Rising Household Income Inequality, Education, and Spousal Income

Summary


  • After being relatively stable from 1947 to 1980, household income inequality has steadily increased.  Political commentary often blames “bad lifestyle choices” by the poor or the “greediness of the rich.”  
  • US Census micro data from 1980 to 2015 suggests that both assertions are wrong.[1]  The data indicates that the rise in income inequality for households is primarily due to changes in (1) household educational composition and educational payoffs and (2) spousal income.


Household Income
The US Census household GINI ratio is one of the most widely used statistics in the discussion of income inequality.  The household is used as the unit of measurement because it is viewed as the critical entity that provides for individuals.  Household members often share expenses, whether or not they are related.  The Census Bureau (CB) sums up the income of all the individuals within a household to arrive at household income (thus from the CB’s perspective, a household of one person making $20,000 is the same as a household with two people making $10,000).  The CB definition of income is pre-tax and includes “...wages and salaries; income from dividends; earnings from self-employment; rental income; child support and alimony payments; Social Security, disability, and unemployment benefits; cash welfare assistance; and pensions and other retirement income.  Census money income does not include non-cash benefits such as those from the Supplemental Nutrition Assistance Program (food stamps), Medicare, Medicaid, or employer-provided health insurance.”[2]  


Because taxes disproportionately redistribute income from wealthier households, GINI ratios calculated on an after-tax basis are lower than the CB numbers.  However, working in the opposite direction, CB inequality numbers have historically been suppressed by the CB’s practice of “top coding.” Top coding means that for reasons of privacy, if an income is over a certain limit set by the CB, it only records the CB “top code” limit amount. Since 1980, the CB has significantly increased the “top code” limit.  In 1980, the top code limit was $75,000 for individuals and households (so a household with two $75,000 wage earners would have been reported as having only $75,000 of income). Approximately 1.7% of the 1980 CB data analyzed here was top coded. However, the 2015 CB micro data has less than 0.00% samples which were top coded because the CB increased the individual income top code to $1.655 million and the household income top code to $2.1 million.  Because top coding suppresses the GINI ratio, the amount of true GINI increase since 1980 is lower than reported.[3] Despite the technical details, there is no dispute that household income has become less equal since 1980.

Figure 1: GINI Coefficients (Household and Family from published Census Bureau numbers; household head and spouse wage from my calculations)

A Big Increase In The Relative Payoff To Education

The Census data indicates that labor income is driven by two primary dynamics: (1) young workers make less than older workers with the same qualifications, and (2) more educated workers make more than less educated workers. The group that makes the least are workers in their 20s with no years of college (“no-college”). Figure 2 shows the interrelationship between education, age, and earnings in 1980 relative to that lowest group. Earnings gaps by education expand in the peak earning years which occur for ages in the 40s and 50s. While workers in their 20s with at least some postgraduate experience only made 1.6 times what their no-college peers made, 50-year old postgrad workers made 2.1 times what a no-college 50-year old made -- and 3.3 times more than 20 year olds with no college experience.

Figure 2: Relative education and age payoff in 1980
    
Figure 3 shows the same relationships in 2015 compared to 1980. Both years are scaled respectively to what no-college workers in their 20s made in those years. In 2015, workers with any postgraduate experience in their 20s started out already making 3.0 times (up from 1.6) what their no-college peers made. Fifty year old postgrad workers made 4.0 times (up from 2.1) what a no-college 50-year old -- and 6.7 times (up from 3.3) more than 20 year olds with no college experience. The increased value of education is entirely consistent with the continuing US economic evolution towards a "knowledge economy", and with the increased globalization of labor supply.

Figure 3: Relative education and age payoff in 2015 versus 1980 for no-college and postgraduate workers
Nearly all of the increased educational payoff went to the top income decile.  First of all, while bottom decile households averaged one income per household in 1980, only 22% of those incomes were from wages (in part because 29% of bottom decile household heads were older than 69). Thus 78% of the bottom decile would not benefit from the increased payoff to education. In addition, 81% of the bottom income decile had no spouse present (due either to being single, divorced or widowed) to benefit from an educational payoff.

The educational level of the bottom decile was low relative to the top decile. If one uses a weighting system that assigns a score of 1 to high school or less, a score of 2 to some years of college but no degree, 3 to a college degree, and 4 to a college degree with any postgraduate experience, the average education of the head of the household in the bottom decile was 1.3 -- slightly above high school. Only 5% of those household heads had a college degree or better. This includes categories 3 and 4 described above and is described as “College+” in Table 1.  


