Edward Nelson on Milton Friedman

While I’m only 275 pages into Ed Nelson’s big 2 volume set entitled “Milton Friedman & Economic Debate in the United States”, I can already say that it’s one of my favorite books on macroeconomics.

One issue that has frequently puzzled me is how to interpret causality in the Phillips Curve relationship. I have always interpreted Friedman’s natural rate model as one where causality went from inflation to output. More specifically, if inflation is higher than expected, then unemployment will be lower than the natural rate, and vice versa (perhaps due to sticky nominal wages.)

Keynesians usually seem to interpret causality in the opposite direction, low unemployment causes high inflation, and vice versa. And today, that seems to be the most widely accepted interpretation. Nelson (p. 273) quotes Friedman offering an interpretation that is consistent with my view:

There was, however, a crucial difference between Fisher’s analysis and Phillips’s, between the truth of 1926 and the error of 1958, which had to do with the direction of causation. Fisher took the rate of change of prices to be the independent variable that set the process going.

But Nelson also cites other statements by Friedman that are consistent with the Keynesian interpretation. He then suggests:

Friedman was also clear that both inflation and unemployment were endogenous variables. That being the case, neither a story based on causation from unemployment to inflation nor a story based on causation from inflation to unemployment can be accepted as a comprehensive description of the Phillips Curve relationship.

You probably know what I’m going to say. What’s really going on is that both inflation and unemployment are affected by NGDP shocks. When NGDP rises faster than expected, it increases inflation and reduces unemployment. When NGDP rises slower than expected, inflation tends to fall and unemployment rises. But what does Nelson say? How does he reconcile various statements by Friedman that seem inconsistent?

The answer offered here is that Friedman’s perspective was that, although inflation and output were jointly determined, the former variable could in large measure be usefully regarded as the driver of the relationship because inflation is a nominal variable and hence ultimately policy determined. Fluctuations in output (in relation to potential) would not occur if the private sector’s expectations of nominal variables corresponded continuously to the actual paths.

I really like this explanation. In Chapter 1 of my book coming out in July, I discuss the tricky issue of causality in macroeconomics. In the end, I conclude that statements about causal relationships should be judged on their usefulness. Thus it’s useful to say that monetary policy caused the Great Recession if another plausible setting of monetary policy instruments would have prevented the Great Recession, but not otherwise.

While Friedman is my favorite macroeconomist, I don’t believe that he got everything right. Like most people of his generation, he focused a lot of attention on inflation. And yet in many cases, his analysis makes more sense if you substitute NGDP growth for inflation. Thus his claim that a slowdown in inflation almost always results in higher unemployment would be more accurate if applied to a slowdown in NGDP growth. After all, inflation can slow due to a positive supply shock.

I’ve recently discovered another Ed Nelson paper on whether it makes sense to think in terms of “nominal shocks”, and will have more to say on this issue in a future post.

PS. Friedman’s quote is referring to a 1926 paper by Fisher that first developed the so-called “Phillips Curve”. It was rediscovered by Phillips in 1958, who gave it a Keynesian interpretation. Both Friedman and I prefer Fisher’s interpretation.



39 Responses to “Edward Nelson on Milton Friedman”

  1. Gravatar of marcus nunes marcus nunes
    2. June 2021 at 18:13

    This post clearly shows that the best explanation for falling/low unemployment & low/stable inflation is a monetary policy that succeeds in keeping NGDP growth on a stable path.

  2. Gravatar of Michael Sandifer Michael Sandifer
    2. June 2021 at 20:48

    I strongly suspect you were writing from the perspective of the Phillips curve model, which you don’t buy, but this quote from you strikes me as wrong:

    “More specifically, if inflation is higher than expected, then unemployment will be lower than the natural rate, and vice versa (perhaps due to sticky nominal wages.)”

    Obviously, not all positive inflation shocks are large enough to take unemployment below the natural rate.

