Behind the Numbers: The Real Electorate

R ecently released data from the Bureau of the Census now reveal a picture of the voting public in 1996 that is substantially different from the one that was available immediately after the election. Initially, analysts believed the electorate was dominated by upscale, college-educated voters. The census data show that view was wrong and suggest that political strategies based on that perception may not work.

There are three powerful sources of data on national elections in the United States: the Voter News Service (VNS) exit polls released on election night; the National Election Study (NES) conducted at the University of Michigan and released in April 1997; and the census survey of voter participation, conducted in November 1996 and released in full in October 1997. Each of these sources has its strengths and limitations. Used judiciously, they allow some fairly definitive judgments on voters and voter behavior in the last election.

In the immediate wake of the election, based on exit polls, the press reported that 43 percent of voters were college graduates. That "fact" appeared to support the New Democrats' strategy of moderate, values-based politics, since it is precisely such a politics that appeals to college-educated winners in the new economy.

But the census data say the true proportion of college graduates in the electorate was only about 29 percent. [See "The Census Shows a Different Electorate," below.] Conversely, more than seven in ten American voters (71 percent) lacked a four-year college degree, far outnumbering college graduates. More American voters (42 percent) had a high school education or less than had a four-year college degree or more. And the 29 percent in the middle—who have a junior college degree or "some college"—are closer in wages and income to high school graduates than to those with college degrees or more. In other words, Americans without college degrees are not just a majority, but a supermajority of voters.

graph--Census Shows a Different Electorate

W hy the huge difference between the exit poll and census results? The answer lies in a chronic bias of the exit polls: The highly educated are much more willing to fill out survey forms in polling places. The resulting overrepresentation of better-educated voters in exit polls is, literally, not to be believed. If 43 percent of 1996 voters really had college degrees, that would imply a turnout rate for college graduate citizens—based on their number in the population—of 102 percent. (Do they vote early and often?)

Of course, the census results are themselves open to criticism since they rely on self-reports of voting, and we know that more people say they vote than really do. But the census could overstate the fraction of noncollege-educated people in the electorate only if lower-education respondents were more likely to overreport voting. In fact, the validated vote data from the Michigan National Election Study show exactly the reverse: High-education respondents are more prone to overreport voting.

Subscribe to The American Prospect

So the exit poll data should not be relied upon to tell us about relative numbers of college-educated and noncollege-educated voters. The exit polls are useful, however, for their original, and still most valid, purpose—analyzing the attitudes and voting preferences of different groups of voters. Conversely, the census, with its much larger sample (about 90,000 respondents of voting age) and rigorous sampling methods, is a superb source for analyzing the demographic characteristics of voters and no use at all for assessing their attitudes and preferences (because the Census Bureau collects no such information).

A nother widely reported "fact" about the 1996 election was a sharp increase in Hispanic turnout. That estimate also came from exit polls, but the census data say Hispanic turnout actually went down by a couple of percentage points. Yet because Hispanics had increased their proportion of the population since 1992 and because Hispanic turnout went down less than the rest of the population in an abysmally low turnout year, Hispanics still modestly improved their representation among voters from 4 percent to 5 percent. In the long run, the growth of the Hispanic population could significantly alter the electorate and—given current voting patterns—provide a substantial boon to the Democrats. But in the 1996 election the magnitude of the change was not large and in the short run does not radically alter our picture of American voters.

Some analysts, particularly within the black community, also argued that black men—presumably mobilized by the Million Man March and related activities—voted in larger numbers in 1996. That too would be a development welcomed by the Democratic Party. But this finding, also based on exit polls, is contradicted by the census data, which show that black men suffered a decline in turnout (down four points) that was actually greater than the decline in turnout among black women (down three points). The representation of black men among voters hardly changed in 1996.

So what explains Clinton's relative success in 1996? The answer is implicit in the data about the preponderance of Americans with less than a college education. Consider the following results from 1996 exit polls: Clinton's support increased by about eight percentage points (over 1992) among the noncollege educated, but by only four points among the college educated. And among women, the disparity was even greater: Clinton's support went up more than 11 percentage points among voters without college degrees and just five points among college graduates. While Clinton may not have delivered much to them, the voters who have most often lost out in the new economy saw the Republicans as having even less to offer.

Interestingly enough, the class disparities among women were actually greatest among married women with children—the ones most closely resembling the fabled "soccer moms." Noncollege-educated mothers increased their support of Clinton by 17 points, compared to just three points among college-educated mothers. Other breakdowns make the same point even more starkly: Compared to 1992, working mothers without a college education increased their support of Clinton by 20 points, while their college-educated counterparts did so by only two points. The median household income of noncollege-educated women was just $34,000 in 1996. Thus the true swing voter starts to look less like an affluent Volvo driver and more like a Cavalier driver struggling to get by economically and build a better life. Clinton deserves credit for reaching so many of these women with his defense of such programs as Medicare, Medicaid, and public education. This is the success Democrats might effectively build on. The Volvo drivers, in contrast, responded only tepidly to Clinton, and there is little reason to build a political strategy around them.

