Why level targeting?

I’ve done a number of posts advocated level targeting, but many readers remains skeptical. Here I’ll try to provide a bit more intuition in support of the idea.

Let’s think about a model with NGDP growth rate targeting at 4%/year. Assume that once every 5 years the central bank makes a mistake and NGDP moves 4% off target (i.e. 0% or 8% growth.) Then they let bygones be bygones and establish a new NGDP trend line from that point forward. Also assume that half of the errors are in an expansionary direction and half are in a contractionary direction. Thus once in every 10 years you get a recession, and once in every 10 years you get overheating and high inflation.

At first glance, it seems like switching to level targeting would be a step backwards. Now you’d see periods of declining NGDP occurring once every 5 years, not once every 10 years. This would occur when there was a contractionary mistake, and it would also occur when the Fed had to correct an expansionary mistake and get back to the old trend line. Twice as many recessions! That sounds bad.

I see two reasons why level targeting would actually produce better results than growth rate targeting. I’m going to focus on the issue of correcting overshoots, as many opponents of NGDP level targeting (including Jay Powell) have no problem with correcting undershoots.

[BTW, I am not suggesting we now correct the recent overshoot, because we don’t have a NGDPLT policy in place. I’ll explain the distinction later, but this is about setting up a new regime when the economy is currently in equilibrium, say back in 2019.]

I believe that NGDP growth targeting proponents overestimate the cost of correcting overshoots for two reasons:

1. They overlook the fact that level targeting would make overshoots (and undershoots) much smaller in the first place.

2. The overlook the fact that the pain of correcting NGDP overshoots from a booming economy is much smaller than the pain of an equivalent reduction in NGDP starting at trend.

Both of these conjectures require some explanation:

Suppose that in mid-2021, the Fed had announced that any overshoot of NGDP would be erased as quickly as possible, using a tight money policy to bring NGDP back to the trend line. In that case, as NGDP started heading for an overshoot in late 2021, interest rates (relative to the natural rate) would have soared much higher, in anticipation of a future tight money policy to bring NGDP back to trend. BTW, most of the contractionary “work” would have been done by lowering the natural interest rate, not raising the actual policy rate.

Instead, the Fed held their policy rate far below the natural rate in late 2021, causing NGDP to dramatically overshoot the pre-Covid trend line.

So yes, under level targeting there’d be twice as many periods of declining NGDP as under growth rate targeting, but they’d be correcting much smaller overshoots—say 1% or 2%, not 4% as with growth rate targeting. As an aside, many New Keynesian models suggest that changes in the current level of aggregate demand are heavily influenced by changes in future expected levels of aggregate demand. In these models, an expectation of future tightening makes money tighter right now. Thus under level targeting, the market helps to stabilize the economy. This may be one reason why New Keynesians such as Michael Woodford have advocated NGDPLT.

In addition, the corrections would be much less painful, even for any give reduction in NGDP. To see why, we need to consider the components of NGDP growth. Recall that NGDP is also gross domestic income. Overshoots of NGDP can be decomposed into overshoots of profits, hours worked, and nominal hourly wages.

When overshoots are corrected quickly, almost all of the action takes places with profits and hours worked. During the period of correction, the economy moves from excess profits and higher than nominal hours worked back to normal profits and normal hours worked. That’s not painful!

What would be painful? Suppose the excessive monetary stimulus were persistent. Eventually, those sticky nominal wages would begin rising. Output would return to the natural rate. In order to reverse these excessive wage gains, you’d probably need to create “slack” in the economy, i.e. higher unemployment. That doesn’t mean a softish landing is impossible, but it’s far more difficult than under level targeting.

The best time to correct the policy mistake is right away, before the NGDP overshoot spills over from profits and hours worked into higher nominal hourly wages. This sort of correction will be much less painful than a reduction in NGDP starting from trend, which would likely create a recession.

Some commenters ask me, “Given that you favor level targeting, what should the Fed do now?” To begin with, the Fed should only do NGDP level targeting if it has already announced a very specific NGDP level target. It has not done so, and hence it would now be inappropriate to return to the pre-2020 trend line. We don’t even know what trend line they would have chosen if (back in 2019) they had opted for NGDP level targeting.

So that’s all a moot point. If the Fed is targeting 2% inflation then they should target 2% inflation. I’m not happy with the target—I’d prefer NGDPLT—but if that’s the target then take it seriously. (Obviously the dual mandate allows some flexibility, but at a minimum the long run inflation rate needs to average 2%.)

