Shifting Mix is Often Ignored as the Reason Behind A Shifting Mean

I have written about this mix effect many times, eg here.  Imagine a corporate division that sells tables and chairs.  The CEO is reviewing this division's performance, and sees that their revenues are increasing but their profit margin is falling.  He asks his analyst to look into it - is it the tables or the chairs or both that are showing falling margins.  Our poor harassed analyst comes back and says, uh, neither.  The profit margins for both tables and chairs went up last year.  Well, the CEO asks, if revenues are up and all their component margins are going up, how is their total margin falling?  It turns out that tables make a much higher margin than chairs, and over the last year the company has seen a much higher growth in chair sales than table sales.  The mix is shifting towards a lower margin product and is bringing the averages down.  By the way, I can say with authority that this conversation is much harder when the analyst is yours truly and the CEO is famed tough (but talented) boss Chuck Knight of Emerson Electric.

Whether the media mentions this effect or not, it is happening all the time.  Here is an example from the WSJ:

One mystery of this economic expansion is that wage growth has remained slow even as the labor market has finally tightened. One widely cited culprit is historically low productivity growth. But a new analysis from the Federal Reserve Bank of San Francisco adds a more optimistic, albeit paradoxical, explanation.

The Bureau of Labor Statistics recently reported that median weekly earnings had risen in July by a healthy 4.2% on an annual basis, the fastest growth in a decade. As labor markets tighten, employers typically increase wages. Until this past year, however, median weekly earnings growth had hovered near 2%, which is significantly less than the 3.25% average from 1983 to 2015.

So why haven’t wages risen faster amid an increase in hiring and unfilled jobs? One answer is that wages have actually been growing at a faster clip—around 4% to 5%—at least for full-time workers with steady jobs. But new full-time workers who are generally paid less than the retirees they replace are dragging down the average wage increase.

Researchers at the San Francisco Fed this week updated their 2016 paper that disaggregated the wages of full-time workers with steady employment from recent entrants—that is, new workers or those returning to full-time work. Their earlier analysis showed that average wage growth had slowed less than expected during the recession while staying relatively flat during the recovery.

That’s because workers who lost jobs during the recession were generally lower skilled and lower paid, so average weekly wages didn’t fall significantly. However, many of those workers have since been rehired at below-average wages, which has depressed the aggregate.

In prior expansions, wage growth has been driven mostly by continuously full-time employed workers, and the researchers find that’s still the case. Wage growth for these workers is now close to the pre-recession 2007 peak. But there are now many more workers who have been on the labor-force sidelines who are moving to full-time employment, thus creating a drag on wages.

This is frequently how mix shifts play out in the news.  Notice that there are actually two pieces of good news here:  1.  Wages for full-time workers who have been employed for a while are growing well and 2.  lower-skilled and less experienced workers who left the labor force are now getting jobs and returning to work.  However, when these are combined, the net is portrayed as bad news, ie wage growth in the US is sluggish.  Because the mix was ignored.


  1. bloke in france:


  2. paul:

    If an above median wage employee in Mexico moves to the US for a better paying job, but one below the median wage, then the median wage in both countries goes down but the person is better off. Nice paradox.

  3. cc:

    Some other aspects of the mix problem:
    Labor participation worries people. However, if you include people who have retired as "not participating", then generous pensions plus wealth that have allowed more people to retire early causes labor participation to go down.
    Likewise, if people are better off, this may cause more wives to be able to stay home (not forced to), thus reducing female labor participation in particular. Also, when families are better off they may decide to take their teenagers on vacation or have them take classes rather than getting a summer job--teenage "unemployment" goes up. I have noticed it being increasingly hard to find teens to do yard work--but that is just my experience.
    The US age distribution is never uniform. A bulge in young or older workers will distort average wages.
    A bulge in young people will cause home ownership % to fall since they are not ready to buy. This is not a "problem" that the gov needs to fix.
    A bulge in retirees will cause average wages to fall since they are living on pensions--but also are not paying for kids or trying to save for retirement. $70,000 annual income for a retiree is much more than for a 45 yr old with kids in college.
    All of these distribution effects tend to be ignored by the press and may falsely be claimed to be a "problem".

  4. antognini:

    In statistics this is known as Simpson's paradox. The most famous example was in a lawsuit alleging gender bias in graduate admissions at UC Berkeley. Overall, women had a lower acceptance rate to graduate programs than men. But nearly every department on its own actually had higher admission rates for women than men. It turned out that women applied to more competitive programs in greater numbers, so although they had higher admission rates to those particular programs than men, they had lower admission rates overall.

  5. stan:

    Point taken. However, using the unemployment rate to conclude that labor markets are tight is foolish.

  6. TruthisaPeskyThing:

    I have heard your line of argument used a lot to down play problems in unemployment and underemployment in the last 8 years. I think the analysis you present assumes a lot without data to back up your wishful assumptions. For example, your cases of generous pension plans allowing earlier retirement and women deciding to stay home. Yes, those movements would lower participation rates for "admirable and positive" reasons. However, such conjecture is more wishful thinking than reality. The labor participation rates for those over 55 -- and especially those over 62 -- have soared in the last 8 years. The plummeting labor participation rates is coming in the ages 25-54. Also, women with children are not decreasing their labor participation rates. Therefore, we have had PROBLEMS in the last 8 years.
    You are right in regards to average wages -- and that is a point that Coyote was making. As a bulge in workers retire, wage earnings will decrease because remaining workers will have less experience and command lower wages. However, that is not nearly the entire story. We have issues because lower educated people are now competing in a global market, and their wages are not keeping up with inflation. We seem to have a huge chunk of the population that apparently thinks that working at McDonald's should be sufficient to take care of a family.
    By the way, finance and tax experts have become much more sophisticated over the past few decades. The more educated workers often find ways to take income not in the form of wages -- so that depresses the reported level of wages in the U.S.