So current and future success for the Democrats depends, most plausibly, not on increasing black turnout (now 11 percent of voters, just about their representation in the adult population) or even on increasing Hispanic turnout (now 5 percent of voters, just one percentage point under their representation among adult citizens) or on seducing the college-educated winners in the new economy. Instead it lies with the strugglers: the overwhelming majority of U.S. voters who lack a college degree and, more particularly, the huge pool of noncollege-educated white voters (57 percent of the electorate) whose support for the Democrats has improved substantially but remains shaky (just 44 percent for the President and 47 percent for the House in the 1996 election).

The Democrats especially need to increase support among nonunion whites without college degrees. The unionized noncollege whites are already solidly in the Democrats' camp: They voted 65 percent Democratic in the 1996 House elections. Partitioning the electorate into several groups underlines the challenge [see "Strengths and Weaknesses in the Democrats' Base," below]. While unionists, nonunion blacks, and nonunion Hispanics made up 33 percent of voters and supported House Democrats at rates of 63, 80, and 69 percent respectively, nonunion noncollege whites made up 45 percent of the electorate and only supported House Democrats at a 44 percent rate. Clearly, Democrats will need to win this group in 1998 and beyond if they are to take back the Congress.


Group Percent of Voters
in 1996
House Support
in 1996
Union households 22% 63%
Nonunion blacks 8 80
Nonunion Hispanics 3 69
Nonunion Asian/other 1 52
Nonunion noncollege whites 45 44
—men 19 40
—women 26 48
Nonunion college whites 21 36
—men 11 32
—women 11 41
Source: Author's analysis of 1996 VNS Exit Poll and 1996 Census Voter Supplement data.

If the Democrats can simply break even among this group, they would increase their share of the House two-party vote to 53 percent. That would almost certainly be enough to win a majority of seats and would bring popular support for House Democrats back to pre-1994 levels. (The Democrats averaged 53 percent of the two-party vote throughout the 1980s and early 1990s.) The hardest part of achieving such a split among nonunion whites without college degrees will be moving the men. While 48 percent of the women supported Democrats in 1996, only 40 percent of the men did so. Of course, if women move up more than two points, men could move up fewer than ten and still produce an even split among the group as a whole.

What could move these voters? Consider some economic data. The median wage of noncollege white men at the end of 1996 remained 10 percent below its 1989 peak, even in the seventh year of an economic expansion. And the median wage of noncollege white women, while showing less erosion over this period, was still 3 percent below its 1989 peak. Household income has risen overall for three years running, but for noncollege white men, the median household income at the end of 1996 was 2 percent lower than in 1989 and, for women, the corresponding figure was 3 percent. So, the material basis for a populist message among these voters is there.

And the political basis for that message may be emerging. One can argue about the economic merits of fast-track legislation, but the politics should be clear. Public opinion, particularly among nonunion noncollege whites, has been hostile to the measure. In a recent poll, 70 percent of nonunion noncollege whites opposed fast track and 69 percent opposed expanding NAFTA to other Latin American countries [see "Why the Democrats Might Win Back More Reagan Democrats," below]. Thus, contrary to what many commentators said at the time, Democrats in Congress shored up their credentials with swing voters by opposing fast track. It was a vote that may help them prevail among precisely the voters whose support they need to take back the House.


White Men
White Women
Oppose fast-track legislation 70% 69%
Oppose expanding NAFTA 66 72
Support improved education via national testing and raising school standards 58 67
Support increasing spending on public schools 57 65
Favor increasing investment in programs like education with higher taxes on the wealthy and corporations 62 77
Favor eliminating subsidies to big corporations and closing tax loopholes for the wealthy 64 77
Support protecting Medicare and Social Security 76 87
Source: Author's analysis of July 18-22, 1997, Hart/AFL-CIO survey (fast track, expanding NAFTA to Latin America); 1996 National Election Study (increasing school spending); October 12-16, 1997, Lake/Greenberg/EMILY's List survey (raising school standards, eliminating subsidies, protecting Medicare and Social Security); and November 5-8, 1996, Greenberg/Campaign for America's Future survey (increasing investment in domestic programs).

The polling data suggest several other areas where nonunion noncollege whites, both women and men, are particularly open to Democratic initiatives. On education issues, while women in the group have stronger views, the men show support for a more activist approach. Both men and women in this group exhibit strong hostility to corporate welfare and attach a very high level of importance to protecting Medicare and Social Security.

Thus, not only are nonunion noncollege whites critical to the success of the Democrats, these voters can be reached with the right program. And this especially includes the men; Democrats who cede them to the Republicans are accepting permanent minority status. Will the Democrats take up the challenge of speaking to this group's interests or continue to rely on platitudes about the virtues of the new economy? The 1998 elections should begin to tell the tale.

You may also like

You need to be logged in to comment.
(If there's one thing we know about comment trolls, it's that they're lazy)