Skip to main content
Menu
Education, Not Income, Predicted Who Would Vote For Trump

Sometimes statistical analysis is tricky, and sometimes a finding just jumps off the page. Here’s one example of the latter.

I took a list of all 981 U.S. counties1 with 50,000 or more people2 and sorted it by the share of the population3 that had completed at least a four-year college degree. Hillary Clinton improved on President Obama’s 2012 performance in 48 of the country’s 50 most-well-educated counties. And on average, she improved on Obama’s margin of victory in these countries by almost 9 percentage points, even though Obama had done pretty well in them to begin with.

COUNTY COLLEGE DEGREE MEDIAN HOUSEHOLD INCOME OBAMA 2012 CLINTON 2016 SHIFT
Average 51.4% $77,768k +17.3 +25.9 +8.5
Arlington, VA 72.0 105,120 +39.8 +60.1 +20.3
Alexandria, VA 61.5 87,319 +43.5 +59.0 +15.5
Howard, MD 60.4 110,133 +22.0 +33.5 +11.5
New York, NY 59.3 71,656 +68.8 +77.2 +8.4
Fairfax, VA 59.2 112,102 +20.5 +36.2 +15.7
Boulder, CO 58.2 69,407 +41.8 +48.7 +6.9
Loudoun, VA 58.0 123,966 +4.5 +16.8 +12.3
Montgomery, MD 57.4 98,704 +43.9 +55.6 +11.7
Orange, NC 56.2 57,261 +42.2 +51.0 +8.8
Douglas, CO 55.9 102,626 -25.8 -18.1 +7.7
Hamilton, IN 55.6 84,635 -34.3 -19.6 +14.7
Marin, CA 54.8 91,529 +51.3 +62.8 +11.5
Williamson, TN 54.1 91,743 -46.5 -35.5 +11.0
District of Columbia 53.4 69,235 +83.6 +88.7 +5.1
San Francisco, CA 52.9 78,378 +70.5 +75.7 +5.2
Johnson, KS 52.1 75,017 -17.4 -2.7 +14.7
Albemarle, VA 52.1 67,958 +12.0 +25.0 +13.0
Somerset, NJ 52.0 100,903 +5.6 +12.5 +6.9
Washtenaw, MI 51.8 60,805 +35.9 +41.5 +5.6
Johnson, IA 51.7 54,985 +35.5 +38.2 +2.7
Benton, OR 51.4 49,338 +28.5 +33.8 +5.3
Middlesex, MA 51.3 83,488 +27.1 +38.9 +11.8
Delaware, OH 51.1 91,936 -23.2 -16.1 +7.1
Morris, NJ 50.6 99,142 -10.8 -4.4 +6.4
Tompkins, NY 50.3 52,836 +40.6 +42.1 +1.5
Norfolk, MA 49.9 86,469 +15.2 +31.6 +16.4
Broomfield, CO 49.5 80,430 +6.0 +14.1 +8.1
Douglas, KS 49.4 50,732 +24.6 +32.7 +8.1
Collin, TX 49.4 84,233 -31.5 -17.0 +14.5
Chester, PA 48.8 86,093 -0.2 +9.3 +9.5
Fulton, GA 48.6 56,642 +29.8 +42.1 +12.3
Story, IA 48.5 51,270 +13.8 +12.2 -1.6
Hunterdon, NJ 48.3 106,519 -17.8 -13.8 +4.0
Wake, NC 48.3 66,579 +11.4 +20.5 +9.1
Chittenden, VT 48.0 64,243 +41.6 +47.4 +5.8
Boone, MO 47.7 49,059 +3.1 +5.9 +2.8
Dane, WI 47.6 62,303 +43.5 +48.0 +4.5
Santa Clara, CA 47.3 93,854 +42.9 +52.3 +9.4
Eagle, CO 47.3 73,774 +14.9 +19.9 +5.0
King, WA 47.1 73,035 +40.6 +50.5 +9.9
DuPage, IL 46.7 79,016 +1.1 +14.1 +13.0
Gallatin, MT 46.7 54,298 -5.0 +1.0 +6.0
Ozaukee, WI 46.4 75,643 -30.3 -19.3 +11.0
Hennepin, MN 46.4 65,033 +27.0 +35.3 +8.3
Madison, MS 46.3 63,156 -15.7 -16.0 -0.3
Montgomery, PA 46.2 79,926 +14.3 +21.1 +6.8
James City, VA 46.1 76,705 -12.0 -5.1 +6.9
Bergen, NJ 46.1 83,686 +11.3 +12.0 +0.7
Westchester, NY 46.0 83,422 +25.1 +32.8 +7.7
Durham, NC 45.6 52,038 +52.8 +60.4 +7.6
Clinton’s margin surged in the 50 most-educated counties

