Beckworth interviews Ozimek

David Beckworth recently interviewed Adam Ozimek for his podcast series, and the discussion focused on monetary policy.  Ozimek pointed out that not enough attention was being paid to the lessons to be learned from the past three years of Fed policy.  The Fed began raising rates in December 2015, and even Fed officials now admit that this was too soon.  Or do they?  That’s one of the issues they discussed.

Ozimek points to interviews with some top Fed policymakers who seem to agree that the natural rate of unemployment is considerably lower than what the Fed estimated back in 2015, and also that the stance of monetary policy back then was less accommodative than the Fed had assumed.  Ozimek wants them to draw the obvious implication from that fact—that policy was too contractionary in late 2015.  There’s really no other plausible interpretation of the statements he cites, but Fed officials don’t quite seem to come out and explicitly make that admission.  This relates to my recent Mercatus policy brief, where I call on Congress to insist that the Fed periodically evaluate previous policy decisions, based on incoming data.

I also tended to view Fed policy as being slightly too expansionary contractionary during 2015-16, although mostly based on their undershooting of the 2% inflation target.  Like the Fed, I underestimated the extent of the decline in the natural rate of unemployment (Ozimek and Beckworth were ahead of me on that point.)  So what implications can we draw from this episode?

1. I believe it’s important to put this policy error in perspective.  It was a consequential error, perhaps costing hundreds of thousands of jobs for a couple of years.  That’s hardly trivial.  At the same time, the error was an order of magnitude less damaging and an order of magnitude less inexcusable that the policy errors of 2008-15.  When figuring out what went wrong with monetary policy, we need to focus almost all our attention on the mistakes of 2008-15.  That’s the elephant in the room.

2.  This episode illustrates the danger of basing Fed policy on estimates of the natural rate of unemployment (Un).  The Fed cannot accurately estimate the natural rate in real time, and hence it’s an exceedingly poor guide to policymakers.  The lesson is not “next time do a better job estimating Un”, it’s “stop trying to estimate Un, and start focusing on variables than you can measure, like NGDP.”  Under NGDP level targeting, there is no need to even measure the unemployment rate—it plays no role in monetary policymaking.

Unfortunately, the Fed is forced to rely on natural rate estimates as long as it targets inflation under a dual mandate approach, as the unemployment rate helps it to determine how much “catch-up” they need to do after a policy miss.  One of the advantages of NGDPLT is that the Fed is no longer forced to try to do the impossible—estimate Un.

They also discussed some research Ozimek did on the relationship between population growth and inflation.  I have not read the research, but if I’m not mistaken the study found a positive correlation between growth in working age population and inflation, both cross sectionally and over time.  Ozimek hypothesizes that this may relate to the impact of population growth on real estate prices.

The cross sectional correlation makes sense to me.  Cities that are growing fast tend to have higher real estate prices than cities shrinking in population, such as Detroit.  The time series correlation is less intuitive. Inflation is determined by monetary policy.  It seems unlikely that the optimal monetary policy calls for higher inflation when population growth is more rapid.  So what’s going on?  My best guess is that the correlation somehow relates to the link between population growth and interest rates, or perhaps population growth and the natural rate of unemployment:

1.  Under the gold standard, price levels were positively correlated with nominal interest rates.  That’s because higher nominal interest rates boosted the opportunity cost of holding gold, which led to less demand for gold, which is inflationary under a gold standard.  Because the expected rate of inflation is roughly zero under a gold standard, anything that boosted the real interest rate, such as faster population growth, also boosted the nominal interest rate, and hence inflation.

2.  In Japan, shocks that reduce the real interest rate seem to be deflationary, as they reduce the opportunity cost of holding yen, thus boosting the demand for yen and the value of yen.  Lower population growth might reduce the real interest rate in Japan.  Of course the BOJ should offset this, but they often do not do so.  Interestingly, they seem to have done better since 2013, enough so that the inflation rate in Japan has risen slightly, even as growth in the working age population has fallen sharply.  Over longer periods of time, however, the inflation/population growth correlation in Japan supports Ozimek’s claim.

3.  Baby boomers started entering the labor force in the late 1960s.  By itself, that’s not inflationary at all, even if it boosted real estate prices.  Recall that real estate prices are a relative price, and relative price increases are not inflationary unless they reduce aggregate supply.  But faster population growth does not reduce AS.  So what explains the Great Inflation?  One factor may have been a misinterpretation of the Phillips Curve relationship.  The faster growth in the labor force (especially among the young and women), led to a rise in the natural rate of unemployment during the 1970s.  At the time, the Fed made exactly the opposite mistake as in 2015—they underestimated Un.  This caused monetary policy to be too expansionary, in a futile attempt at holding down the unemployment rate.  Just one more reason not to use monetary policy to target unemployment.

4.  Reverse causation.  If monetary policy is procyclical then inflation will also be procyclical.  In that case, rising inflation will be associated with growing RGDP.  It will also be associated with a rising population, as the boom draws in immigrants from places like Mexico.