Some people still fail to see that monetary policy was too expansionary during 2022. Admittedly, by that time it was too late to return to the old trend line, but we should at a minimum have reduced NGDP growth to the old trend line. Instead, the economy rose further and further above trend and the labor market continued to be wildly overheated. In 2022, we could have had a tighter monetary policy, lower inflation, and a better functioning labor market.

If I switch from eating three donuts every breakfast to two donuts every breakfast, that certainly doesn’t mean I’m dieting. Money was too expansionary in 2022, even if NGDP growth slowed late in the year. Growth was still far too rapid.

[BTW, the real consumption figures from 2022 are highly suspect, as they don’t account for the horrendous decline in the quality of service faced by consumers of things like travel and hotels. As we return to normal, the rate of productivity growth rate will be understated.]

PS. Here’s a shorter version of this post. Macroeconomic instability is caused by fluctuations in the growth rate of the aggregate nominal wage rate. Level targeting stabilizes nominal wages more effectively than growth rate targeting.



26 Responses to “Why level targeting?”

  1. Gravatar of Randomize Randomize
    1. August 2023 at 10:02

    Dr. Sumner, I struggle to reconcile what would happen if we had a LARGE one-time overshoot. Imagine we were targeting 50% NGDP growth over 10 years (~4% annualized). Then in year one, we screwed up and had 100% growth in a single year. Level targeting would prescribe -50% over the following 9 years.

    I would hope most everyone would agree that while sudden and large NDGP growth (likely accompanied by large inflation) is obnoxious, declining NGDP and deflation causes depressions as sticky wages would drive falling labor hours and business simply can’t survive selling goods for less than they paid for them. 9 years of depression seems like too big a price to pay for the sake of hitting a target number. What do I have wrong here? Your thoughts?

  2. Gravatar of John Hall John Hall
    1. August 2023 at 11:08

    “Some commenters ask me, “Given that you favor level targeting, what should the Fed do now?””

    Oh, I think I’m one of those! Thanks for addressing in more detail.

    I think the other side to the question is something along the lines of “suppose the central bank has already adopted NGDPLT and the country gets into a war and the legislature significantly increases deficit spending, would the legislature continue to let the central bank pursue NGDPLT?”

  3. Gravatar of Siddharth Siddharth
    1. August 2023 at 13:59

    Why use NGDP and not Nominal Gross Output (Mark Skousen’s metric that BEA began publishing). It seems that would give a much broader picture of economic output. If I remember correctly, Nicolas Cachanosky has also spoke favorably about Nominal Gross Output Targeting, if I remember correctly.

  4. Gravatar of Siddharth Siddharth
    1. August 2023 at 14:02

    Why do you use NGDP and not Nominal Gross Output (Mark Skousen’s metric that BEA began publishing). It seems that would give a much broader picture of economic output. If I remember correctly, Nicolas Cachanosky has also spoke favorably about Nominal Gross Output Targeting, if I remember correctly.

  5. Gravatar of ssumner ssumner
    1. August 2023 at 18:41

    Randomize, That would not happen with a credible NGDPLT regime, so it’s a moot point. The purpose of NGDPLT is to prevent those sorts of overshoots in the first place. I doubt that overshoots would ever exceed 1% or 2% under NGDPLT.

    John, I think they would allow it.

    Siddharth, How does that differ from NGDP? If it double counts intermediate inputs, then it’s not a good target.

  6. Gravatar of Siddharth Siddharth
    1. August 2023 at 19:36

    Does the relevance of NGDP not come from MV=PQ? The amount of times a dollar changes hands should in theory include spending on intermediate goods, and also financial instruments (which unfortunately Gross Output does not include). Ff we’re really targeting monetary equilibrium, Gross output + financial assets should be targeted.

  7. Gravatar of ssumner ssumner
    1. August 2023 at 19:46

    Siddharth, You said:

    “Does the relevance of NGDP not come from MV=PQ?”

    No, that’s just an identity, and has no bearing on what policy is best. NGDP is relevant because it’s correlated with the labor market. Gross output is not.

  8. Gravatar of Michael Sandifer Michael Sandifer
    1. August 2023 at 20:11

    If the US economy was in equilibrium in 2019, why was earnings growth still so slow, while stock prices soared nearly 29%, with well-below target inflation and even lower inflation expectations?

