Tuesday, September 17, 2013

Median Incomes

Some new census data is out.  Tyler Cowen has a couple of links.

As usual, there is some basic disaggregation that can provide some interesting details.

Here are the numbers, by age:

For 15 to 54 year olds, the 25 year real change in cash income is negative.  The average is pulled up entirely by 55+ year olds.

However, this mostly reflects changes in labor force participation.  Here is a similar graph of changes in labor force participation:

Generally, age groups with higher labor force participation have seen increased household income and vice versa.  (I have left out the 15-24 age group.  It declined a lot, but most of the decline was among younger workers who would not be heads of households.  Given the well-publicized drop in young persons' labor force participation, the recent behavior of 15-24 year old household income looks relatively normal.)

It is difficult to make this adjustment quantitatively, but adjusting for number of workers would tend to pull all the age groups toward the aggregate median income number, and raise the worker-adjusted aggregate income by a few percentage points, which is clearly still lower than we would prefer.

We can look at household income by number of earners per household:

What is interesting is if we measure these by the relative levels, compared to 1987, we see the phenomenon that appears every now and then Simpson's Paradox, where the aggregate number is lower than any of the disaggregated numbers:

This is because, contrary to conventional wisdom, the average number of earners per household has been declining over the last 30 years, and pretty steeply over the last 15 years:

If we use the weighted average of each earner group (the orange bars above), we add about 8% to the relative income growth since 1987.  Or, if we simply adjust the 2012 amount by the relative change in average number of earners, we would add more than 10% to measured income growth.

Obviously, there are still a lot of caveats here.  This measure doesn't include non-cash income, so it doesn't include food stamps, health insurance, pensions, etc.  These account for a significant, and growing, portion of income.  This makes any definitive comments on "No Earner" households difficult.  And, it might explain the difference between the growth of 1 earner and 2 earner households.  Some 2nd earner cash income might be enlarged through negotiation, since health benefits would not be required.  The 13% in additional growth for 2 earner households probably reflects a portion of the cost of non-cash benefits, although there are clearly frictions that keep individual incomes from completely reflecting this factor.

Most references to these numbers bemoan the lack of growth, especially among low income households.  But, the difficulty for that interpretation is that most of the difference between household income levels is a product of the number of earners.  If households really have been in a troublesome 25 year period with stagnant incomes, we should expect to see more earners per household, or at least a level number.  Families eager to increase cash income would not be shedding workers.

I wonder how much of the income stagnation is a secondary product of increasing non-cash compensation.  Many non-cash benefits would be redundant to a 2nd wage earner.  If labor market frictions prevent cash wages from fully compensating 2nd wage earners for foregone benefits, the increasing value of those benefits may be causing potential 2nd wage earners to remain out of the labor force.

I have shown that, once we account for age and gender, long term labor force trends among the core working ages are very stable, going back decades.  The trend in lower earners per household appears to accelerate downward at about the same time that women's LFPs peak, in the late 1990's.  Possibly this is another place where the increase in female labor force participation was masking the broader trend of smaller households and fewer earners.  When female LFP peaked, this trend became clear.

The issue with non-cash income sounds plausible, but maybe the American labor force has simply reached a stage where we are making lifestyle choices to exchange leisure for compensated work.

This is a place where I think that debunking the "demoralized worker" myth of the declining labor force is important.  The idea that workers who would like to be in the labor force have been kept out by an unfriendly labor market for a period of 10 or 15 years, extending over several business cycles, is implausible.  The declining LFP rate is enough of a nudge to make that idea more plausible.  But, the long term LFP rate is a demographic and cultural issue.  The adjusted trends in the male data go back at least 50 years in surprisingly linear fashion.  The easy answer given by the misinterpreted LFP leads us down the wrong path.

The graph above shows a long term decline in earners per household.  There is a cyclical exchange of about 2% of household members into and out of the labor force, but outside that cyclical pattern, there is no clear long term trend in the number of non-earners - it is pretty flat.  That suggests that Americans are willing to accept the costs of fewer household earners in order to accommodate some cultural tendency to smaller households.

There surely are important trends to understand here.  For starters, households with less than two earners probably wouldn't choose for most of their income gains to be swallowed up by health insurance costs.  In that case, what is causing so many households to proceed with only one or no earners?  Retirement, schooling, divorce, single-parenthood, and other factors must be heavily at play here.  These reflect trends that are a product of being a wealthier society.  I believe that there is a tendency for pundits to protect their political credibility by ignoring these factors and treating these matters as purely matters of involuntary poverty and distributive justice.  I'm afraid that that posture causes us to miss most of what is important, and could lead to misconstrued public policy.

ADDENDUM:  Looks like another candidate for my "It's all demographics" series.  Here is a graph, since 1980, comparing the percentage of households in the 65+ age group and the percentage of households with No Earners:
The portion of households over 65 will probably rise to 30% over the next 15 years.  If this leads to a demographic rise in "No Earner" households to 30%, that will cause another 5% or 6% drag in the stated aggregate median household income measure.

No comments:

Post a Comment