Sources: American Community Survey, U.S. Election Atlas, ABC News

Although they all have highly educated populations, these counties are otherwise reasonably diverse. The list includes major cities, like San Francisco, and counties that host college towns, like Washtenaw, Michigan, where the University of Michigan is located. It also includes some upper-middle-class, professional counties such as Johnson County, Kansas, which is in the western suburbs of Kansas City. It includes counties in states where Clinton did poorly: She improved over Obama in Delaware County, Ohio, for example — a traditionally Republican stronghold outside Columbus — despite her numbers crashing in Ohio overall. It includes extremely white counties like Chittenden County, Vermont (90 percent non-Hispanic white), and more diverse ones like Fulton County, Georgia, where African-Americans form the plurality of the population. If a county had high education levels, Clinton was almost certain to improve there regardless of the area’s other characteristics.

Now here’s the opposite list: The 50 counties (minimum population of 50,000) where the smallest share of the population has bachelor’s degrees:

COUNTY COLLEGE DEGREE MEDIAN HOUSEHOLD INCOME OBAMA 2012 CLINTON 2016 SHIFT
Average 13.3% $41,108 -19.3 -30.5 -11.3
Liberty, TX 8.8 47,722 -53.3 -58.0 -4.7
Starr, TX 9.6 25,906 +73.3 +60.1 -13.2
Acadia, LA 9.9 37,684 -49.8 -56.7 -6.9
Apache, AZ 10.1 32,396 +34.3 +36.9 +2.6
Duplin, NC 10.4 34,787 -11.6 -19.2 -7.6
Walker, AL 10.7 36,712 -52.8 -67.5 -14.7
Edgecombe, NC 10.7 33,892 +36.2 +32.2 -4.0
St. Mary, LA 11.1 41,956 -18.8 -27.6 -8.8
DeKalb, AL 11.3 37,977 -54.7 -69.4 -14.7
Anderson, TX 11.3 42,511 -52.1 -58.1 -6.0
McKinley, NM 11.4 29,812 +46.9 +39.5 -7.4
Henry, VA 11.5 34,344 -14.7 -29.2 -14.5
Putnam, FL 11.6 32,714 -24.5 -36.6 -12.2
Darke, OH 11.6 43,323 -44.4 -61.2 -16.8
Halifax, NC 11.9 32,834 +32.3 +26.9 -5.4
Laurel, KY 11.9 35,746 -63.6 -69.1 -5.5
Sampson, NC 12.1 35,731 -10.9 -16.7 -5.8
Maverick, TX 12.1 32,536 +58.1 +55.8 -2.3
Mohave, AZ 12.2 38,456 -42.1 -51.5 -9.4
Blount, AL 12.3 44,409 -73.9 -81.4 -7.5
Robeson, NC 12.4 30,581 +17.4 -4.8 -22.2
Kings, CA 12.5 47,341 -14.9 -17.4 -2.5
Talladega, AL 12.5 35,896 -16.0 -25.5 -9.5
Pike, KY 12.5 32,571 -50.5 -62.7 -12.2
Marion, OH 12.5 42,904 -6.4 -34.4 -28.0
Lea, NM 12.6 55,248 -49.8 -48.3 +1.5
Columbus, NC 12.7 34,597 -7.8 -22.1 -14.3
Terrebonne, LA 12.9 49,932 -41.2 -48.4 -7.2
Wilkes, NC 12.9 32,157 -42.4 -55.2 -12.8
Jackson, AL 12.9 36,874 -41.8 -62.5 -20.7
Le Flore, OK 12.9 35,970 -41.1 -58.7 -17.6
Merced, CA 13.0 43,066 +8.7 +7.9 -0.8
Hawkins, TN 13.0 37,432 -46.9 -63.4 -16.5
Vermilion, LA 13.0 47,344 -52.8 -59.6 -6.8
St. Landry, LA 13.1 33,928 -4.3 -11.9 -7.6
Rockingham, NC 13.1 38,946 -21.1 -30.0 -8.9
Huron, OH 13.1 49,315 -8.3 -36.4 -28.1
Clearfield, PA 13.2 41,510 -28.9 -49.5 -20.6
Tulare, CA 13.3 42,863 -15.0 -16.2 -1.2
Rusk, TX 13.3 46,924 -51.1 -56.6 -5.5
Ashtabula, OH 13.4 40,304 +12.8 -19.0 -31.8
Imperial, CA 13.4 41,772 +32.0 +41.8 +9.7
Bullitt, KY 13.4 56,199 -35.7 -49.8 -14.1
Caldwell, NC 13.4 34,853 -35.5 -50.6 -15.1
Montcalm, MI 13.4 40,739 -8.6 -34.0 -25.4
Madera, CA 13.5 45,490 -17.1 -17.3 -0.2
Dickson, TN 13.5 45,056 -28.4 -45.7 -17.3
Tuscola, MI 13.5 44,017 -10.8 -38.0 -27.2
Pearl River, MS 13.5 40,997 -59.3 -66.7 -7.4
Columbiana, OH 13.6 43,707 -11.8 -41.6 -29.8
Clinton collapsed in the 50 least-educated counties