Note that all four of these explanations are entirely ad hoc, so I wouldn’t necessarily expect the correlation between population growth and inflation to hold in the future, at least at the national level.  Basically any correlation one finds is evidence of sub-optimal monetary policy, just as the Phillips Curve relationship is evidence of suboptimal monetary policy.



20 Responses to “Beckworth interviews Ozimek”

  1. Gravatar of James Alexander James Alexander
    20. November 2018 at 13:39

    Other things being equal a higher population growth will lead to higher RGDP growth. If you also want stable NGDP growth inflation will have to go lower. Is that a good thing?

  2. Gravatar of Brian Donohue Brian Donohue
    20. November 2018 at 14:31

    The more obvious data item is that something like 9-10 million people had jobs in January 2008 who did not have jobs by February 2010.

    This is so far off the charts compared to any post-WWII experience that we shouldn’t need to ponder the natural rate of unemployment. It’s fair to assume that most of these people who had actual jobs would again like to have actual jobs.

    The narrative was that “these jobs aren’t coming back”. Kind of a middle finger to the people who lost their jobs, most of whom actually have come back eventually, to the surprise of economists. Oops.

  3. Gravatar of Matthew Waters Matthew Waters
    20. November 2018 at 15:48

    Population growth vs inflation is an interesting finding. Population growth did not reduce AS, but AD increased more than AS. Housing has an inelastic component in its supply: land.

    The US also exhibits a long bias against denser new construction. Consumers may care less about land than the housing on the land. So the supply-side issue is more than usual market dynamics.

    Supply-side issues then create big issues for *real* wage stickiness. The “natural rate of unemployment” in the 70’s may have been real wage stickiness similar to nominal wage stickiness in 2008-2013.

  4. Gravatar of Matthew Waters Matthew Waters
    20. November 2018 at 17:09

    BTW, this paper provides best breakdown of workforce changes for 1969-82.

    In 1970-75, Ages 16-24 contributed 40% of the unemployment increase. Of the 40%, 30% was composition effect and 10% was increased unemployment. There was a smaller composition effect for women ages 25-54.

    For 1975-80, overall unemployment decreased by 0.8 percentage points. Ages 16-24 contributed 65% of the decrease, while men Ages 25-54 contributed -30%. In other words, their unemployment went up 1975-80. Perhaps the Ages 25-54 had a composition effect within the cohort.

    However, while the composition effects were real, real unemployment also increased substantially for 1970-80. Blue-collar occupation predominantly accounted for the change. Demand moved from manufacturing while workers still expected wage increases.

  5. Gravatar of Benjamin Cole Benjamin Cole
    20. November 2018 at 20:00

    Some have posited that changes in US immigration policy decreased immigration after 2006, and thus changed the outlook for housing, resulting in house-price declines.

    The reverse of the above sentiment is that immigration pre-2006 was one or perhaps the prime cause of a huge house price run-up in the US—rising demand for a relatively fixed supply (due to property zoning).

    There are physical immigrants, but then there are also financial immigrants, those who buy housing in the US as a second home. Both physical and financial immigrants express demand for housing through money—they buy or rent.

    Given the large and chronic trade deficits that define the US trade scene, one might guess financial immigrants play a role in housing inflation as well.

    The above Fed study concludes nations with current-account trade deficits have house price booms. Incoming capital seeks a home (pun, haha).

    As housing feeds into the PCE core, and as the supply of housing in the US is restricted, these are notable observations regarding physical and financial immigration and current-account trade deficits.

    The Fed appears very concerned about “tight” labor markets, repeated dutifully and completely in every Beige Book.

    I will exaggerate to make a point: The Fed is fighting inflation, which is caused in large part by property zoning and physical and financial immigration, by beating up on domestic labor.

    Kevin Erdmann has run many charts showing CPI core ex housing at 1% or so.

    The Fed is way under its 2% IT, except for the effects of housing. So time to undercut labor, right?

    Paul Krugman has said the concept of free trade and immigration is sacralized among orthodox Western economists. So has David Glasner.

    But when theory becomes theology…well, honest assessments become difficult. What happens when free-sailing theologies founder on the rocks of huge structural impediments?

  6. Gravatar of ssumner ssumner
    20. November 2018 at 23:10

    James, I favor stable growth in NGDP per capita.

    Brian, In fairness, I think most economists understood that those unemployed workers in 2010 wanted jobs.

    Matthew, Population growth only increases inflation if the Fed wants it to increase inflation.

  7. Gravatar of Matthew Waters Matthew Waters
    21. November 2018 at 00:19

    “Population growth only increases inflation if the Fed wants it to increase inflation.”

    I should say inelastic housing supply makes inflation closer to NGDP.

  8. Gravatar of Matthew Waters Matthew Waters
    21. November 2018 at 00:24

    Bah, that last comment was wrong if inflation is more than NGDP. Inelastic housing/land increases inflation relative NGDP, i.e. decreases RGDP.

  9. Gravatar of Michael Sandifer Michael Sandifer
    21. November 2018 at 03:03

    Interestingly, Jason Smith has posited that there might be a causal relationship between the increase in women entering the workforce and higher inflation.

