The Causes of Income Inequality

The Causes of Growing Inequality

Between 1979 and 2018, real GDP per capita increased by 89.1% A939RX0Q048SBEA while median pre-tax wages increased just 6.3% LES1252881600Q. We'll call this discrepancy The Gap. What explains it?

A Didactic Math Note

Suppose X = A * B * C. Suppose ∆X means "% increase in X".

Then, 1+∆X = (1+∆A) * (1+∆B) * (1+∆C)

So, ln(1+∆X) = ln(1+∆A) + ln(1+∆B) + ln(1+∆C)

I implicitly use this observation to determine what "%" of the gap various hypotheses explain.

Accounting Issues

Laborers are being paid more via benefits these days relative to historically. In particular, salaries and wages as a percent of compensation have decreased from 84.8% to 81.3% Table 2.1. This explains 7% of The Gap.

The hawk-eyed reader will also note a definitional foible: the employment-population ratio. At the moment, we are comparing the median full-time wage to wages per capita. If more people are working now (as a percentage) relative to 1979, then that could explain part of The Gap.

The hours-worked-per-person has increased 4.5% during this time period Table 6.9B Table 6.9D. The worst case scenario is that a X% increase in the hours-worked-per-person can explain an X% gap between GDP-per-capita and average full-time compensation. In this case the change in the employment-population ratio explains just 0-7% of The Gap.

Finally, economists use CPI when converting nominal wages to real wages, but they use the GDP deflator when converting nominal GDP per capita to nominal GDP per capita. During this time frame, CPI increased 3.460-fold CPIAUCSL while the GDP deflator increased 2.848-fold USAGDPDEFAISMEI. In other words CPI has grown 21.5% faster. This accounts for 34% of The Gap.

This raises the question of why the CPI grows faster than the GDP deflator. There are really two possibilities.

The first is that the two indices measure different sets of goods and services. CPI is constructed to represent what typical consumers buy while the GDP-deflator is based on price of all finished goods and services produced in the country. Loosely speaking, if GDP is represented as C + I + G + X - M, CPI includes C and some of M while the GPD deflator includes C, I, G, X, but not M. It's possible that prices have grown faster for M than for I and X. Alternatively, its possible that the CPI's basket of goods doesn't actually represent all consumption.

The second explanation is that relates to book-keeping minutia: CPI is a Laspeyres index, which means its biased upwards, while the GDP deflator is a Paasche index, which means it's biased downwards. It's worth noting that there isn't some objective measure of inflation, so its not that any of these indices need be wrong, just different attempts with similar goals.

I don't know of a perfect way to distinguish between these two explanations, so I'll us an imperfect one: if we apply GDP methodology to just consumption (instead of GDP), we find CPI has grown 19.6% faster DPCERD3Q086SBEA, which represents 92% of the CPI-GDP-deflator difference.

This suggests (but doesn't prove) that the different sets of goods and services accounts for only ~1.6% of the growth while the index type accounts for the vast, vast majority of it.

The above issues suggest that between 41% and 48% of The Gap isn't due to underlying economic conditions but due to measurement differences. The Gap was 77.9%, but The True Gap is between 41% and 48%.

Capitalists

Maybe the owners of capital are making off with all the money rather than the laborers. During the same time period, labor compensation as a percent of GDP decreased from 57.2% to 53.2% Table 2.1. Table 1.10. This explains about 20% of The True Gap.

Labor's share of GDP has shrunk 7% since 1979, almost entirely since 2000.

Taxes and Transfers

The measurement of median income we've been using has been pre-tax median income. However, taxes have changed since 1979 and this could affect median post-tax. Fortunately The Economist has looked into this Inequality or middle incomes: which matters more?:

In short, they've found that changes in taxes and transfers have caused median household income to increase ~22% during the time period.

On a related note, the effective tax rate paid by the top 1% has ebbed and flowed, but hasn't trended any particular direction during the time period Brookings Institution:

Labor Compensation Distribution

This leaves us with 33-40% of The Gap and 80% of The True Gap to explain. As far as I can tell, the only factor left that we haven't accounted for is the distribution of compensation for labor.

There are lots of theories regarding why the distribution of labor compensation has grown less equal Economic inequality, some are

  • The difference in average earnings between various education level is growing.
  • Unions are becoming less common.
  • The minimum wage is falling relative to inflation/the median wage/the mean wage/GDP per capita.
  • Technology is making low-skill jobs obsolete/lower-paying
  • The ability to trade freely is causing low-skill jobs to be offshored.
  • Some wealthy groups are influencing government policy to make themselves yet richer.

I'm not going to go into which of these is true or how much of the gap they account for - mostly because I'm deeply uncertain and that kind of analysis is difficult.

General Causes

Twin Studies

Twin studies Twin study can be used to estimate what proportion of income differences are due to genes, shared environment, and unshared environment. Shared environment is defined as the portion of environment siblings share while unshared environment is a catch-all bin for everything else.

At least five twin studies have ben conducted in the US to estimate the effect of nature and nurture on lifetime income Hyytinen. The median estimates are 40% genetic, 11% shared environment, and 46% random chance.

One use of this data is to evaluate the narrative that people born into poor families having fewer opportunities and, subsequently, earn less.

One view is that 11% looks pretty small: even if we raised all children in the same environment income inequality would only shrink by about 11%.

The other view is that we should be looking at effect size rather than variance explained. Taking the square root of 11% yields 0.33. We can interpret this as meaning

Take two average clones, B and G. Suppose B is raised in a bad environment (bottom 10th percentile), and G is raised in a good environment (top 90th percentile). Then (on average), we'd expect B to end up in the 33rd percentile of income while G would end up in the 67th. That is, G would earn nearly twice as much as B due solely to their environments.