Sources: American Community Survey, U.S. Election Atlas, ABC News, Alaska Division of Elections

These results are every bit as striking: Clinton lost ground relative to Obama in 47 of the 50 counties — she did an average of 11 percentage points worse, in fact. These are really the places that won Donald Trump the presidency, especially given that a fair number of them are in swing states such as Ohio and North Carolina. He improved on Mitt Romney’s margin by more than 30 points (!) in Ashtabula County, Ohio, for example, an industrial county along Lake Erie that hadn’t voted Republican since 1984.

And this is also a reasonably diverse list of counties. While some of them are poor, a few others — such as Bullitt County, Kentucky, and Terrebonne Parish, Louisiana — have average incomes. There’s also some racial diversity on the list: Starr County, Texas, is 96 percent Hispanic, for example, and Clinton underperformed Obama there (although she still won it by a large margin). Edgecombe County, North Carolina, is 57 percent black and saw a shift toward Trump.

How do we know that education levels drove changes in support — as opposed to income levels, for example? It’s tricky because there’s a fairly strong correlation between income and education.4 Nonetheless, with the whole country to pick from, we can find some places where education levels are high but incomes are average or below average. If education is the key driver of changes in the electorate, we’d expect Clinton to hold steady or gain in these counties. If income matters more, we might see her numbers decline.

As it happens, I grew up in one of these places: Ingham County, Michigan, which is home to Michigan State University and the state capital of Lansing, along with a lot of auto manufacturing jobs (though fewer than there used to be). The university and government jobs attract an educated workforce, but there aren’t a lot of rich people in Ingham County. How did Clinton do there? Just fine. She won it by 28 percentage points, the same as Obama did four years ago, despite her overall decline in Michigan.

And in most places that fit this description, Clinton improved on Obama’s performance. I identified 22 counties5 where at least 35 percent of the population has bachelor’s degrees but the median household income is less than $50,0006 and at least 50 percent of the population is non-Hispanic white (we’ll look at what happened with majority-minority counties in a moment, so hang tight). Clinton improved on Obama’s performance in 18 of the 22 counties, by an average of about 4 percentage points:

