The Clinton-Trump Vote and the Socioeconomic Progress of the White Working Class

January 27, 2017

CEPR

By Dean Baker and Sarah Rawlins

Since the presidential election, there has been an ongoing debate about the extent to which support for Donald Trump by white, working-class voters was driven by racism, xenophobia, and misogyny, as opposed to economic hardships and insecurity. An aspect of this debate that is worth considering is that the size of the white working class (defined here as non-college educated) is itself dependent on the socioeconomic progress of this group.

Specifically, as the situation of the white working class improves, more children from white, working-class families will graduate from college. This means that the size of the white working class will shrink by this definition as they become more prosperous.

As we show below, if the percentage of college grads among the young had continued to increase in the years since 1979 at the rate it did in the years from 1959 to 1979, and we assume the same voting patterns among college grads and non-graduates as we saw in November, Hillary Clinton’s margin in the popular vote would have increased by 1.8 million.

Slowing Progress in College Graduation Rates

A big part of the story of the upward redistribution of the last four decades has been a slowing in the rate of growth of college graduates. The share of people age 25 to 29 who were college graduates increased by 12.0 percentage points from 1959 to 1979. Over the next twenty years it increased by just 5.1 percentage points. This slowdown affected both men and women and blacks and whites. Table 1 shows the percentage of college grads among this age group, by race and gender, for 1959, 1979, 1999, and 2015, the most recent year for which data are available.[1]

	All races			White			Black
	Total	Male	Female	Total	Male	Female	Total	Male	Female
1959	11.1	14.8	7.6	11.9	15.9	8.1	4.6	5.6	3.7
1979	23.1	25.6	20.5	24.3	27.1	21.5	12.4	13.3	11.7
1999	28.2	26.8	29.5	29.3	27.6	30.9	15.0	13.1	16.5
2015	35.6	32.4	38.9	36.7	33.2	40.3	20.5	16.5	24.2

	Source: Census Bureau.

If the white working class had continued to see the improvements in wages and living standards they saw in the period prior to 1979, a much larger share of whites would have graduated college in subsequent decades. It is important to note that the slower rate of progress is even more pronounced for blacks and especially black men. The upward redistribution worsened the situation of the working class regardless of race. However, the focus of this discussion is on the situation of whites because they had a sharp difference in voting patterns by educational attainment, which is not the case for blacks or Hispanics.

As a counterfactual, we can imagine a situation in which the rate of growth in the share of college graduates among whites in the 25–29 age group continued at the same pace as it had in the years 1959 to 1979. For men this implies an increase of 2.8 percentage points every five years and for women an increase of 3.35 percentage points. The actual share of college grads for each relevant age group in 2016, compared with the share in this hypothetical scenario are shown in the table below.[2]

		Percent of College Grads At Different Ages in 2016

	Actual		Continued Growth		Difference
	Men	Women	Men	Women	Men	Women

Ages
25-26	31.6	38.2	46.7	45.0	15.1	6.8
27-31	31.6	38.2	46.7	45.0	15.1	6.8
32-36	27.0	36.0	43.9	41.6	16.9	5.6
37-41	25.8	32.1	41.1	38.3	15.3	6.2
42-46	27.6	30.9	38.3	34.9	10.7	4.0
47-51	23.6	24.8	35.5	31.6	11.9	6.8
52-56	24.8	24.0	32.7	28.2	7.9	4.2
57-61	24.3	21.9	29.9	24.9	5.6	3.0
62-66	27.1	21.5	27.1	21.5	0.0	0.0

	Source: Census Bureau and authors’ calculations.

Using these figures and data on the number of people in each age group as of July 1, 2016 from the Census Bureau, we calculated the difference in college grads in this hypothetical scenario compared with the actual number. This calculation showed an increase in the number of male college grads of 7.2 million and female college grads of 3.2 million. The table below shows the impact of this increase in the number of college graduates on the outcome of the 2016 election, assuming that the voting patterns of the additional college graduates would have been the same as the college graduates who voted in 2016. The calculations assume a 60 percent turnout among both groups.

		Trump	Clinton
		Percent
Men	Non-College	71	23
	College	53	39

Women	Non-College	53	39
	College	44	51

		Change (millions)
	Men	-0.76	0.68
	Women	-0.17	0.22

	Total	-0.93	0.90

Source: Washington Post and authors calculations, see text.

The totals imply an increase in the margin for Clinton of 1.83 million. This would have raised her popular vote margin to 4.7 million, implying a margin of 3.5 percentage points of the total vote.

There are many caveats that need to be applied to this sort of calculation. A country in which working-class voters have far more educational opportunities than is currently the case would be very different from the United States today. It is not clear how the better prospects for the working class might affect their voting patterns. Nor is it clear how the worsening, in relative terms, of the situation of college grads could affect their attitudes. Also, the different institutional structure that would support more equal educational outcomes would affect worldviews in unpredictable ways.

However, if we abstract from these issues, and assume that the country had continued to make the same progress in educational attainment in subsequent years as it did in the period 1959 to 1979, Hillary Clinton would have won a decisive victory over Donald Trump in both the popular vote, and in all probability, the electoral college as well.

[1] The table doesn’t have data for Hispanics and Asian Americans since the data do not go back as far.

[2] The data in this table are showing actual and counter-factual percentages of college grads based on the cohorts’ ages in 2016. Since it is based on 2014 data, we don’t have data for those who are now 25 and 26. We assumed that the percentage of college graduates among these cohorts, and gap with the counter-factual, is the same as for the slight older 27-31 cohort.