”

    An increase in real interest rates is associated with an increase in the velocity of circulation. Banks don’t lend deposits so velocity will continue to decelerate. Then the FED tries to offset the decline in AD by adding new money. That will slow growth even further. We are obviously headed for another Great Depression as the FED is driving the economy in reverse.

  13. Gravatar of Michael Sandifer Michael Sandifer
    21. May 2021 at 13:03

    Spencer Bradley Hall,

    I’ve been wrong about many things, and perhaps I was right for the wrong reasons about NAIRU being lower than most thought. I think potential real GDP is higher than most think, and I was hoping average inflation targeting and an eagerness to climb out of such a deep hole would allow that idea to be tested, but I increasingly think that my framework will instead be tested by testing my pessimism.

    I shudder to think what will happen politically if we have another slow recovery. Let’s hope we’re wrong.

  14. Gravatar of John Trainor John Trainor
    21. May 2021 at 13:51

    This is also slightly off topic, a bleg.

    I teach economics to high school students and like to find and use original sources for my understanding and theirs. Good examples are quotes from Adam Smith and Bastiat, Smith’s discussion of compensating differentials, the abstract of Friedman’s 1968 paper on Monetary Policy, and the first chapter of Irving Fisher’s “The Theory of Interest”, and Armen Alchian’s “Universal Economics”.

    Keynes (via Hicks perhaps) used AD, but in the Keynesian Cross. Robert Barro criticized it in 1994, referring to a few published articles and several textbooks. The earliest textbook cited, written by the same Robert J. Gordon, was published in 1987.

    Here’s the bleg: where and how did the AD/AS model begin?


  15. Gravatar of Lizard Man Lizard Man
    21. May 2021 at 14:54

    “China has a plan against America?”

    I think it is a bit solipsistic to say that China has a plan against the US, but it also seems unrealistic to say that China doesn’t have plans that if realized would greatly reduce the influence of the US throughout the world. Calling that a plan against the US I think will blind folks in the US to their own interests, which aren’t by necessity incompatible with a more influential China.

    As for inflation, I don’t quite understand why any of the models would have problems differentiating between demand driven inflation and inflation from a supply shock. It seems easy to understand the difference when using plain old language, and I thought that a key part of being an economist was becoming proficient in multiple branches of mathematics, such that adjusting a toy model to distinguish between demand and supply shock inflation should be pretty easy for economists. Is there something about the building of toy model economics I am missing?

  16. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    22. May 2021 at 06:47

    Monetary inflation subsides in Feb 2022:

    monetary flows:

    parse date; inflation index

    12/1/2020 ,,,,, 1.26
    01/1/2021 ,,,,, 1.31
    02/1/2021 ,,,,, 1.41
    03/1/2021 ,,,,, 1.51
    04/1/2021 ,,,,, 1.54 peaks/plateaus
    05/1/2021 ,,,,, 1.47
    06/1/2021 ,,,,, 1.44
    07/1/2021 ,,,,, 1.47
    08/1/2021 ,,,,, 1.45
    09/1/2021 ,,,,, 1.40
    10/1/2021 ,,,,, 1.40
    11/1/2021 ,,,,, 1.28
    12/1/2021 ,,,,, 1.37
    01/1/2021 ,,,,, 1.42
    02/1/2021 ,,,,, 1.06 subsides
    03/1/2021 ,,,,, 0.84
    04/1/2021 ,,,,, 0.77
    05/1/2021 ,,,,, 0.70

  17. Gravatar of ssumner ssumner
    22. May 2021 at 08:35

    John, I’m not sure, but I was teaching AS/AD well before 1987, so it must be in older textbooks.

    Lizard, One problem is that economics define these terms in different ways, which leads to disputes that are mostly about language.

  18. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    23. May 2021 at 07:18

    “The last 20 years’ steady decline in money velocity is indicative of a shrinking economy.”

    Isn’t that reasoning from a price change?

  19. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    23. May 2021 at 10:48

    The current deceleration in R-gDp during the last half of 2021 will become a negative rate-of-change in the 1st qtr. of 2022. But a new injection of money, a new stimulus package, in the “sweet spot” or in the 1st qtr. downswing, will then push the much longer-term flexible “average” level of inflation to a higher plateau. This action can only be done so many times before the inflation level becomes counterproductive/consumes us. I.e., the continuation of the FED’s short-term pushes will force the long-term average too high.

  20. Gravatar of dtoh dtoh
    23. May 2021 at 16:09


    Good post. I think there are a few things that are confusing to policy makers right now.

    1. Different types of inflation that you mention.
    2. Impact of unemployment benefits on “full employment”, growth and inflation.
    3. Uncertainty about tax increases which impacts the relationship between equity and fixed income pricing.

  21. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    25. May 2021 at 08:07

    re: “the 1978‐era Phillips Curve”

    Every economist missed their forecast for the rate of inflation in 1978. In 1978 (when Vi fell, but Vt rose) all economist’s forecasts for inflation were drastically wrong. Put into perspective: There were 27 price forecasts by individuals & 9 by econometric models for the year 1978 (Business Week). The lowest (Gary Schilling, White Weld), the highest, (Freund, NY, Stock Exch) & (Sprinkel, Harris Trust & Sav.).

    The range CPI, 4.9 – 6.5 percent. For the Econometric models, low (Wharton, U. of Penn) 5.7%; high, 6.6% U. of Ga.). For 1978 inflation based upon the CPI figure was 9.018% [and Leland Prichard, in his Money and Banking class, predicted 9%].

  22. Gravatar of ssumner ssumner
    25. May 2021 at 15:38

    dtoh, Agree on all three points.

  23. Gravatar of Jasur Urunboyev Jasur Urunboyev
    28. May 2021 at 02:00

    Hi dtoh, I think that the third point -“Uncertainty about tax increases which impacts the relationship between equity and fixed income pricing” is key. Nobody wants to work to the detriment of their own pocket.

Leave a Reply