    That seems more like a weak economy with the expectation of some catch up growth as wages adjust.

  9. Gravatar of Michael Sandifer Michael Sandifer
    1. August 2023 at 20:18

    Earnings only grew about 3% in 2019. Core PCE rose just over 1.8%. The 5 year breakeven rate was 1.88% at its height for 2019 in March, and hit a low of 1.27% in October, while finishing the year at 1.66%. Also, employment growth slowed.

  10. Gravatar of Siddharth Siddharth
    1. August 2023 at 21:17

    Perhaps this is just a difference of opinion in rationale, but it seems to me that monetary equilibrium is correlated with the entire macroeconomy. As Garrison and Horwitz have argued, when there’s a disequilibrium in money, this leads to disequilibrium across all markets, since money is one of two markets that every transaction involves (the other being the market for time, aka interest).

    Thus monetary equilibrium, when money supply and money demand (inverse of V) are in tandem, is crucial for macroeconomic stability, and related to all markets including that for labor. a stable MV, provides monetary equilibrium.

    So if MV= Nominal Gross Output + financial assets, that is what should be targeted.

  11. Gravatar of Sara Sara
    1. August 2023 at 21:19

    It’s not possible to target 2%. To achieve the 2% rate, at this point, after a decade of easy money and massive printing, you’d have to raise interest rates several points above the inflation rate, and that would make the debt almost impossible to pay. Servicing the debt will already hit 1 trillion per year which is about 20% of the receipts. You can spend 80B on more tax agents, and like the Roman empire in its final days, or like Hitler’s gestapo, desperately harass and torment blue collar folks in an effort to viciously and maliciously increase inflows, or you can stop printing money. The latter is obviously the best way to reduce inflation. That means ending unnecessary wars, like the one in Ukraine, and recognizing that the nation is essentially broke.

  12. Gravatar of Dr Richard Dr Richard
    1. August 2023 at 21:24

    “at a minimum the long run inflation rate needs to average 2%.”

    Can you explain why the average needs to be around 2% inflation instead of 0% inflation? What about 4% inflation? I don’t know much about how different the economy reacts between 0% inflation and 4% inflation.

  13. Gravatar of Michael Sandifer Michael Sandifer
    2. August 2023 at 02:23

    Looking at the numbers I presented, an equilibrium NGDP growth rate in 2019 would have been just over 5%, minimum, reflecting the gap bewteen 3% S&P 500 earnings growth and NGDP growth (just over 4%) amd that between NGDP growth and stock price appreciation (1.29 × roighly 4%).

  14. Gravatar of John Hall John Hall
    2. August 2023 at 04:51

    “John, I think they would allow it.”

    I’m a bit more skeptical…

  15. Gravatar of steve steve
    2. August 2023 at 06:36

    As usual, Sara is correct.

  16. Gravatar of spencer spencer
    2. August 2023 at 08:55

    Sara is not correct: “you’d have to raise interest rates several points above the inflation rate”

    Interest is the price of credit. The price of money is the reciprocal of the price level.

    And the ECB has the right idea:

    And the FRB-STL also has the right idea:

  17. Gravatar of Kevin Erdmann Kevin Erdmann
    2. August 2023 at 09:45


    The month-over-month inflation rate has been effectively at the 2% target since July 2022, when the fed funds rate was still about 1%. There is no reason to think that the real fed funds rate going forward has to be much above zero.

  18. Gravatar of spencer spencer
    2. August 2023 at 11:17

    re: “which would likely create a recession”

    You reverse any recession, i.e., disintermediation of the nonbanks, by driving the banks out of the savings business. The 1966 Interest Rate Adjustment Act is the model.

    I.e., lending by the banks is inflationary, whereas lending by the nonbanks is noninflationary (other things equal).

  19. Gravatar of Philo Philo
    2. August 2023 at 12:59

    It is hard to predict what NGDP is going to be, a year or two into the future; one will get it wrong repeatedly. It is much easier to gauge the market’s opinion about what NGDP is going to be. And for the purposes of monetary policy, the market’s present opinion is much more important than the eventual reality. If the Fed kept the market’s forecast of NGDP on a smooth track, it needn’t worry about occasional bumps in the pattern of actual NGDP (which, under such a policy regime, would be only slight deviations from trend).

    The Fed should target the *market forecast* of NGDP.