COUNTY COLLEGE DEGREE MEDIAN HOUSEHOLD INCOME OBAMA 2012 CLINTON 2016 SHIFT
Average 40.2% $43,862 +4.8 +8.8 +4.0
Brazos, TX 38.3 39,060 -35.3 -23.6 +11.7
Champaign, IL 42.5 46,680 +7.0 +18.4 +11.4
Clarke, GA 39.3 33,430 +28.8 +38.0 +9.2
Harrisonburg, VA 35.6 38,807 +13.4 +21.9 +8.5
Fayette, KY 40.2 48,667 +1.0 +9.4 +8.4
Riley, KS 45.5 44,522 -12.0 -4.5 +7.5
Davidson, TN 36.5 47,434 +18.6 +26.0 +7.4
Benton, OR 51.4 49,338 +28.5 +33.8 +5.3
Alachua, FL 40.8 42,045 +17.4 +22.6 +5.2
Watauga, NC 38.0 35,491 -3.1 +1.5 +4.6
Monroe, IN 44.2 41,857 +19.1 +23.7 +4.6
Boone, MO 47.7 49,059 +3.1 +5.9 +2.8
Buncombe, NC 35.1 45,642 +12.5 +14.6 +2.1
Montgomery, VA 44.3 44,810 -0.3 +1.3 +1.6
Leon, FL 44.3 46,620 +23.6 +25.1 +1.5
Lafayette, MS 36.9 41,343 -15.3 -14.8 +0.5
New Hanover, NC 37.2 49,582 -4.6 -4.1 +0.5
Payne, OK 36.4 37,637 -28.4 -28.3 +0.1
Ingham, MI 36.5 45,278 +27.8 +27.7 -0.1
Monongalia, WV 38.8 46,166 -9.5 -10.4 -0.9
Tippecanoe, IN 35.2 44,474 -3.6 -5.7 -2.1
Missoula, MT 40.2 47,029 +17.8 +15.7 -2.1
High-education, medium-income white counties shifted to Clinton

Counties shown have a population of at least 50,000. At least 50 percent of residents are non-Hispanic whites, at least 35 percent of the age-25-and-older population has a bachelor’s degree or higher, and the median household income is below $50,000.

Sources: American Community Survey, U.S. Election Atlas, ABC News

Are these so-called “white working-class” counties? You could argue for it: They’re mostly white, and they have average or below-average incomes. But, of course, “class” is a slippery term, and definitions vary. It is worth noting that many of the counties on the list are home to major colleges or universities, although there are some exceptions. Clinton made substantial gains in Nashville, Tennessee (Davidson County), and modest gains in Asheville, North Carolina (Buncombe County), for instance, and both places have reputations as intellectual and cultural havens but aren’t really college towns.7

There are also some counties where incomes are high but residents aren’t particularly well-educated. Take Suffolk County, New York, for instance, which comprises the eastern three-quarters of Long Island. The median household income there is around $88,000, but only about a third of the population has college degrees (as compared to a national average of around 30 percent). Suffolk County turned into Trump Territory, voting for him by 8 percentage points after Obama had won it by 4 points in 2012. Trump made even larger gains in Staten Island, New York (Richmond County), winning it by 17 points after Obama won it by 3 points in 2012.

Long Island and Staten Island might be peculiar cases because voters there may have a cultural affinity with Trump, who grew up in Queens. Even so, they reveal something about how cultural and educational fault lines can mean more than economic circumstances. Clinton improved over Obama’s performance in suburban Westchester County, New York, for instance, which has broadly similar income levels to Long Island and Staten Island but higher education levels and a different mix of occupations.8 (Staten Island is famous for its large population of police and firefighters, but you’ll meet a lot more journalists who have homes in Westchester.9)

Trump improved on Romney’s performance in 23 of 30 counties where median incomes are $70,000 or higher but less than 35 percent of the population have college degrees and the majority of the population is white. For example, Trump won by a much larger margin than Romney in Calvert County, Maryland, which has some commonalities with Long Island.10 And he substantially improved on Romney’s performance in Chisago County, Sherburne County and Wright County in the Minneapolis exurbs, even though Clinton made major gains in Minneapolis’ Hennepin County. There’s probably some degree of cultural self-sorting at play here. These communities have plenty of nice homes and good schools — they’re not cheap to live in — but they have fewer cultural amenities or pretensions (think big-box retail as opposed to boutiques) than you usually find in nearer-in suburbs and small towns such as those in Westchester County.