By 2015, the percent with college degrees or better had only increased to 15% and their average education was still only 1.6. For all these reasons, the bottom household income decile did not benefit much from the increased wage payoff to education shown in Figure 3.  

Table 1: 1980 to 2014: Key Characteristics of Top and Bottom Household Income Deciles (my calculations from data cited in footnote 1)



Average
Bottom Income Decile
Top Income Decile

1980
2015
Change
% Change
1980
2015
Change
% Change
People/Household
1.8
1.7
-0.2
-8%
3.5
3.0
-0.4
-13%
Incomes/Household
1.0
0.9
0.0
-1%
2.2
2.1
0.0
-2%
Household Income
$2,175
$6,972
$4,797
221%
$51,723
$274,520
$222,797
431%
Household Wage Income
$569
$1,967
$1,398
246%
$40,404
$204,810
$164,406
407%
% No Spouse Present
81%
88%
7%
9%
13%
18%
5%
37%
Income of "Head"
$1,983
$6,223
$4,240
214%
$37,264
$164,173
$126,909
341%
Income of Spouse
$505
$2,893
$2,388
473%
$10,343
$110,073
$99,730
964%
Spouse Income/Household Income %
23%
41%
18%
79%
20%
40%
20%
101%
Top Decile Spouse Income/Bottom Decile Household Income




4.8
15.8
11.0
232%
Education of "Head"
1.3
1.6
0.3
26%
2.3
3.0
0.7
31%
Education of Spouse
1.2
1.6
0.3
28%
1.9
2.8
0.9
48%
% No College "Head"
82%
58%
-24%
-29%
37%
10%
-27%
-73%
% No College Spouse
84%
63%
-21%
-25%
48%
14%
-34%
-72%
% College+ "Head"
6%
14%
8%
129%
43%
74%
30%
70%
% College+ Spouse
5%
15%
10%
201%
27%
66%
38%
140%
% Head Older Than 70
29%
23%
-6%
-20%
4%
9%
6%
149%
% Head Younger Than 30
21%
15%
-6%
-28%
6%
2%
-3%
-60%
% Head 30 to 70 (Peak Earnings)
50%
62%
12%
23%
91%
88%
-2%
-2%
% With Wage Income
22%
20%
-2%
-8%
62%
74%
12%
20%
% of Household Income From Wages
26%
28%
2%
8%
78%
75%
-4%
-4%
 
In contrast to the bottom decile in 1980, the top decile of household income averaged 2.2 incomes, 78% had income from wages, and 87% had a spouse present. Forty three percent of the household heads -- who would have been born in the 1930s and been of college age in the 1950s -- had a college degree or better. Twenty-seven percent of spouses in these households also had a college degree or better. The average education for the whole decile was 2.3 for household heads and 1.9 for spouses (using the scale described above). That average was less than a full college degree, but many household heads had some years of college. These households started from a better position to get the increased payoff to education.  


The top decile dramatically increased their educational advantage by 2015.  By then, 74% of household heads -- born in the mid ‘60s and of college age in the ‘80s -- had a college degree and 66% of their spouses did as well. Their average educations were 3.0 and 2.8 respectively, both at or near the college degree level. The education gap between top decile spouses and top bottom household heads which had been relatively narrow in 1980 at 1.9 versus 1.3, expanded to 2.8 versus 1.6 in 2015.   

The Significant Increase In Spousal Income And Assortive Mating

Because households in the lowest income decile have a very low percentage with spouses, an increase in spousal income had very little impact on its overall decile statistics. While spousal income in the lowest decile increased 473% from 1980 to 2015, it only affected about 15% of those households. Thus, spousal income accounted for only 6% of the household income increase for this decile.


The converse is true for the top decile in which nearly all households have spouses.  Because spouses in this decile benefited from the increased payoff to education, their income increased 964% compared to only 341% for household heads.  Because 82% of these households had spouses in 2015, these large increases in spousal income contributed 64% of the total increase in household income for this decile.