    3. August 2023 at 05:53

    Have you described anywhere how an NGDP futures market would actually be structured? What it would look like to lay people? Could you please point me to it?

    I presume only one market could be in existence at any one time, because players could game the system if multiple markets were running concurrently?

    I presume the market would be for NGDP about one year in the future? Would each market only last for a quarter, then a new market would be created for one year from the new date? And the payoff one year forward would be based on actual NGDP?

    I’m trying to understand, but I can’t get my brain around what an NGDP market would actually look like. Thanks for your help! And thanks for all you’re doing: the living pdf book seems like a potentially effective method to keep monetary policy moving toward a better future.

  21. Gravatar of ssumner ssumner
    3. August 2023 at 08:01

    Michael, We’ve discussed that issue before.

    Siddharth, You said:

    “So if MV= Nominal Gross Output + financial assets, that is what should be targeted”

    That would be a disaster. Financial transactions are orders of magnitude greater than goods transactions. You’d essentially be targeting the financial markets.

    Richard, I don’t favor a 2% inflation target. I said if that’s the target then the Fed needs to hit it.

    John, My point applies to small wars like Iraq. If there’s WWIII you might be right, but then monetary policy will be the least of our worries!!

    Philo, Exactly.

    Todd, There’s an explanation of the idea in chapter 5 of my new book:


  22. Gravatar of Michael Sandifer Michael Sandifer
    3. August 2023 at 19:58


    Yes, we’ve discussed the issue more than once, but there’s a subtle difference in my approach this time which leads to a not-so-subtle conclusion, which I believe in inescapable. This time I’m explicit about the earnings growth over the relevant period, which allows me to make my point more clearly.

    Let’s start with a hypothetical. Do you buy the notion that NGDP growth is a big driver of earnings growth? If expected earnings growth falls 50%, what happens to stock prices, in the case of the S&P 500? What happens to the price multiple? What happens to earnings?

    I argue that stock prices will fall 50%, though the multiple will double. This means earnings must fall at least 75%, and this is in fact what happens empirically.

    Whether you buy my model, which uses the expected NGDP growth rate as the discount rate under earnings to determine the S&P 500 price, or the Gordon growth model, something has to give with regard to earnings.

    So, what happens to stock prices when expected NGDP growth falls 50%? Is the outcome much different for earnings? In general, do you disagree that the price and earnings will fall, while the earnings multiple rises? If so, then obviously you agree earnings must fall more than the price.

    The implication is that in a booming economy, earnings should be booming even more than stock prices, if economic growth is at or near potential. That never happened during the recovery from the Great Recession, and the opposite happened in 2009, specifically. Earnings growth was lower than ngdp growth by more than 1%, as stock prices soared.

  23. Gravatar of Michael Sandifer Michael Sandifer
    3. August 2023 at 20:01

    That should read, “the opposite happened in 2019,…”, not 2009. In 2019, earnings grew more slowly than NGDP by more than 1%, while stock prices soared 29%.

  24. Gravatar of spencer spencer
    5. August 2023 at 02:53

    Siddharth is right. Financial transactions aren’t random.

    Example: In 1978 (when Vt rose, but Vi fell) 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 Dr. Leland Prichard, in his Money and Banking class, predicted 9%].

    See: Analysis of bank debits as a business cycle indicator (richmond.edu)

  25. Gravatar of spencer spencer
    5. August 2023 at 04:22

    re: “That would be a disaster. Financial transactions are orders of magnitude greater than goods transactions. You’d essentially be targeting the financial markets.”

    Pure B.S.

    The G.6 release’s primary proponent William G. Bretz (Juncture Recognition), died & the Bank Credit Analyst in Canada didn’t complain that they were losing their invaluable “debit/loan ratio”.

    See: New Measures Used to Gauge Money supply WSJ – 6/28/83

  26. Gravatar of Bob Bob
    5. August 2023 at 13:38

    I do not understand macro anywhere near as well as Dr Sumner, but therefore I also have a simpler story: Every loan and long term contract is, implicitly, also a bet on NGDP growth. If we targeted like we do inflation, we are still going to be gambling: Undershoots or overshoots would not get corrected, ever, except by coincidence. Withlevel targeting, we gamble less, as we will be far closer at guessing what ngdp will be like 30 years from now. This lets loan evaluation be all about things the parties are experts on, not guessing what the central bank’s biases are going to be.

Leave a Reply