    From the market monetarist perspective, perhaps some labor market shocks helped cause the Fed to underestimate inflation risks.

  10. Gravatar of Matthew McOsker Matthew McOsker
    21. November 2018 at 03:38

    Two items.

    One, can we just kill NAIRU, and this bolsters Scott’s preference for NGDPT.

    Two, there is another way to think about population growth. The growth in world labor supply and manufacturing capacity. The latter probably effects inflation, because if labor and/or capacity is short here, the it can be obtained abroad.

  11. Gravatar of Adam Ozimek Adam Ozimek
    21. November 2018 at 05:13

    Hey Scott, here’s a link to the paper if you’re interested:

    I certainly agree so far the argument is under-theorized from a macro perspective. You note the case of Japan, in the paper I cite some research that used Japanese prefecture data similarly to how I used US metro data and found similar results.



  12. Gravatar of Iskander Iskander
    21. November 2018 at 05:45

    Isn’t population growth deflationary in that (for a given level of NGDP) it raises output?

    Now I think its really strange that people blame Japan’s aging population for its deflation, since retirement lowers aggregate supply, we should expect inflation. Of course this depends on the actions of the CB, which were deflationary.

  13. Gravatar of Michael Rulle Michael Rulle
    21. November 2018 at 07:12

    I did not realize some economists think that the entire field of Monetary Policy is akin to Astrology (Ed Yardeni, for example). I am convinced he is not entirely serious—-but certain points are interesting. He seems to have Fed critiques I see you have too. He believes the Fed is tightening too fast. He questions the usefulness of the unseen and constantly changing natural rate of interest in setting policy.

    De Facto, he agrees with this essay in that he thinks they should slow down the hike and assess what has been accomplished before going further. Then he states Powell is a “pragmatist”, I.e. let’s get to 3.5% as fast as we can (well as steady as we can) so we can drop it if we need to! We have seen that argument before, which seems hard to believe. Also, the tightening labor market is called into question. It is unclear to me why we measure the unemployment rate when what we really want to know is the employment rate.

    The reason I read you is at bottom I like the idea of NGDP targeting. So I One reason I like it is you can measure it, and relies less on unknowables. I still have a problem with it’s implementation. Not the futures idea, but what actions should the Fed take to put it back in line. It seems they still need to rely on unknowables. As an aside, I still find it absolutely unbelievable that the Fed bought 4.5 trillion of treasures and mortgages. It simply trained the market to not worry about the importance of free markets. It’s like welfare, or inherited money. One learns to be unproductive.

  14. Gravatar of bill bill
    21. November 2018 at 14:40

    I found this confusing. Wouldn’t undershooting the target mean the policy was too contractionary?

    “I also tended to view Fed policy as being slightly too expansionary during 2015-16, although mostly based on their undershooting of the 2% inflation target.”

  15. Gravatar of ssumner ssumner
    21. November 2018 at 16:10

    Matthew, Yes, inelastic housing supply reduces RGDP, but population growth does not.

    Iskandar, Yes, it’s deflationary for any given NGDP. This is why I say the effect of population growth depends on the Fed’s reaction.

    bill. That was a typo—I meant contractionary.

  16. Gravatar of bill bill
    21. November 2018 at 17:32

    I hang on your every word. ;- )

  17. Gravatar of Matthew Waters Matthew Waters
    21. November 2018 at 23:17

    As I thought about it, population growth can have either positive or negative economies of scale. I wouldn’t flatly say population growth does not have an effect on RGDP per person.

    Looking at the paper, Ozimek first found strong cross-sectional correlations. Downward population growth has twice the coefficient of upward population growth, which makes sense. The buildings remain built for a long time.

    The very long run correlation was based on 17 different metro areas with data sets going back to early-1900’s. Population growth predicted higher inflation over this time. All 17 metro areas used the dollar. Certainly housing restrictions played a role, but I believe housing has a negative economy of scale regardless.

    The correlation did leave out the higher productivity through specialization in more populous areas. So while RGDP per person is dragged down by housing costs, specialization could more than offset that effect.

  18. Gravatar of ssumner ssumner
    22. November 2018 at 08:23

    Bill, I’m glad someone is paying attention!

    Matthew, Keep in mind that population growth has very little impact on GDP/person–not enough to have a noticeable impact on inflation.

  19. Gravatar of ssumner ssumner
    22. November 2018 at 09:01

    Adam, Thanks for the paper. I certainly agree that there is good theory explaining the cross sectional results, and that real estate plays a role at that level.

    Note that regional price level differences are a real phenomenon, whereas aggregate inflation is a nominal phenomenon.

  20. Gravatar of Petja Ylitalo Petja Ylitalo
    22. November 2018 at 12:05

    It seems to me that all this is explained by central banks being “careful” in their actions, rather doing too little than too much.

    This might be related to humans usually using copenhagen interpretation of ethics: the more you are involved with something, the more you are to blame if something goes wrong with it.

Leave a Reply