More on Population Growth as a Cause of Inflation: Mark Sadowski and Steve Waldman

     Yes, it was a pleasure just being nominated. Steve Waldman has a response to the debate that started between him and Sumner whether or not the Great Inflation of the 70s was a monetary phenomenon-for Monetarists it always is; Waldman on the other hand says the GI was caused by a spike in population due to the Baby Boomers reaching the workplace along with women working in much higher numbers along with the 70s drop in productivity.

       As Waldman notes, Mark Sadowski claimed not to be too impressed with this theory yet sure expanded a lot of intellectual firepower trying to disprove it.  

      ’"Ive had a wonderful nemesis and helper the last few days in commenter Mark Sadowski, who challenged me to provide evidence for a demographic effect on inflation in international data. Mostly I made a fool of myself (twice actually, and not unusually). Looking at thegraphs — after Sadowski helped me get them right! — I see support for a relationship between labor force demographics and inflation in the United States, Japan, Canada, and Finland. Italy is a strong counterexample — it disinflated in the middle of its labor boom. The rest you can squint and tell stories about. (I now have 14 graphs now.) Italy notwithstanding, the claim “it’s hard to disinflate when labor force growth is strong” looks more general than “inflation correlates with labor force growth”. Decide for yourself."

     "Sadowski is not much impressed by my demographic view of the Great Inflation. But he paid me the huge compliment of devoting time and his considerable expertise to testing my speculations. He writes

I took your set of eight nations plus the four from my original set of counterexamples that you excluded (West Germany, Ireland, Luxembourg and the Netherlands), combined civilian labor force data from the OECD with CPI from AMECO, and computed 5-year compounded average civilian labor force growth rates and CPI inflation rates. The time periods ran from 1960-65 through 2007-12.
Then I regressed the average CPI inflation rates upon the average labor force growth rates. Five of the twelve were statistically significant, and all at the 1% level. The average civilian labor force growth rate and average CPI inflation rate were positively correlated in the U.S. and Japan, and negatively correlated in Spain, the Netherlands and Luxembourg.
Next I conducted Granger causality tests using the Toda and Yamamato method on the level data over 1960-2012 for the U.S., Japan, Spain, the Netherlands and Luxembourg.
The U.S. data is cointegrated, so although the majority of lag length criteria suggested using only one lag, since Granger causality in both directions was rejected at a length of one, I went to two lags based on the other criteria. The results are that CPI Granger causes civilian labor force at the 10% significance level but civilian labor force does not Granger cause CPI.
In Japan’s case civilian labor force Granger causes CPI at the 1% significance level but CPI does not Granger cause civilian labor force.
Granger causality was rejected in both directions for the other three countries.
In short, out of the 12 countries I looked at, only five have a significant correlation between average civilian labor force growth and average CPI inflation, and only two of five have a positive correlation. Of the five, only the two with positive correlation demonstrate Granger causality. But in the US case the direction of causality is in the opposite direction to that which you predict. Only Japan seems to support the kind of story you are trying to tell.

and follows up

I added your set of seven new nations (Canada, Finland, Greece, Italy, New Zealand, Switzerland and Turkey) plus seven additional nations (Belgium, Denmark, Iceland, Korea, Norway, Poland and Portugal) to the set of 12 that I commented on last time. I did the same analysis as I did last time for this new set of 14, that is I combined civilian labor force data from the OECD with CPI from AMECO, and computed 5-year compounded average civilian labor force growth rates and CPI inflation rates. The time periods ran from 1960-65 through 2007-12 with the exception of Korea which started with 1967-72. I regressed the average CPI inflation rates upon the average labor force growth rates. Ten of the fourteen were statistically significant, and all at the 1% level with the exception of Poland which was at the 10% significance level. The average civilian labor force growth rate and average CPI inflation rate were positively correlated in Canada, Denmark, Finland, Greece, Iceland, Italy, Korea, New Zealand and Norway and negatively correlated in Poland.
Next I conducted Granger causality tests using the Toda and Yamamato method on the level data over 1960-2012 (except for Korea which was over 1967-2012) for the ten countries which had statistically significant correlations.
In Finland, Poland and Korea civilian labor force Granger causes CPI at the 5% significance level but CPI does not Granger cause civilian labor force. In Greece CPI Granger causes civilian labor force at the 1% significance level but civilian labor force does not Granger cause CPI. In Iceland CPI Granger causes civilian labor force at the 10% significance level but civilian labor force does not Granger cause CPI.
So out of the 26 countries I have looked at, fifteen have a significant correlation between average civilian labor force growth and average CPI inflation with eleven of the fifteen having a positive correlation. Of the eleven with positive correlation six demonstrate Granger causality with three showing one way causality from civilian labor force to CPI and three showing one way causality from CPI to civilian labor force. Of the four with negative correlation one demonstrates Granger causality from civilian labor force to CPI.
Only three countries (Japan, Korea and Finland) out of the 26 support the kind of story you are trying to tell.