In 1980, a spouse in the top household income decile had average individual income that by itself was 4.8 times larger than the total average household income of the bottom decile.  By 2015, the top household income decile spouse’s average individual income by itself was 15.8 times larger than the total average household income of the bottom decile. These statistics dramatically illustrate the inequality effect of spousal income growth in the top household income decile.
The data also suggests that people in the top decile of household income are increasingly likely to marry spouses with similar educational levels. In 1980, the “college plus” gap between household heads and spouses in the top decile was 16% (43% vs. 27%).  By 2015, that gap narrowed to 8% (74% vs. 66%). This phenomenon is known as “assortive mating” and has been cited by others as a significant contributor to income inequality. It can be a self-perpetuating dynamic if the top decile’s children in turn, marry people at their educational level.

Magnitude Check: Could Education Payoff and Spousal Income Changes Account For The Magnitude of The GINI Coefficient Increase?

Let us imagine a two household world. In 1980, the first household is a bottom decile household: one single worker in their 20s with no college. This worker receives the typical income of a 20s no-college worker standardized to $1. The other household is a typical top decile married couple in their early 50s. The head receives the standard relative income of $2.7 per Figure 2 for someone with a college degree, and their spouse receives the standard relative income of $1.7 for someone with a few years at college but no degree.  The GINI coefficient for these two households with respective household incomes of $1 and $4.4 is 0.31.


The clock has advanced to 2015.  The bottom decile household is the same (as the data suggests) and receives the same $1 standardized to 2015 dollars.  If the second household had the same educational make up (one college, one with some college), it would now receive standardized incomes of $4.7 and $2.6.  These two household incomes in 2015 of $1 and $7.3 yield a GINI coefficient of 0.38.  If, as suggested by the data, the top decile household instead has changed due to assortive mating such that both the head and spouse had college degrees (they both receive $4.7), the GINI coefficient rises to 0.40.


This is a simplified example that does not incorporate other household compositional changes (e.g., percent of a decile in peak earning years). But it does illustrate that GINI coefficient increases due solely to increased education payoffs and spousal education in the top decile are consistent with the actual GINI coefficient increases observed since 1980 (Figure 1).  

Debunking Ad Hominem Arguments That Blame The Poor Or The Rich

Those who attempt to blame increasing inequality on the poor for allegedly bad lifestyle choices typically claim that government benefits and/or moral decay have reduced the incentive to work and promoted the formation of households lead by young people without spouses. The data in Table 1 does show that non-spouse households did in fact increase by 7% in the bottom household income decile, but they also increased by 5% in the top income decile.  Importantly (not shown in the table), the non-spouse rate for those less than 30 years of age in the bottom decile actually dropped from 21% to 15%. Young household heads in the bottom decile were more likely to have spouses in 2015 than in 1980! Similarly, while the percent of bottom decile household heads that had wages did drop very modestly from 22% to 20%, the percent of household income attributable to wages actually increased from 26% to 28%. Indeed, there is some evidence that government benefits were reallocated away from the poorest households in favor of lower middle class households in the 1980’s and 1990’s. The data in Table 1 shows no compelling evidence that the inclination to work declined within the bottom decile.


Those who attempt to blame increasing inequality on the rich often talk about increasing “greed” -- frequently making reference to the salary of CEOs relative to the poorest paid workers in their companies.[4] Most damning to this line of argument are the increases in educational payoffs and spousal income illustrated above. Has more education made people greedier? Are spouses inherently greedy? Also the numbers simply don’t work. There are 500 large company CEOs in the S&P 500, but there are 12.5 million households in the top household income decile that have shown significant income increases relative to the bottom decile.  Finally, there is some suggestion that CEO/worker pay ratio has been relatively stable within companies, but has changed dramatically across companies.


Neither ad hominem line of argument is credible.


Transparency and reproducibility:  All of the labeled Figures and Tables can be generated by my code using the free, publicly-available R program to analyze the free publicly available data available from the links in the article.[5]  


[1] IPUMS-USA, University of Minnesota, www.ipums.org. The data is provided for free, but is subject to their licensing restrictions.
[3]  To get a sense of the impact of top coding changes, the ‘92 to ‘93 GINI increase from 0.433 to 0.454 was primarily driven by a change in the top code.
[4] The dramatic increase in the relative wages of professional athletes and film stars does not seem to receive the same critical attention.
[5] Results from the R code were combined in a spreadsheet to produce the figures and tables. The spreadsheet is available by request to disciplinedthinkingblog@gmail.com.



Comments

Popular Posts