    I have a similar problem with this other quote:

    “Thus his claim that a slowdown in inflation almost always results in higher unemployment would be more accurate if applied to a slowdown in NGDP growth.”

    Inflation can surprise on the downside without raising unemployment. It could just slow employment growth, for example.

    I think unemployment results when NGDP growth is expected to fall below that of wages. Usually, unemployment rises during a recession.

    I think the inflation rate can only change to the degree wage growth is flexible, or that output falls. When below full employment, this limits inflation growth, unless there’s a sufficiently large shock as to set the expectation that output will exceed sustainable limits for the foreseeable future, after which inflation expectations can boost wage growth, allowing inflation more room to soar.

    When approaching full employment, wage growth and inflation begin to rise with real GDP, as illustrated in the SR AS/AD model. My view is entirely consistent with that traditional model, though, implicitly, I don’t think many economists take it seriously.

  3. Gravatar of Michael Sandifer Michael Sandifer
    3. June 2021 at 03:34

    Another way of putting it is that the lack of flexibility in the change in wages limits the flexibility of prices, unless changes in the quantity of output or labor are restricted as demand rises, or changes in prices become less restricted as employment falls in line with demand.

  4. Gravatar of Michael Sandifer Michael Sandifer
    3. June 2021 at 03:52

    My guess is that core CPI or PCE prices can temporarily be more flexible downward during recessions, due to sudden inventory overhangs.

  5. Gravatar of Carl Carl
    3. June 2021 at 06:52

    What are your assumptions about the relationship of automation to rises in NGDP? Do you expect it to be proportional to the drop in unemployment?

  6. Gravatar of ssumner ssumner
    3. June 2021 at 09:24

    Carl, I’m not sure what you are asking. Are you asking if there is a causal relationship between automation and NGDP? Can you rephrase the question?

  7. Gravatar of Brian Donohue Brian Donohue
    3. June 2021 at 11:45

    I think if Friedman were around today, he’d agree 100% with your modification.

    Friedman tried to pin down the tricky “quantity of money” variable, which is basically the Fed tool for hitting NGDP targets, right?

  8. Gravatar of Carl Carl
    3. June 2021 at 11:54

    Yes, I was trying to find out if there is also a causal relationship between unexpected NGDP growth and automation (I forgot to include the word “unexpected” in my first question). I originally assumed the answer was a simple yes because there is a relationship between NGDP growth and hiring and because automation is a substitute for more hiring. But, if the NGDP growth is unexpected, maybe that would mean it would spur more spending on hiring than automation because automation has a longer investment horizon.

  9. Gravatar of ssumner ssumner
    3. June 2021 at 11:58

    Carl, There may be some correlation at cyclical frequencies, but over the long run NGDP doesn’t impact automation.

  10. Gravatar of marcus nunes marcus nunes
    3. June 2021 at 12:18

    Missing Friedman!

  11. Gravatar of David S David S
    3. June 2021 at 13:25

    Marcus reminded me of a question I’ve been meaning to ask:
    Are the following scenarios effectively the same with respect to achieving sustainable employment stability?

    Scenario 1: 2% PCE & 5% NGDP
    Scenario 2: 4% PCE & 8% NGDP

    I’m trying to frame this as a hypothetical–despite current economic conditions bearing similarity to Scenario 2, the long term goal is a Scenario 1, right?

    Happy to get a smackdown on this.

  12. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    3. June 2021 at 15:39

    Friedman was one dimensionally confused. Banks aren’t intermediaries. When a bank makes a loan, the money stock is altered. If a commercial bank was just a conduit, there wouldn’t be a change in the volume of money. How do you explain the growth in the money stock over the decades? It is a fallacy of composition.

    After the monetization of time deposits between 1961 and 1981 (which increased velocity), the payment’s system incessantly bottled-up savings. This idling of savings, the universal mistaken belief that banks are intermediaries (the complete deregulation of interest rates for just the banks), continually destroyed money velocity.

