BY JAMES G. KANE
As Broward Democrats lament the relatively low turnout for this gubernatorial election, reasons for this decline as expressed by political operatives and pundits almost universally point to a lack of an effective “get out the vote” campaign. Yet, as measured by Broward’s early voting figures that portion of the campaign seemed quite successful. Over 54% of Broward’s voters cast their ballot prior to the election, either by absentee or at early voting locations.
Early Voting Does Not Always Produce Higher Turnouts
Although early voting was designed to increase voter participation by allowing voters the convenience of participating in the election over a longer period of time, many studies have shown little or no significant increase in election participation. In fact, one study (Election Laws, Mobilization, and Turnout: The Unanticipated Consequences of Election Reform, Burden, et.al., 2010) showed a “decrease in voter turnout caused by that early voting has created negative unanticipated consequences, reducing the civic significance of elections for individuals, and altering the incentives for political campaigns to invest in mobilization.”
Put in practical terms, early voting has decreased the value of Election Day voting as a community event that energized voters’ civic pride and participation. As a personal example, my Irish father never worked on Election Day but instead worked on getting everyone he could to the polls. When he was done, he would go to the pub and drink beer and argue with his buddies about who would win (I have, of course, carried on this tradition). Early voting has minimized the importance of Election Day and the value of participating in a community event.
In addition, many campaigns rely on early voting programs almost exclusively that, more often than not, depend on getting people to early voting locations who almost certainly would have voted on Election Day.
In other words, we have just moved the likely voters to the early voting column and left the less likely voters to vote on Election Day.
Trends in Broward Mid-Term Participation
In this mid-term election, 44.45% of registered voters participated, well below the state average of 50.3%. Media reports tend to suggest that this is an unusually low turnout for Broward County in a mid-term election. Reality is, however, that this is a typical mid-term election participation rate at least since 1998, as Figure 1 shows.
As this Figure demonstrates, Broward’s participation rate has been in the mid to low forty percent range since Jeb Bush was first elected governor. The turnout rate in the 1980’s and 1990’s, however, were in the fifty and sixty percent level. In four years, Broward’s participation rate dropped a remarkable 16%. The question, of course, is why did this happen?
In their classic political work, Who Votes?, Wolfinger and Rosenstone (1980) empirically demonstrated which demographic groups participated regularly in American Elections. When controlling for other variables like party, income, etc., they found the two most influential demographic traits were education and age. As expected, the better educated voter was more likely to show up at the polls than those with the least amount of school years. And older folks were more likely to vote than those who were younger.
So I became curious if some fundamental shift in Broward’s demographics had somehow altered the political landscape and contributed to the decline of mid-term voting. Using election and census data going back to 1986, I explored whether there was any statistical pattern that could affect turnout over such a short period of time.
First, the decline of mid-term turnout is not associated with Broward’s educational levels. In fact, Broward residents are much better educated now than they have ever been. The percentage of residents with a Bachelor’s Degree increased from 18% in 1990 to 30% in 2014. According to the Wolfinger and Rosenstone model, turnout should have increased during this period.
Changes in older voters during this period, however, show a significant decline. In 1986, some 23.5% were 65 or older, but by 2014 this segment of the population dropped to only 15%, as shown in Figure 2 below.
The connection between turnout and the decline of older voters is quite remarkable. Correlation analysis between the two variables demonstrates this fact. The correlation (1 equals perfect correlation/0 equals no correlation) for age and turnout shows an almost perfect relationship at .948 (sig. <.0001 level) as graphically depicted in Figure 3.
The blue bars represent the percentage of residents 65 and older, while the red line is the turnout for mid-term elections since 1986. As the figure displays, as older voters declined so did the percentage of voter turnout.
As I often tell my students, however, correlation is not necessarily causation. We can conclude, of course, that age is likely causing turnout decline and not the other way around. Thus, we can show how the significant and independent effect of age has on turnout by regressing (simple regression) the percent of 65 and older residents on the dependent variable, the percent of turnout. In Figure 4, we show a direct linear the relationship between older voters and turnout.
(Warning: there is some math involved below)
This scatter plot demonstrates that as the percent of older voters increases so also does the percent of turnout increase as well. And the statistical relationship is strong with an R square nearly .90 (1.0 is a perfect relationship), which measures the amount of variation in the data.
But one variable alone, like age, could be masking other more important variables, such as party or incumbency, to affect turnout. To control for these factors, we have to create a statistical model that measures the effect of other variables (independent variables) on voter turnout (dependent variable).
Using multiple regression, we can control for other independent effects on voter turnout and estimate the impact of age compared to other factors. For example, an incumbent Democrat might stimulate more turnout in Democrat rich Broward rather the number of older voters (such as Lawton Chiles in 1994).
To control for this possibility, I have included two more variables: candidate incumbency and the party of the winning candidate (governor), along with the percentage of voters 65 and older. The R square for this equation is .967, which statistically means the model is an excellent fit in predicting turnout.
But the proof of the model’s accuracy is in how well it can predict past elections’ turnout. Using the regression model’s coefficients, we can test its accuracy on previous elections (model equation and coefficients available on request). In Table 1, the model estimates for previous elections was quite accurate, generating an average error of only 0.746%.
ELECTION ACTUAL TURNOUT % PREDICTED TURNOUT % ERROR %
TABLE 1 TOTAL ERROR= 5.97; AVG ERROR 0.746
This table contains the actual turnout for the mid-term elections in Broward from 1986 to 2014. Up until 1998, Broward’s turnout rate was 54% to 61%, which included two Democratic victories. Our model predicted a turnout that was within 2% of the actual for each election cycle. As expected, the percentage of older voters had the most effect (beta) on turnout of the three independent variables.
In 2004, Florida authorized the early voting program as we know it today. To test the hypothesis that early voting does not increase turnout, at least in mid-term elections, I included this variable in the final model and, as expected, it had no effect (positively or negatively) on turnout at all when the other variables were included.
In plain talk, early voting is a bust when it comes to increasing turnout.
Broward County’s 65 and older population dropped below the statewide average in 2010 according to the Census Bureau. In that year, Florida’s oldest population group was 17% of the total population. Although projected estimates of the retiree population in Broward vary, it is unlikely that it will ever achieve its past glory. Between housing costs and transportation difficulties for seniors in urban areas, the percent of older Broward residents will likely stabilize over the next few years or, possibly, decline even further. Older retirees are now moving to central and panhandle areas of Florida and these retirees tend to favor the Republican Party.
Consequently, for the foreseeable future, Broward candidates and parties will have to deal with increasing turnout using other means than relying on older voters and early voting (including absentee voting) to increase Broward’s turnout to at least statewide averages during mid-term elections. In statewide mid-term elections, Democratic candidates can no longer rely on Broward’s voters to save the day.
If you don’t believe me, just ask Charlie Crist.
(James G. Kane is a long-time local political operative and an adjunct professor in the University of Florida Department of Political Science’s Political Campaigning graduate program. For more information on Kane click here.)