COUNTY COLLEGE DEGREE MEDIAN HOUSEHOLD INCOME OBAMA 2012 CLINTON 2016 SHIFT
Average 30.4% $76,701 -11.0 -15.8 -4.8
Richmond, NY 30.6 74,043 +2.6 -16.8 -19.4
Chisago, MN 21.5 70,223 -12.6 -30.6 -18.0
Sherburne, MN 26.2 73,621 -22.0 -37.1 -15.1
Litchfield, CT 33.7 72,068 -3.6 -16.0 -12.3
Orange, NY 28.6 70,794 +5.7 -6.4 -12.1
Suffolk, NY 33.5 88,323 +3.7 -8.2 -11.9
Wright, MN 27.4 73,085 -21.7 -33.2 -11.5
Gloucester, NJ 28.7 76,213 +10.8 -0.5 -11.3
Calvert, MD 29.3 95,425 -7.5 -18.4 -10.9
Warren, NJ 29.5 70,934 -15.5 -25.6 -10.1
St. Mary’s, MD 29.8 88,190 -14.8 -24.6 -9.8
Sussex, NJ 33.1 87,397 -21.4 -30.2 -8.8
Dutchess, NY 33.4 72,471 +7.5 -1.1 -8.6
Anoka, MN 27.3 70,464 -2.6 -9.7 -7.1
Livingston, MI 33.0 73,694 -23.3 -29.6 -6.3
St. Croix, WI 32.4 70,313 -12.1 -18.4 -6.3
Harford, MD 33.4 81,016 -18.4 -24.5 -6.1
Spotsylvania, VA 28.3 78,505 -11.5 -16.8 -5.3
Fauquier, VA 34.3 92,078 -19.9 -24.7 -4.8
Carroll, MD 32.7 85,532 -32.9 -36.9 -4.0
Chesapeake, VA 29.4 70,176 +1.0 -1.3 -2.3
Ascension, LA 25.8 70,207 -34.3 -36.0 -1.7
Elko, NV 17.5 72,280 -53.2 -54.7 -1.5
Will, IL 32.6 76,142 +5.5 +5.6 +0.1
McHenry, IL 32.2 76,345 -8.8 -8.0 +0.8
Kendall, IL 34.3 83,844 -3.3 -1.5 +1.8
Plymouth, MA 34.0 75,816 +4.2 +10.1 +5.9
Napa, CA 31.9 70,925 +28.7 +35.3 +6.6
Kane, IL 31.8 70,514 +1.1 +9.0 +7.9
Davis, UT 34.6 70,388 -61.9 -22.9 +39.0
High-income, medium-education white counties shifted to Trump

Counties shown have a population of at least 50,000. At least 50 percent of residents are non-Hispanic whites, less than 35 percent of the age-25-and-older population has a bachelor’s degree or higher, and the median household income is above $70,000.

Sources: American Community Survey, U.S. Election Atlas, ABC News

Education levels are also increasingly dividing majority-minority communities from one another. For example, let’s look at a set of counties that were a sweet spot for the Obama coalition — those that are both diverse and highly educated. In particular, there are 24 counties (minimum population 50,000) in the U.S. where at least 35 percent of the population has college degrees and less than half the population is non-Hispanic white. Obama did really well in these counties in 2012, winning them by an average of 41 percentage points. But Clinton did even better, winning them by 47 points, on average. The only two such counties that Obama had lost, Clinton won: Fort Bend County, Texas, in suburban Houston, which voted for a Democrat for the first time since 1964, and Orange County, California, which hadn’t voted Democratic since 1936.