    Unless the upper income quintiles’ savings are expeditiously activated, completing the circuit income velocity of funds, a dampening economic impact is generated. Subpar economic growth is about the paradox of thrift.

  13. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    4. June 2021 at 04:41

    “Moody’s data showed that nearly two-thirds of the excess savings (2 trillion), in the US is by households in the top 10% of the income distribution, and three-quarters is by those in the richest 20%.”

    Bidens’ trillions of dollars infrastructure stimulus package contains: “The belief is that the U.S. economy can’t grow faster than 1.9% over the long term because the U.S. population is aging and demographics is destiny. Productivity growth is fated to slow down, and tax and regulatory policy doesn’t matter.”

    Krugman’s modern-day description? “What we’re looking at here is a world awash in savings with nowhere to go.”

    See: Should Commercial Banks Accept Savings Deposits, Conference on Savings and Residential Financing (1961 proceedings), United States Savings and Loan League, Chicago, 1961, pp 42, 43

    Completely wrong. 1963 “Profit or Loss from Time Deposit Banking” in Banking and Monetary Studies, Comptroller of the Currency, United States Treasury Department, Irwin, pp. 369-386

    “Furthermore, it seems highly improbable (and in contradiction to Professor Chandler’s theoretical analysis) that the stoppage in the flow of these funds (time deposits) is entirely compensated for by an increased velocity of the remaining demand deposits. It is quite probable that the growth of time deposits shrinks aggregate demand and therefore produces adverse effects on GNP…”

    “Monetary savings (funds held beyond the income period in which received), cannot be used to finance investment until their owners (nonbank public) so decide, and so long as the nonbank public chooses to hold savings in the form of time deposits, the means-of-payment velocity of these funds is zero and the funds are lost to investment…”

    The ”Taper Tantrum” (reduction in the FDIC’s unlimited transaction deposit insurance), was prima facie evidence. Monetary policy was concurrently tightened, real interest rates rose and the dollar strengthened.

    The Taper Tantrum’s policy is diametrically opposed to Covid-19’s policy.

  14. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    4. June 2021 at 05:34

    Krugman: “To maintain full employment, a market economy must persuade businesses to invest all the money households want to save.”

    “Why not borrow cheaply and use the funds to rebuild our crumbling infrastructure, invest in the health and education of our children, and more?”

    2/3 of the U.S. Golden Era in Capitalism was financed with velocity, i.e., by putting savings back to work.

    Money products have a reduced rate of turnover as compared to savings products. i.e., savings transferred through the nonbanks represents an exchange in pre-existing deposit liabilities in the payment’s system, a velocity relationship. That’s why velocity has fallen. Bernanke destroyed the nonbanks.

  15. Gravatar of Michael Sandifer Michael Sandifer
    4. June 2021 at 05:39

    That we should accept that negative real rates years out on the yield curve with above 4% positive NGDP growth expectations as normal, is a perspective only an economist could hold. To be clear, that’s not complimentary, from my perspective.

  16. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    4. June 2021 at 05:57

    The utilization of savings has a ripple effect, it increases the real rate of interest.

  17. Gravatar of ssumner ssumner
    4. June 2021 at 08:22

    David, I’m not sure what you are asking. Those are not two policy options.

    Michael, You said:

    “That we should accept that negative real rates years out on the yield curve with above 4% positive NGDP growth expectations as normal, is a perspective only an economist could hold.”

    Most non-economists don’t know what that sentence means, and hence don’t hold any “perspective” on the question.

  18. Gravatar of Randomize Randomize
    4. June 2021 at 09:32

    Nice post. It’s refreshing to get back to econ now that he-who-shall-not-be-named is out of the news.

    While reading this, it was helpful to me to focus on monetary stance rather than inflation/deflation. Tighter money could *cause* lower nominal rates, higher real rates, and higher unemployment and loose money could cause the opposite. Am not sure that this adds anything to the conversation but it’s what went through my head.

