Amazon is not a macro phenomenon

I noticed a new post that pushed back against my claim that slow NGDP growth explains slow wage growth:

Amazon Has Become A Macro Phenomenon

.  .  .  There is no shortage of explanations on why low unemployment isn’t sparking any wage (and subsequent price) inflation, for instance, here are two of the most common explanations:

  • Low productivity growth
  • Low nominal growth and hence low inflation expectations

.  .  .

More convincing is Scott Sumner’s explanation that it is simply the result of low inflationary expectations, which keep wages in check. We see this even more in Japan, where the labor market is arguably considerably tighter but wages simply fail to take off.

The Amazon effect

But there could be another explanation, which centers around the huge size and growing influence of the online (and increasingly offline) behemoth called Amazon (AMZN).

Perhaps like Wal-Mart (WMT) before, Amazon exerts downward pressure on both wages and prices by its sheer scale and efficiency.

There are three problems with this explanation.  First, while rapid productivity growth in retailing might hold down inflation (if the Fed were targeting NGDP), it would not hold down wage growth.  Second, the Fed is targeting inflation, so rapid growth in productivity should boost NGDP growth, not reduce inflation.  And third, productivity growth is actually quite slow.  Although Amazon is an impressive company, it’s just a tiny share of GDP.

The economy is really, really big.  Don’t use micro reasoning to think about macro phenomena.

I keep going back to NGDP.  NGDP (i.e. monetary policy) explains the slow wage growth.  Period.  End of story.



24 Responses to “Amazon is not a macro phenomenon”

  1. Gravatar of bill bill
    7. September 2017 at 11:31

    I think our Fed often gives explanations for the lack of inflation that are only a little bit better.

  2. Gravatar of ssumner ssumner
    7. September 2017 at 12:07

    Bill, Yes.

  3. Gravatar of Randomize Randomize
    7. September 2017 at 13:34

    NGDP does explain it all but I’d quibble with one point:

    It’s not accurate to dismiss the companies’ influences based on their respective shares of the economy as the effects of Amazon and Walmart’s bargaining power are not limited to their own transactions. Their price pressure ripples throughout the entire retail community as every consumer with a phone will price check major purchases online even while at brick-and-mortar stores.

  4. Gravatar of Benjamin Cole Benjamin Cole
    7. September 2017 at 16:21

    It is probably time to retire the “higher productivity leads to higher wages” totem.

    For full-time male US workers, median weekly wages are down since 1979. (Mercifully, this chart does not go back further in time)

    Yet workers produce more than double per hour today what they did in 1979:

    Okay, so output per hour has doubled, yet real pay has fallen.

    Likely this reflects globalization of wages, and perhaps some institutional shifts in who has situational leverage in the U.S.

    But, greater productivity has not resulted in higher wages in the U.S. in the last 40 years…and counting.

  5. Gravatar of Major.Freedom Major.Freedom
    7. September 2017 at 19:50

    NGDP (i.e. monetary policy) explains the slow wage growth. Period. End of story.

    No, not period and not end of story.

    NGDP is the money and spending on final goods and services, not on labor. In fact, expenditures on labor is in direct competition with, meaning mutually exclusive to, expenditures on everything else not labor, including final goods and services of which NGDP is supposed to track according to the various politburos who do this service for governments and their academic apologists, as it certainly isn’t valued in the market.

    Labor income is not a factor in NGDP. They necessarily move in opposite directions, ceteris paribus. If I spend more on labor, that additional money has to come at the expense of reduced spending somewhere else. You cannot spend the same dollar on more than one thing. To be sure, the muddled framework of anti-market monetarism tends to lead people into conflating all sorts of concepts, such as spending on labor and spending on that which makes up GDP. A is A, but sometimes we can believe A “as if it were” B, or some other lazy and incoherent appeal to end the analysis out of being tired.

    You speak of “correlations” between NGDP and wages? Good, there is also a correlation between global warming and pirate attacks. Correlations do not indicate causation. The reason why wages have tended to historically be correlated with NGDP (and even here the correlations have not been entirely persistent) is because both wages and NGDP are affected roughly similarly by the same underlying factors. I bet you would think it reasonable for me to say that if everyone’s wages doubled, that we would likely see NGDP go up. Doesn’t this prove wages drive NGDP? Is that why there is an apology for wage targeting as well? Because correlations cannot parse out what is causing what, so pick one and hope for the best?

    Or perhaps the final backstop is the “hot potato” theory? This is a flawed theory because it is putting the cart before the horse. Money doesn’t become “hot” in this way until an actual increased supply of money is brought about and spent, but that leaves unanswered the question of why the initial receivers of said money spent it in the first place before money become hot on account of THEIR spending first. And they must spend first. The member banks control what they spend. If they don’t “trade” more, through increased lending or purchases or investment, then all the increased reserves from the counterfeiters just doesn’t translate into increasing the temperature of money. Money can go cold here. Having said that, historically and empirically we see banks lending more when reserves have or are expected to increase, and this increased lending and subsequent spending and respending of money throughout the economic system is what then makes money hotter to hold. The initial drive to inflation was not hot potato, and everything after is not hot potato either. It is an adjustment in people’s valuations of money relative to capital, goods and services.