COUNTY COLLEGE DEGREE NON-HISPANIC WHITE OBAMA 2012 CLINTON 2016 SHIFT
Average 42.9% 41.9% +41.2 +47.5 +6.3
Fort Bend, TX 42.3 35.5 -6.8 +6.6 +13.4
Fulton, GA 48.6 40.6 +29.8 +42.1 +12.3
Montgomery, MD 57.4 47.4 +43.9 +55.6 +11.7
Orange, CA 37.3 42.9 -6.2 +5.2 +11.4
San Mateo, CA 45.0 41.2 +46.7 +57.2 +10.5
San Diego, CA 35.1 47.5 +7.6 +17.1 +9.5
Santa Clara, CA 47.3 34.1 +42.9 +52.3 +9.4
New York, NY 59.3 47.4 +68.8 +77.2 +8.4
Yolo, CA 38.3 48.8 +34.0 +42.1 +8.1
DeKalb, GA 40.3 29.7 +56.8 +64.7 +7.9
Suffolk, MA 41.0 47.1 +56.7 +64.6 +7.9
Contra Costa, CA 39.4 46.6 +35.2 +42.9 +7.7
Durham, NC 45.6 42.1 +52.8 +60.4 +7.6
Mecklenburg, NC 41.5 49.6 +22.4 +29.9 +7.5
Richmond, VA 35.4 39.7 +57.3 +63.8 +6.5
San Francisco, CA 52.9 41.4 +70.5 +75.7 +5.2
District of Columbia 53.4 35.4 +83.6 +88.7 +5.1
Prince William, VA 38.1 47.0 +16.0 +20.1 +4.1
Alameda, CA 42.1 33.3 +60.7 +64.4 +3.7
Cook, IL 35.3 43.4 +49.4 +53.0 +3.6
Richland, SC 36.2 44.6 +32.0 +32.9 +0.9
Santa Fe, NM 39.9 43.4 +51.1 +50.8 -0.3
Hudson, NJ 36.8 29.6 +56.1 +51.9 -4.2
Middlesex, NJ 40.7 47.0 +27.6 +19.7 -7.9
Highly educated majority-minority counties shifted toward Clinton

Counties on this list have a population of at least 50,000. Less than 50 percent of residents are non-Hispanic whites and at least 35 percent of the age-25-and-older population has a bachelor’s degree or higher.

Sources: American Community Survey, U.S. Election Atlas, ABC News

By contrast, Clinton struggled (relatively speaking) in majority-minority communities with lower education levels. Among the 19 majority-minority countries where 15 percent or less of the population has a bachelor’s degree, she won by an average of only 7 percentage points, less than Obama’s 10-point average margin of victory in 2012. We need to be slightly careful here because of the potential ecological fallacy — it’s not clear if minority voters shifted away from Clinton in these counties or if the white voters who live there did. Still, Trump probably gained overall among Latino and black voters compared to Romney, and it’s worth investigating divisions within those communities instead of treating their votes as monolithic.

COUNTY COLLEGE DEGREE NON-HISPANIC WHITE OBAMA 2012 CLINTON 2016 SHIFT
Average 12.8% 30.3% +10.1 +7.0 -3.1
Robeson, NC 12.4 26.7 +17.4 -4.8 -22.2
Cumberland, NJ 13.8 49.0 +24.2 +5.3 -18.9
Starr, TX 9.6 3.4 +73.3 +60.1 -13.2
McKinley, NM 11.4 10.1 +46.9 +39.5 -7.4
Crittenden, AR 14.6 44.7 +14.9 +8.9 -6.0
Halifax, NC 11.9 39.3 32.3 26.9 -5.4
Edgecombe, NC 10.7 37.2 +36.2 +32.2 -4.0
San Patricio, TX 14.8 41.0 -20.7 -24.0 -3.3
Kings, CA 12.5 34.5 -14.9 -17.4 -2.5
Maverick, TX 12.1 3.1 +58.1 +55.8 -2.3
Tulare, CA 13.3 31.3 -15.0 -16.2 -1.2
Merced, CA 13.0 30.5 +8.7 +7.9 -0.8
Madera, CA 13.5 36.8 -17.1 -17.3 -0.2
Navajo, AZ 14.5 43.0 -7.8 -7.9 -0.1
Lea County, NM 12.6 40.6 -49.8 -48.3 +1.5
Apache, AZ 10.1 19.6 +34.3 +36.9 +2.6
Yuma, AZ 14.0 34.0 -12.6 -5.5 7.1
Ector, TX 14.3 38.3 -48.9 -40.6 +8.3
Imperial, CA 13.4 13.0 +32.0 +41.8 +9.7
Low-education majority-minority counties shifted toward Trump

Counties shown have a population of at least 50,000. Less than 50 percent of residents are non-Hispanic whites and less than 15 percent of the age-25-and-older population has a bachelor’s degree or higher.