  19. Gravatar of postkey postkey
    5. June 2021 at 01:46

    ‘An empirical test is conducted, whereby money is borrowed from a cooperating bank, while its internal records are being monitored, to establish whether in the process of making the loan available to the borrower, the bank transfers these funds from other accounts within or outside the bank, or whether they are newly created. This study establishes for the first time empirically that banks individually create money out of nothing. The money supply is created as ‘fairy dust’ produced by the banks individually, “out of thin air”. . . .
    ‘Thus it can now be said with confidence for the first time – possibly in the 5000 years’ history of banking – that it has been empirically demonstrated that each individual bank creates credit and money out of nothing, when it extends what is called a ‘bank loan’. The bank does not loan any existing money, but instead creates new money. The money supply is created as ‘fairy dust’ produced by the banks out of thin air.32 The implications are far-reaching.’

  20. Gravatar of postkey postkey
    5. June 2021 at 01:51

    “This paper presents a second empirical test, using an alternative methodology, which allows control for all other factors. The financial intermediation and the fractional reserve theories of banking are rejected by the evidence . . . “

  21. Gravatar of postkey postkey
    5. June 2021 at 01:54

    Empirical evidence offered to this ‘religious’ site?

  22. Gravatar of ssumner ssumner
    5. June 2021 at 09:42

    Randomize, Trump is still in the news. He recently suggested that he’ll be back in office by August. His former national security advisor is advocating a Myanmar-style military coup. He’s the GOP frontrunner for 2024.

    Postkey, You said:

    “This study establishes for the first time empirically that banks individually create money out of nothing. The money supply is created as ‘fairy dust’ produced by the banks individually, “out of thin air”. . . .”


  23. Gravatar of Carl Carl
    5. June 2021 at 10:05

    It strikes me as odd that there is a causal relationship between NGDP growth and employment, but not between NGDP growth and something that is a substitute for employment, automation. I must be missing something about the nature of the relationship between employment and automation.

  24. Gravatar of Michael Sandifer Michael Sandifer
    5. June 2021 at 10:36


    True, but there are plenty in the finance field without econ degrees, not that they seem to do as well as economists on economics overall. Quite the contrary.

  25. Gravatar of Michael Sandifer Michael Sandifer
    5. June 2021 at 10:41

    Marcus Nunes,

    Having read some of your Substack posts recently, it seems we have very similar views about the economic outlook, but I got there by interpreting the forward expected path of NGDP growth implicit in the S&P 500 index, and then relying on TIPS spreads to deem inflation expectations versus those of real growth.

    Many people are going to be surprised by how weak the real recovery is, after perhaps an initial burst.

  26. Gravatar of ssumner ssumner
    6. June 2021 at 08:24

    Carl, There’s no long run relationship between NGDP and either employment or automation. There may be a short run relation with either, but employment is easier to change in the short run than automation.

  27. Gravatar of Spencer Bradley Hall Spencer Bradley Hall
    6. June 2021 at 09:44

    Lending by the Reserve and commercial banks is inflationary (results in an addition to the supply of money relative to the supply of goods offered in exchange) Lending by the nonbanks is noninflationary (matches savings with investments).

    So lending by the Reserve and commercial banks decreases the real rate of interest whereas lending by the nonbanks increases the real rate of interest. That in and of itself is a reason N-gDp targeting can’t ultimately work.

  28. Gravatar of Carl Carl
    6. June 2021 at 13:49

    Thanks. That makes sense.

  29. Gravatar of Willy2 Willy2
    7. June 2021 at 01:36

    – (Price) Inflation is actually EXTREMELY simple.

    – In the 1960s & especially the 1970s workers/employees received wage increases at or above inflation. So, when inflation was at e.g. 5% then workers would easily receive a wage increase of 5%, 6%, 7% and up to even 10%. But when a worker receives a 8% wage increase then he/she can spend 8% more. And that allows a manufacturer to increase its prices by 8% as well (= price inflation). Now inflation rises from the afore-mentioned 5% to(wards) 8%.
    – That changed in the (very) early 1980s. Then employers started to keep a lid on wage growth. Let’s assume that inflation was at 5% then workers/employees would get a wage increase of only say 3%. Then workers can spend 3% more. But a manufacturer that increases its prices by say 7% is guaranteed to lose (3 – 7=) 4% in revenues. So, manufacturers are forced to limit the price increases to say 3% as well. Now (price) inflation now falls from 5% to(wards) 3%.

