The premise of Occupy Wall Street is simple: American society is becoming more unequal as a privileged minority takes control of an increasingly larger share of wealth and power. As I mentioned in my post last week, not all scholars agree with this assessment.
Robert J. Gordon at Northwestern University in a 2009 NBER paper argues, “The rise in American inequality has been exaggerated both in magnitude and timing.” In reviewing Gordon’s paper I want to emphasize areas where there is genuine disagreement, and where there is consensus. The punchline: Gordon focuses on how potential mismeasurement has overstated inequality within the bottom 99% of the income distribution, but he actually does not disagree that the share of the top 1% has dramatically outpaced at other groups.
Gordon claims inequality has been exaggerated in three ways:
- 1. The conventional measure showing a large gap between growth of median real household income and of productivity greatly overstates the increase compared to a conceptually consistent alternative gap concept.
- 2. The increase of inequality is not a steady ongoing process; after widening most rapidly between 1981 and 1993, the growth of inequality reversed itself and became negative during 2000‐2007.
- 3. An emerging literature documents an exaggeration of the rise of inequality due to the use of common price indexes across income groups.
Comparing Apples and Oranges?
Gordon’s wants to undermine a now common observation: total economic productivity has dramatically outpaced median household income growth over the last thirty years. The conventional claim might illustrate that typical households have not been sharing in the spectacular growth of the American economy.
The conventional method is wrong, Gordon argues, because it does not adjust for factors including shrinking household sizes, the use of GDP deflators used to compare incomes over time, and changing number of hours worked. The conventional estimate for the income-productivity gap between 1979-2007 is 1.46, but after making all adjustments it is reduced to 0.16. Moreover, mean and median incomes grew at remarkably similar rates during this period, even after accounting for topcoding.
The failure to adjust conventional income series is a potentially serious problem. I have already written about work by Richard Burkhauser et al. that suggests that household size adjustments and choice of deflators can substantially change estimated household income growth rates. One question that we should be asking, however, is whether workers adding additional hours or women and teenagers entering the workforce reflects some fundamental change in the labor market over the previous three decades that is a response to declining real wages. If so, then we may not want to completely adjust for it in inequality calculations.
How Plausible is Skills Biased Technological Change?
One popular explanation for rising inequality is SBTC – the mechanization of low or middle-skilled professions leading to a dampening of wages in those sectors. The figure below shows that there was a steady increase in the ratio of income between the 90th percentile and the 50th percentile from the 70s onward, and a very large divergence between the 50th and the 10th that began in the early 1980s. The apparent compression between the 50th and the 10th percentile in the late 1990s would be inconsistent with the SBTC hypothesis if the mechanization were only affecting the lowest skilled workers, but as Gordon points out, there is evidence that the middle sectors (such as clerical workers and factory workers) have been hardest hit by SBTC. Outsourcing is another piece of the puzzle, but there is a lingering question about the persistence and scope of outsourcing after the current recession.
The more interesting question is what has been happening above the 90th percentile. The highest earners are a heterogeneous group (as I reviewed last week), and the degree to which their compensation is tethered to the productivity of their sector in the market varies. Gordon actually emphasizes this point for CEO pay: “While superstars and top professionals have their incomes chosen by the market, CEO compensation is chosen by their peers, a system that gives CEOs and their hand-picked boards of directors, rather than the market, control over top incomes.” Beginning in the 1990s, stock options became a very important piece of the compensation of executives. While the compensation of CEOs and the performance of markets are clearly correlated (so that they take in less during recession years, see below), other forms of income such as salaries are much better protected from the market.
Concerning the top 1%, Gordon favors market-based explanations for rising compensation: rising CEO pay reflects a closer alignment of CEO compensation with market fundamentals. However, there have also been changes in the structure of financial regulations and taxes on the wealthiest that have favored this group. While Gordon acknowledges other arguments showing greater rent seeking among CEOs, he is quick to dismiss them. He shouldn’t be if Hacker and Pierson are right.
Getting to the bottom of the inequality puzzle is not easy. The Gordon paper is one example of how reasonable scholars can interpret income and earnings data and develop a somewhat different story than the conventional narrative. None of this should conceal points of obvious agreement: inequality in income has been growing at most points in the income distribution (it’s only a question of how much and why), and the very top is continuing to take in a larger share of national wealth. Should this bother you? We need to have a deeper understanding of the micro-mechanisms that mediate inequality at the very top (is it politics? Changing mores about executive compensation? Financial shenanigans?). We also need to get a better handle on the complex relationship between the productivity of workers and their take-home pay.