Earlier this month, Republicans were shocked to find out that, no, the diversity of the 2008 electorate didn't decrease, it increased. Some Republicans did seem to at least understand the long-term trend prior to the election: As Sen. Lindsey Graham infamously put it, “We’re not generating enough angry white guys to stay in business for the long term.”
There's still plenty of white guys around, of course, just not enough of the ones who are angry about the things Republicans want them to be angry about. But the pool of white guys from which to draw angry white guy voters is shrinking over time.
The census numbers provide a more detailed look at this phenomenon than the exit polls. The census numbers are plotted below for elections from 1972 to 2008. Each bar on the graph represents an age group during a certain election. We see in the upper left-hand corner, for example, that the oldest voters of the 1970s elections were 90-95 percent non-Hispanic white. On the opposite end of the curve, the youngest voters in the 2008 elections (lower right-hand corner) were just 67 percent non-Hispanic white. The 2012 census data are not yet available.
Update: The graph above is a complicated one. An enlarged, simplified version, with points instead of bars, can be found here.
We can see above that the older voters leaving the voter rolls were about 85 percent white in 2008, while the youngest voters were a little more than 65 percent white. This is the most obvious process and the one that is most often discussed. It is not the only thing that is changing the demographics of the electorate, though.
Isolating the Boomers
In the graph above, you may notice that for the population born after about 1920, the electorate has grown more diverse over time for a given age group. One cohort can be tracked from their 20s to their 60s—those born between 1945-1960 (approximately the Baby Boomers); this group is highlighted in blue.
We can pull the data for the Boomers into a separate graph to see the change over time more easily. Note that, as you can see from the first graph, the census data do not usually match the birth year range exactly, so numbers shown below will be approximate.
However, you might notice that although the proportion of the Boomers who are black has remained fairly steady (declining slightly), the proportion of the Boomer electorate that was black (in the previous graph) still increased from 1972 to 2008. The reason for that is increased participation.
Crude turnout is much lower among other racial groups, partly because of lower rates of citizenship, but still increasing over time, as we would expect as a population ages. Turnout among Latino citizens, for all ages, was about 50 percent in 2008.
Enough with the Boomers!
We can also look at Generation X, the electorate born between 1965-1980, although there's only half as much data. This generation has also become more diverse over time—again, both the electorate and the population—and has done so even more rapidly than the Boomers.
Republicans were confident that decreased turnout for minorities and/or increased turnout for white voters would result in a whiter, more Republican electorate this year compared to 2008. The numbers show that, historically, that just hasn't happened in any sequential presidential election since the first census numbers published in 1972. Turnout does go up and down among different racial groups from one presidential election to another, but racial diversity is pretty consistently increasing as any changes in participation are swamped by immigration and replacement.
Remember, though, this post is only about presidential election years. Midterm elections have lower participation rates, and the dropoff is particularly acute among the younger (more diverse and Democratic-leaning) voters. The electorate in 2014 will almost certainly be less diverse than 2012, just as the 2010 electorate was less diverse than the 2008 electorate.
Fine print: Census numbers show an electorate that is a few percent more white than exit polls report. Also, the census has changed the way it collects data several times over the time period of 1972-2008, and this will have small effects on the numbers. For example, prior to 2000, the Census Bureau did not report numbers for the category of non-Hispanic white, so I had to estimate by subtracting the Hispanic or Spanish-origin population from the white population. This estimate was off by about 0.1-0.5 percent in the year 2000 when it could be compared to the non-Hispanic white numbers given by the census.