    – A LOT OF people give the FED (+ Volcker) for “killing” inflation but it was actually corporate America that was responsible for killing inflation.

    (J.M. Keynes sends his greetings)

  30. Gravatar of ssumner ssumner
    7. June 2021 at 07:38

    Willy, You are mixing up nominal and real variables.

  31. Gravatar of Willy2 Willy2
    7. June 2021 at 08:41

    – “Mixing up nominal and real variables” ??? Please explain …

  32. Gravatar of Friedman as a critic of Keynesian economics – Econlib Friedman as a critic of Keynesian economics - Econlib
    8. June 2021 at 01:17

    […] a recent post, I quoted Friedman on the Phillips Curve, an area where he was also skeptical of Keynesian […]

  33. Gravatar of ssumner ssumner
    8. June 2021 at 08:20

    Willy, Market power in the labor market impacts real wages. Inflation is a nominal process, where all wages and prices rise over time. In economics, it’s called the “classical dichotomy”.

    Cost push theories of inflation were popular in the 1950s and 1960s, but went out of style for good reason; they don’t explain NGDP growth. NGDP growth was 11% from 1971-81, and that requires easy money. Cost push doesn’t do that.

  34. Gravatar of postkey postkey
    9. June 2021 at 07:34

    L.O.L. is that a religious retort? 🙂

  35. Gravatar of postkey postkey
    10. June 2021 at 00:32

    “Saving does not by itself increase the deposits or ‘funds available’ for banks to lend. Indeed, viewing banks simply as intermediaries ignores the fact that, in reality in the modern economy, commercial banks are the creators of deposit money. This article explains how, rather than banks lending out deposits that are placed with them, the act of lending creates deposits — the reverse of the sequence typically described in textbooks.”

    ‘L.O.L.’ ≡ ‘economic ignoramus’?

  36. Gravatar of Doug M Doug M
    10. June 2021 at 13:50

    Why are we still discussing the Phillips curve? The Phillips curve is flat. There may have been an obvious empirical relationship between inflation and unemployment that was visible in the historical data prior to 1958. But, over the last 50 years, the relationship not so clear.

    We entered a phase of bad economic policy when we treated the historic correlation between these two variables as something rigid. We arrogantly thought that we could trade one for the other, and the relationship broke down. It is a classic post hoc ergo propter hoc fallacy.

    It may be a reasonable model to say that changes in NGDP drive changes in both variables, but there is no causal relationship between the two.

  37. Gravatar of Lewis Lewis
    12. June 2021 at 01:40

    May I ask for your opinion on the Fed’s monetary policy now? From what I understand from your posts in 2009, you would think that the rising inflation (or NGDP) is a thing to cheer for, right?

    In the same regard, in your opinion, when should the Fed worry (or control) inflation? Once the unemployment gets stabilized to pre-covid level?

  38. Gravatar of ssumner ssumner
    12. June 2021 at 08:03

    Lewis, It seems about right. The Fed should always be concerned with inflation. They have a 2% AIT target. That’s not my preference, but if that’s their target then they should hit it.

    As an analogy, would someone ask, “Is this a time when the captain should be concerned with steering the boat?” Yes, this and all other times.

    It’s hard to judge the unemployment rate right now, as it’s being heavily distorted by the supplemental UI program which ends in September.

  39. Gravatar of Willy2 Willy2
    15. June 2021 at 01:12

    – The entire “Cost Push” idea is nonsense. Because it’s a “Demand Pull” issue.

    J.M. Keynes sends his greetings to M. Friedman.

Leave a Reply