Sources: American Community Survey, U.S. Election Atlas, ABC News

In short, it appears as though educational levels are the critical factor in predicting shifts in the vote between 2012 and 2016. You can come to that conclusion with a relatively simple analysis, like the one I’ve conducted above, or by using fancier methods. In a regression analysis at the county level, for instance, lower-income counties were no more likely to shift to Trump once you control for education levels.11 And although there’s more work to be done, these conclusions also appear to hold if you examine the data at a more granular level, like by precinct or among individual voters in panel surveys.

But although this finding is clear in a statistical sense, that doesn’t mean the interpretation of it is straightforward. It seems to me that there a number of competing hypotheses that are compatible with this evidence, some of which will be favored by conservatives and some by liberals:

  • Education levels may be a proxy for cultural hegemony. Academia, the news media and the arts and entertainment sectors are increasingly dominated by people with a liberal, multicultural worldview, and jobs in these sectors also almost always require college degrees. Trump’s campaign may have represented a backlash against these cultural elites.
  • Educational attainment may be a better indicator of long-term economic well-being than household incomes. Unionized jobs in the auto industry often pay reasonably well even if they don’t require college degrees, for instance, but they’re also potentially at risk of being shipped overseas or automated.
  • Education levels probably have some relationship with racial resentment, although the causality isn’t clear. The act of having attended college itself may be important, insofar as colleges and universities are often more diverse places than students’ hometowns. There’s more research to be done on how exposure to racial minorities affected white voters. For instance, did white voters who live in counties with large Hispanic populations shift toward Clinton or toward Trump?
  • Education levels have strong relationships with media-consumption habits, which may have been instrumental in deciding people’s votes, especially given the overall decline in trust in the news media.
  • Trump’s approach to the campaign — relying on emotional appeals while glossing over policy details — may have resonated more among people with lower education levels as compared with Clinton’s wonkier and more cerebral approach.

So data like this is really just a starting point for further research into the campaign. Nonetheless, the education gap is carving up the American electorate and toppling political coalitions that had been in place for many years.


VIDEO: Nate Silver discusses the method to our forecast


Footnotes

  1. If we’re being technical, “counties” include county equivalents such as parishes (in Louisiana), boroughs (in Alaska) and independent cities (in Virginia and other states).
  2. Unless otherwise specified, all figures in this article are as of the American Community Survey’s five-year estimates from 2010 to 2014.
  3. Aged 25 and older — this is the age many government documents start from when reporting how much of the population has a college degree.
  4. Specifically, the correlation coefficient is .69 between median household income and the share of the population 25 and older with at least a bachelor’s degree, among counties with a population of 50,000 or larger.
  5. Minimum population 50,000.
  6. This is a roughly middle-of-the-pack income — the Census Bureau estimates that the median U.S. income was about $53,600 in 2014, the last year included in this data set.
  7. Nashville is home to Vanderbilt University, but it’s small in comparison to Davidson County’s overall population.
  8. Also note that Clinton has a home in Chappaqua, New York, which is in Westchester County.
  9. Just so you know this isn’t an unproven assertion: 3 percent of the people who are employed in Westchester County said they work in the Census Bureau category called “arts, design, entertainment, sports, and media occupations”; only 1.9 percent of workers in Staten Island said the same.
  10. Namely, being on the outer edge of reasonable commuting distance from a major employment hub — Washington is a long commute from Calvert County in the same way that Manhattan is a long commute from Suffolk County — while also having some resorts and beaches.
  11. In fact, in some regression specifications — such as if you weight by a county’s population — lower incomes were actually associated with a slight shift toward Clinton, when education levels were held constant.

Nate Silver is the founder and editor in chief of FiveThirtyEight.

Comments