  6. Gravatar of Major.Freedom Major.Freedom
    7. September 2017 at 19:51

    Vote fraud in New Hampshire 2016. Decided the state for democrats:

  7. Gravatar of ssumner ssumner
    7. September 2017 at 21:24

    Ben, As I’ve said, I don’t agree with your data.

  8. Gravatar of ssumner ssumner
    7. September 2017 at 21:25

    Randomize, That should imply falling profits. But corporate profits are strong.

  9. Gravatar of JMCSF JMCSF
    7. September 2017 at 22:07

    The Amazon effect will be real to whichever city is chosen for their 5billion, 50k employee HQ2. Although it will likely be higher wages for that metro area.

    I wonder where they will ultimately decide (which city/state will give them the best offer). Which would you like to see?

  10. Gravatar of TravisV TravisV
    8. September 2017 at 07:45

    Tyler Cowen: “Sticky wages are part of the macro labor market story, but not at the center”

  11. Gravatar of Randomize Randomize
    8. September 2017 at 09:19

    Dr. Sumner,

    It would imply slimmer margins. They could (and I suspect they do) make it up on volume.


    I too disagree with your presentation of the data. First, the CPI that’s used in the “real” wages computation is generally considered to be inflated. If you use the Fred system to graph nominal wages / PCE index (either with or without food and energy), you’ll see a different story.

    You’re right that’s it time to stop talking about productivity and wages as if they’re linked. In the manufacturing world, for example, imagine that the boss buys a new machine for the factory. Productivity jumps as the new machine will make more widgets than the old machine but the employees who operate it are not necessarily doing anything that requires new skills. In this case, the boss would rightfully reap the benefits of that investment and wages would stay the same.

  12. Gravatar of ChrisA ChrisA
    8. September 2017 at 10:17

    Why is the idea that people can reduce their expected wages to get a job such a difficult idea for people to accept? Low unemployment could reflect high demand for labor, or it could reflect people pricing themselves into jobs. Looked at this way, low wage growth coupled with low unemployment is no mystery. It’s even consistent with lower labor force participation – there are quite a few people (say retirees or people on various benefits) who don’t have to lower their wage expectation so they don’t and stay out of the labor market.

    Increase overall demand for labor and I would expect wages to rise more rapidly, as happens often in recoveries from recessions. The question is, is demand low for structural or monetary reasons? If structural maybe it is really the Amazon effect after-all, not be a competitive effect, but by improvements in labor efficiency for delivery logistics. Of course it is not just Amazon but many companies all doing Amazon like things to logistics.

  13. Gravatar of Matthew McOsker Matthew McOsker
    8. September 2017 at 10:44

    I agree weak NGDP growth is a major source of economic problems.

    But, higher NGDP growth does not guarantee an unwanted level of inflation.

    Be careful on corporate profits, foreign trade has been a good propellant over the years. I have not looked at 2017, and not sure what the long term correlation has been to US NGDP growth.

    This is just the S&P 500 which of course excludes private cos. But gives one an idea.

  14. Gravatar of Major.Freedom Major.Freedom
    8. September 2017 at 19:25

    Julian Assange, who has a 100% accuracy track record in what he publishes, which is better than ANY source referenced by Sumner on this blog, has admitted:

    “We can say, we have said, repeatedly that over the last two months that our source is not the Russian government and it is not a state party”

    Got that? NOT RUSSIA.

    He also implied in a BBC interview (about the DNC emails Wikileaks published) that his source was a young man who was murdered for leaking the emails to Wikileaks. The interviewer picked up on the implication.


  15. Gravatar of ssumner ssumner
    8. September 2017 at 19:49

    Randomize, But you were talking about the entire sector.

  16. Gravatar of Benjamin Cole Benjamin Cole
    8. September 2017 at 21:46


    The data is from FRED, as cited. Not my data!

    Even if wages have drifted a little higher than expressed by real median weekly wages of full-time employed males (due to measurement issues), still wages are long way from doubling, as has output per hour since 1979.

    You can pick some other series if you like. Just cite it, and let us review.

    But the Big Picture is that the “productivity leads to wage gains” story is dead for the last 40 years.

  17. Gravatar of Scott Sumner Scott Sumner
    9. September 2017 at 07:11

    Ben, You simply need to read some of my older posts:

    Both profits and wages are a stable share of national income over the past 50 years.

  18. Gravatar of Major.Freedom Major.Freedom
    9. September 2017 at 11:39

    “Both profits and wages are a stable share of national income over the past 50 years.”

    No they aren’t.

    Wages as a percent of GDP is DECREASING over the last 50 years. Indeed, Wages/GDP was stable up until the early 1970s, after which the share has declined.

    Gee, wonder what big event about money and spending occurred during the early 1970s that might explain this. Hmmm, nope, won’t see it on this blog. There’s an agenda to push!

    Even if the share was stable, which of course it isn’t, but assuming for the sake of argument it was, to suggest this relationship will continue is what is called “naive extrapolation”.