Which interpretation is right? Both. They only seem contradictory because human statistical intuition is mediocre.

Ultimately, the former (11%) interpretation is more useful if you care about the inequality of outcomes while the latter (0.33) is more useful if you care about inequality of opportunity. We're focusing on the former here.

Race, Sex, and College

The National Longitudinal Survey of Youth NLSY97 is following about 9000 students born between 1980 and 1984 and asked them thousands of questions over several decades. It's also one of the most mined datasets in social science - the kind of thing where, upon reading it, you see it pop up all over the place.

It lets us find, for instance, that

  • Whether you're black explains 3% of income variance.
  • Whether you're male explains 4% of income variance.
  • Whether you have a Bachelors degree explains 9% of income variance

These correlational estimates probably place an upper limit on how much each of these factors cause income inequality. Education in particular probably only causes about half of its premium, which means education probably only causes about a quarter of the variance. So ultimately these three factors probably cause at most 10% of income inequality.

Moreover, the black gap hasn't really changed over the last several decades and the male-female gap has shrunk, so neither of those explains the increase in inequality. Finally, even if Bachelor's degrees hadn't existed in 1979, the degree would be responsible for, at most, 4% of the increase in income inequality.

It's worth noting, the fact that these three factors explain little of total income inequality or its change in the past several decades does not imply these three factors are "small" in an intuitive sense. As previously mentioned, the gap between high school graduates and college graduates is ~70%, the gap between blacks and whites is of a similar magnitude, and the gap between men and women is ~30%. These are obviously large differences - they just don't explain a lot of overall income variance because income is distributed so nonuniformly.

Summary: Why Is Income Less Equal?

Real GDP-per-capita has increased 1.65% per year over the past 39 years. After accounting for various accounting issues, the median after-tax compensation has grown a slower: about 0.8% per year. The reason for the remaining 0.85% gap is split between capitalists earning more of the pie (0.17%) and labor compensation growing less equal (0.68%).

On the other hand, increasingly progressive taxes and transfers have added 0.59% to median after-tax income growth. This, combined with the accounting issues, has meant that after-tax income has lagged GDP-per-capita by only 0.26% per year - a far cry from the 1.49% lag the unadjusted numbers show.

Genetic differences cause about 40% of income inequality, while shared environment explains just 11%. Even so, childhood environment can cause large changes in expected child income.

Obvious "categories" like sex, race, and educationoal degrees explain at most 12% of overall income inequality, some of which certainly overlaps with the shared-environment percentage. These categories explain little of income inequality growth over the last 4 decades.

U.S. Bureau of Economic Analysis, Real gross domestic product per capita [A939RX0Q048SBEA], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/A939RX0Q048SBEA, January 12, 2020. U.S. Bureau of Labor Statistics, Employed full time: Median usual weekly real earnings: Wage and salary workers: 16 years and over [LES1252881600Q], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/LES1252881600Q, January 12, 2020. Table 2.1. Gross Domestic Product. The Bureau of Economic Analysis. https://apps.bea.gov/iTable/index_nipa.cfm. U.S. Bureau of Labor Statistics, Consumer Price Index for All Urban Consumers: All Items in U.S. City Average [CPIAUCSL], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/CPIAUCSL, April 3, 2020. Organization for Economic Co-operation and Development, GDP Implicit Price Deflator in United States [USAGDPDEFAISMEI], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/USAGDPDEFAISMEI, April 4, 2020. U.S. Bureau of Economic Analysis, Personal consumption expenditures (implicit price deflator) [DPCERD3Q086SBEA], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/DPCERD3Q086SBEA, April 27, 2020. Table 1.10. Gross Domestic Income by Type of Income. The Bureau of Economic Analysis. https://apps.bea.gov/iTable/index_nipa.cfm. Inequality or middle incomes: which matters more? (2017, January 7). The Economist. https://www.economist.com/united-states/2017/01/07/inequality-or-middle-incomes-which-matters-more Urban Institute & Brookings Institution. Historical Average Federal Tax Rates for All Households: 1979 to 2016. Retrieved from https://www.taxpolicycenter.org/statistics/historical-average-federal-tax-rates-all-households Wikipedia contributors. (2020, April 3). Economic inequality. In Wikipedia, The Free Encyclopedia. Retrieved 15:11, April 4, 2020, from https://en.wikipedia.org/w/index.php?title=Economic_inequality&oldid=948949510#Various_proposed_causes_of_economic_inequality Wikipedia contributors. (2020, May 4). Twin study. In Wikipedia, The Free Encyclopedia. Retrieved 16:04, May 5, 2020, from https://en.wikipedia.org/w/index.php?title=Twin_study&oldid=954870727 Hyytinen, A., Ilmakunnas, P., Johansson, E., & Toivanen, O. (2013). Heritability of lifetime income. Helsinki Center of Economic Research Discussion Paper, (364). http://dx.doi.org/10.2139/ssrn.2253264. NLSY97 (National Longitudinal Survey of Youth 1997). (n.d.). Retrieved from https://www.nlsinfo.org/investigator/pages/search Table 6.9B. Hours Worked by Full-Time and Part-Time Employees by Industry. The Bureau of Economic Analysis. https://apps.bea.gov/iTable/index_nipa.cfm. Table 6.9D. Hours Worked by Full-Time and Part-Time Employees by Industry. The Bureau of Economic Analysis. https://apps.bea.gov/iTable/index_nipa.cfm.