Is decoupling real?

Since the 1970s, income growth for middle-class American households has become decoupled from growth of the economy. The chart below offers one way to see this. It shows trends in GDP per capita and median family income, with each series displayed as an index set to equal 1 in the initial year. From the late 1940s through the mid-to-late 1970s, the two moved in lockstep. After that, GDP per capita continued its steady upward march (through 2007), but median income rose much less rapidly.

This is disappointing, but seemingly not surprising. After all, income inequality increased sharply during these years. The share of income going to the top 1% of households jumped from 8% in 1979 to 17% in 2007. With a larger and larger portion of economic growth going to those at the top, a divorce between growth of the economy and growth of middle-class incomes is exactly what we would expect to see.

But according to some (here, here, here, here), this picture may significantly overstate the degree of decoupling.

One objection is that the price deflator typically used to adjust GDP per capita for inflation differs from the deflator used for median family income. I’ve addressed that here by using the same deflator for both.

A second concern has to do with GDP per capita as an indicator of economic advance. Since the 1970s a larger portion of GDP has gone to replace old capital equipment and therefore can’t go to household income. Also, the number of persons has increased less rapidly than the number of households, so a per capita (per person) measure of GDP could mislead.

A third worry is that the income measure used to calculate median family income is too thin. If a growing portion of GDP has gone to employer benefits, that would help middle-class households, but it wouldn’t show up in these income data.

To address these second and third concerns, we can turn to a more encompassing measure of household income. The data are from the Congressional Budget Office (CBO). The measure includes all sources of cash income. It adds in-kind income (employer-paid health insurance premiums, food stamps, Medicare and Medicaid benefits), employee contributions to 401(k) retirement plans, and employer-paid payroll taxes. Tax payments are subtracted.

We can use average household income in these data as a substitute for GDP per capita. The CBO data set doesn’t tell us the median income, but it provides something quite similar: the average income of households in the middle quintile of the distribution (from the 40th percentile to the 60th). The following chart adds these two series. The story is virtually identical.

Decoupling is real and sizable.

35 thoughts on “Is decoupling real?

  1. The average household income line is interesting. The gap between the average and the median implies that there are big outliers on the high side.
    Essentially the average includes all incomes, particularly the top 1% ones, while the median is only the middle of the pack incomes.
    You should elaborate on “We can use average household income in these data as a substitute for GDP per capita.”, that’s an important point.

  2. It would also be interesting to see top 1% household income on this graph.

  3. Yes, definitely overlay both top quintile average income and bottom quintile average income on the second graph. That will complete the picture and demonstrate what is happening.

  4. I believe the premise, decoupling is real. Next question Why has it occurred? Jacob Hacker and Paul Pierson present interesting, in depth and compelling explanation over the past 40 years. Do you think they are correct in their explanation(s)? Or are there other, better explanatory variables?

  5. but but but
    what about household size change ?
    did I miss how you addressed that criticism ?
    to put it another way, given that the media are rife with stories about how single person households are growing, can you do a sensitivity test or something ?

  6. Another manifestation of decoupling is that the income distribution (AGI) is exponential (with a time-independent average) up to about $150k and Paretto after that.

    Given that most people make <$150k it seems the US has turned into a trap as far as economic mobility goes. And that is what the special interests wanted all along.

  7. @ezra abrams
    That is why the average household income line is important. It tracks GDP which implies that changes in the make up of households are negligible.
    Take the single person household example, if that were really the cause then there would be a spike in the number of households and that would drag the average down (same money spread across more houses).
    Instead the average line shows that money per household is still tracking GDP, but the median line shows that most of it is being soaked up somewhere in the top half of society.

  8. Assignable cause George Mason University 1972
    ALEC 1973
    Heritage Foundation 1973
    CATO 1977
    Ronald Reagan 1981 – 1989
    This series of events occurred at the right time to start and create the bifurcation of income A corporatist form of a union effort

  9. ~ 1920 to 1945 the trend went backwards from 2012 to 1950 because of unions & the view of shared responsabilty

    WWII was the pinch point ~ an hour glass shape with a neck length between 1945 to 1981

  10. Is there a reason why you use household for the average line and family for the median? These concepts matter as median household income has risen more than median family income due to increases in cohabitation, etc. So the graph would be more meaningful if consistent series were used.

  11. Does the CBO measure of income include things like employer-paid “business meals,” company cars, and other in-kind benefits that, unlike 401(k) contributions, health insurance, and the others that you list, never show up on a payroll statement or W2?

    We should expect all of those to have decreased since 1980 with lower marginal tax rates, as workers would have preferred not to get those cash substitutes.