    In economics, where subjective valuations reign supreme, past events are not necessarily indicative of any law of what choices will be made in the future. You are not Nostradamus. The method of physics does not work in economics, because the method of physics assumes causal relations between variables is time invariant.

    Any attempt to impose a “system” by anti-market activity, will destroy any supposed causal connection anyway, which of course there isn’t as mentioned.

    Anti-market monetarism fails to account for Lucas’ critique:

    “It is naive to try to predict the effects of a change in economic policy entirely on the basis of relationships observed in historical data, especially highly aggregated historical data.”

    It does not follow from wages being historically X% of NGDP, that counterfeiters changing their “policy” to NGDP % when hitting CTRL-P will automatically transition wage payment decisions away from employers and employees, to the counterfeiters, such that the counterfeiters can maintain X% when changing NGDP. It doesn’t work that way. If it did, then communism would have worked the world over already and we’d all be wealthy socialist citizens by now. The failure is not a failure of number estimates. The failure is assuming such a number can be known as superior than market driven NGDP.

    NGDP and wages are mutually exclusive types of expenditures. They move in opposite directions ceteris paribus. Any historical regularity between the two variables is brought about by the two variables themselves being affected by independent factors that make both wages and NGDP move co-extensively. It does not follow that people spending more money on milkshakes and cars that wages will go up. If all else is equal, the more people splurge on their own consumption, the less investment in labor is made and the lower wages will become, not the higher.

    John Stuart Mill wrote almost 140 years ago:

    “Demand for commodities is not demand for labor.”

    F.A. Hayek wrote of this truth:

    “John Stuart Mill’s profound insight that demand for commodities is not demand for labor, which Leslie Stephen could in 1878 still describe as the doctrine whose “complete apprehension is, perhaps, the best test of a sound economist,” remained for Keynes an incomprehensible absurdity (Collected Works, vol. 9., p. 249).”

    Will this blog reach beyond 19th century superstitutions and understand this so it no longer appears as incomprehensible absurdity?


    In Russia news:

    Hillary Clinton advisor (still on her payroll) insinuates Bernie Sanders is being paid off by Russians:

    Who else is being paid off by the Russians you may ask? Anyone who opposes the corrupt Clintons. Sumner, good thing you have only made churlish ad hominem posts regarding Clinton, or else you would be accused of colluding with the Russians as well, bahahaha

  19. Gravatar of Major.Freedom Major.Freedom
    9. September 2017 at 11:41

    Voter fraud in NH likely determined election for Clinton

  20. Gravatar of Major.Freedom Major.Freedom
    9. September 2017 at 11:42

    Trump likely won the popular vote if all the fraudulent votes are removed

  21. Gravatar of Benjamin Cole Benjamin Cole
    9. September 2017 at 21:58

    Scott Sumner:

    I think you and Randomize have a point—what is the best deflator to use, CPI or PCE? Or another measure? The CPI does run a higher rate of inflation. (Just remember, if you use PCE, then the top annual rate of inflation in the Arthur Burns era was…8%. That’s it? Yes.)

    On the labor share of national income issue (which I was not addressing) I sense one can get into “statistics wars” or “dueling models” without result (for objective bystanders anyway).

    For example, here is post from the BLS, and it says labor share has dropped off a lot since 2001:

    Okay, this so this BLS fellow uses a different methodology than you use.

    Kevin Erdmann suspects much of labor’s share of declining income is really just rising rents,which get back to ubiquitous property zoning and artificial scarcity (or housing “shortages.”).

    In any event, I think the looking at median wages, even if you choose to adjust by the PCE rather than the CPI, will find that wages are not keeping up with productivity, nor should they, for reasons Randomize mentions. The argument that higher productivity necessarily leads to higher wages is hollow.

    Living standards are something else, I hope higher productivity can help there.

    As I am sure you know, median household income stats are interesting, but have huge flaws, such as a rise in single-parent households (think divorces or out-of-wedlock births) could depress the median, while a rise in two-income (mom and dad both work) households could hike the median. An increase in retirees could decrease the median etc.

    The stereotype joke in Los Angeles in the 1990s-2010 era was “I bought the house next to my parent’s house! They bought theirs when dad worked 9-5 and mom was mom. I bought mine with my spouse and we are professionals who work 60 hours a week each. We may have one child someday.”

    Does anyone think their son or daughter could troop off to college, and graduate and start fresh in a NYC, Boston, L.A., Bay Area and make it on their own? Buy a house and raise a family? You could 40 years ago, especially so 60 years ago. Not today.

  22. Gravatar of Scott Sumner Scott Sumner
    11. September 2017 at 22:38

    Ben, Wages are keeping up with productivity. Read Matt Rognlie’s comments in the post I linked to.

  23. Gravatar of Jose Jose
    14. September 2017 at 10:34

    What about another micro factor that may impact macro variables: isn’t it possible that online tools and apps that target HR and labor placement are making the labor market more efficient and therefore the NAIRU may have fallen significantly because of that?

  24. Gravatar of ssumner ssumner
    16. September 2017 at 08:57

    Jose, Possible, but I’d like to see some evidence.

Leave a Reply