     "Mostly I am very grateful to Sadowski for his work."

      http://www.interfluidity.com/v2/4706.html

       Now I admit when we start discussing Granger Causality, we're not really in my first language but is this really the most charitable analysis? However, I can understand Waldman just appreciating all this time taken by Sadowski-who is always very impressive with his amount of knowledge-as I feel the same way about Waldman himself linking to me at he top of his page. I told you I'd stop talking about being linked at Delong when something else impressive happened. This is the kind of thing I had in mind.

       "About a week ago, I argued that the Great Inflation of the 1970s was largely a demographic phenomenon. That claim has provoked a lot of debate and rebuttal, in the comment sections of several posts here, and elsewhere in the blogosphere. See Kevin Erdman, Edward Lambert [1, 2, 3], Marcus Nunes [1, 2], Steve Roth [1,2], Mike Sax, Karl Smith [1, 2, 3], Evan Soltas, and Scott Sumner [1,2], as well as a related post by Tyler Cowen. I love the first post by Karl Smith. My title would have been, “Arthur Burns, Genius.”

    Important thing is that I actually garnered a mention at Interfluidity. Luckily Steve is not so overawed by the term Granger Causality as I am. He points out that GC is not the same thing as strict causality. As I suspected Sadowski hasn't necessarily demolished Steve's theory. I get one link and so now I'm calling him Steve. Still I'm sandwiched between Steve Roth and Karl Smith. 

    Kudos to Tom Brown as I didn't realize I was mentioned by Waldman until reading his comment on my cell phone on the way home from work. 

     In any case, here's Steeeeve:

      "Alas, I am not at all dissuaded from my view. At the margin I’m even a bit encouraged. The direction of Granger causality is not very meaningful here. (Granger causality, in the econometric cliché, is not causality at all but a statement about the arrangement of correlations in time. Expectations matter and near-future labor force growth is easy to predict, so there’s no problem if CPI changes can precede labor force changes.) I see some support for my thesis in the significant and usually positive correlations Sadowski observes in many countries. However, much as I am grateful, I don’t take this work as strong evidence either way. Sadowski overflatters my graphical analysis technique by translating it directly to an empirical model. Collapsing growth into overlapping 5-year trailing windows smooths out graphs that would otherwise just look like choppy tall-grass noise. But it creates a lot of autocorrelation unless the data is chunked into nonoverlapping periods. (Sadowski may well have done that! It’s not clear from the write-ups.) More substantively, to generate a good empirical model we’d have to think hard about other influences and controls that should be included. One wouldn’t model inflation as always and everywhere a univariate function of domestic labor force growth."

    He sure has a polite way of saying someone's wrong. He allows that maybe he has been decisively refuted here:

    "Maybe my view has been definitively refuted and I’m just full of derp! You’ll have to judge for yourself. In any case, I thank Sadowski for the work and food for thought, and for help and correction as I made a fool of myself."

     I doubt it. That Sadowski had to expand this much firepower to refute him-assuming that he did at all-means that you can't claim that Steve's argument is 'easily refutable.' Indeed, even Sumner has treated it with respect and Evan Soltas also has a piece about it-further buttressing the idea that it's not a puerile idea. So, yes, I can see why he's not discouraged.  As I have to leave my house right now for the train to begin my commute to Melville I'll have to wait till later to get into Soltas' analysis. A telemarketer's job is never done.