  12. Business entertainment, like meals, are also less deductible than it used to be.

  13. Very interesting that the most dramatic change from the solid black line to the solid red line is in the post 2000 era. Does that demonstrate that wage stagnation there was because of increased cost of benefits, but that was not the case for the earlier years?

    Some of this would be expected because of the increase in education. Someone who takes off 3-6 years for professional school or graduate school earns a lower income while in school but a higher income afterwards, skewing the distribution a bit. Couple that with increasing returns to education, and some of this is not that surprising.

    Also a fair amount of measured GDP increase is because of the increased tendency of upper middle class and upper class women to work outside the home. By definition, those benefits accrue to them.

  14. I find it noteworthy that counting benefits has so little effect from 1978 to 2000, but then a strong effect starting in 2000.

  15. Jeff: Household is a better unit. I use family for the median in order to go back farther in time; family income data from the Census (CPS) begin in 1947, household income data in 1967. But the trend would look no different had I used median household rather than median family.

  16. So, this would seem to convincingly imply the Golden Rule- ‘He/She with the Gold, makes the rules’ Or some variant thereof. While we may not be headed toward Serfdom, it does seem the lines do NOT portend the American Dream fable for many of us. Despite what one hears so often on the national office campaign trail.

    Any predictions of societal impact if the lines continue on their present course for the next 10, 20, 50 years?

    Dr. Kenworthy is after all, in the Sociology dept, which I think deals with Society in the broadest sense. . .

  17. I came here from Marginal Revolution where there is a post linked to here. Very interesting and obviously a lot of thought went into it. Thank you.

    I think you are still missing an important component or two that accounts a lot for the gap. It’s the number of non working seniors living on SS in the main and maybe some other small sources They are filling out the bottom half of the population disproportionately. So they cause the median to be re-set lower all the time.

    I would also like to know how Medicaid is accounted for. I know in Census work it is accounted for as a function of the recipient’s income which artificially depresses its contribution to their lives and accentuates inequalty ratios equally artificially.

    Last is that people have to spend more years in school than in the past and that suppresses the contribution they would otherwise have made to income.

  18. I also think SS Disability Income recipients have grown rapidly in the post-80 period and that has the same effect on lowering the median as I stated above.

    I think what you are seeing is in considerable measure the welfare state working, frankly.. Welfare recipients fill out the bottom half of the population and exert great downward pressure on the median.

  19. @Lane:

    I plotted the hourly salary of blue-collar workers over the same time period and this shows the exact same divergence in the late 1970’s.

    The American middle class was essentially blue-collar and perhaps the stagnation is primarily a response to markets opening up in the 70’s forcing American workers to equalize wages with global competition.

  20. when the number of people per household changed (by >20%), GDP per capita and per household can not track.
    the red dashed plot must be wrong.
    The difference betweeen gdp per capita and median household just reflects household size and larger fraction of individual retirees.

    Just do the math correctly !

    The only thing you are plotting here is the change in demography, and no social change

  21. @steve
    Your question is I guess rhetorical but the demographics are indisputable. In 1980, against a population of roughly 227MM, there were roughly 35MM recipients of SS benefits, In 2010, against a pop of roughly 310MM, there were roughtly 63MM recipients of SS benefits. Total pop went up about 35% while SS recipients went up about 80%. That will affect where the median is set and draw it downward.

    Note I did not say that this explained it all. I think it is a considerable factor.

  22. I would also note that immigration has soared in the period under question, again a rate of increase that far outstrips the growth in the general population. According to the census, immigrants, seniors and disability recipients account for about 30% of the US population, up from less than half that in 1970. Again, these groups are filling out the bottom half of the population much more than the top, and dragging the median downward meaningfully.

  23. I suggest that Mr Kenworthy stops outputting for a while, and takes some classes in demography and statistics 101

  24. What the graph does illustrate is that over this time period households are more likely to have multiple bread winners. Both men and women are more likely to hold down full time positions in a single household. This makes the above graphics even more illustrative. As the economy has grown and more families have two full-time workers, household incomes have STILL stagnated relative to the overall economy.

  25. Working hours per family has doubled but family income has remained flat. Cost of going to work has doubled (cost have gone up) but net disposable income has gone down. Median single income $26 K, sustainability income level $44 K Lack of a balance of power in a real open competitive market.

  26. Very interesting post, with quite a few variables taken into account. There are always many factors affecting real world comparisons, so it depends on how you interpret the data. I believe difference in the increase is due to the widening income gap though.

  27. I notice that the decoupling coincides with the stabilization of inflation rates at 2%. Has the central bank broken the mechanism by which wages were coupled to GDP growth?

Leave a Reply