Differences in Registration and Turnout By Age, 1996 – 2006 (PDF)
Note: I have an updated presidential elections chart.
Draft, January 2008
I examine the differences in registration, turnout, and turnout of registered individuals across ages, separately for presidential and midterm elections from 1996 to 2006. The patterns between these variables are complex, and traditional analyses using cross-tabulations, summary statistics, or regression oversimplify these relationships. As a complimentary analysis, I visualize Current Population Survey data for each election using polynomial approximations that facilitate review of trends. In particular, this method highlights a steeper registration curve for the 18-24 age group, higher than expected turnout for registered 18 year olds (particularly in midterms), and declining participation for the 70+ age group.
Data for this project come from the November voting supplement of the US Census Bureau’s Current Population Survey for the most recent six federal elections, spanning from 1996 to 2006. I examine three basic indicators of participation in elections: registration rate, turnout rate, and turnout rate out of those who are registered. All statistics are computed for the eligible voter population. Rather than displaying the point estimate at each age, which would be noisy and overly reflect sampling error, I use a six-term polynomial smoother to approximate the distribution and display a clean curve.
Turnout for the past three presidential elections has maintained its overall shape, with 1996 and 2000 having almost exactly the same curves. In 2004 we see a notable turnout increase across all ages, but with the largest increases occurring in the youngest voters. Registration levels were also similar for 1996 and 2000, with a slightly lower level in 2000 for most ages. 2004 shows increased registration across all ages compared to 2000, but compared to 1996 only the youngest Americans had higher registration rates in 2004. Turnout among registered individuals shows consistent increases across all ages from 1996 to 2000 and from 2000 to 2004, and that registered youth are turning out at rates nearly equal to older Americans. The statistic displays an interesting lack of linearity and multiple inflection points.
Recent midterm elections display a surprising consistency in registration and turnout. Although increases for youth are visible in 2006, the magnitude is smaller than the presidential election increase. The statistic that does show large variation is turnout of registered voters. In 2002 there was a substantial increase in the turnout of the very youngest registrants (18-22 year olds), and in 2006 there was a more balanced increase across 18-29 year olds. As in presidential elections, the turnout of registered voters shows several inflection points over age. Overall we see a stronger relationship between age and political participation in midterm elections compared to presidential elections, and this is especially distinct for registration rates of 18-29 year olds – note the steepness of the registration curve for this age group.
This analysis has shown the complexity of the interaction between age and political participation, which is not as simple as a positive linear relationship and varies between presidential and midterm elections. The method of using polynomial smoothing to display large amounts of data makes this clear. For any future work on this line of research I will utilize a more advanced, localized smoothing function (which will particularly improve estimation at the edges) and will provide additional numerical tables as supporting information.
i David King used a similar polynomial smoothing technique to compare voter turnout in 2004 and 2000. See Young Voter Mobilization in 2004 (2006), a report conducted by Ryan Friedrichs for Skyline Public Works.