In my previous post I made some projections about the likely percentages of Democratic, Republican, and unaffiliated and other voters in the upcoming general election, projections that I thought were reasonably well supported by the evidence. In this post I wanted to publish more data for anyone who’s interested, and also explore a little further a topic that I touched on in the last post.

First, the data. From various Maryland State Board of Elections reports1 I’ve collected together two related data sets, one of which I discussed in the last post and one that’s new:

  • Turnout statistics for general elections in Howard County from 1988 through 2008 (available as a Google spreadsheet, Excel file, or a text file suitable for loading into R). Note that there’s one very minor problem with this data: In one of the years a total reported by the Maryland State Board of Elections for all parties and unaffiliated and other voters was off by 3 compared to the sum of the individual numbers for all parties and unaffiliated voters. I made a correction to eliminate the discrepancy (adding three voters to the unaffiliated and other category), but unfortunately have now forgotten which year I made this correction for. If I ever have time I’ll go back and track this down and restore the numbers to what was actually reported (which I think would have been the best course of action).
  • Turnout statistics for primary elections in Howard County from 1988 through 2008 (like the other results, available as a Google spreadsheet, Excel file, or R-compatible text file). Unfortunately this data set is missing data for 1996, a year in which the Maryland State Board of Elections apparently didn’t report turnout numbers for the primary. Again, if I ever have time I’ll try to track down these numbers somewhere else.

The open question has to do with the relative percentage of unaffiliated and other voters in recent elections. (Incidentally, I use the term “unaffiliated and other”because it’s more accurate than “independent”; this group does include some people who belong to political parties, albeit small ones, and as I’ve previously written there are good reasons for believing that unaffiliated voters actually have consistent partisan preferences.)

One of the things I found interesting is that the percentage of those voting in general elections who are unaffiliated has shown an almost perfectly linear upward trend over the years, closely matching the growth in the number of voters who register as unaffiliated. By contrast the share of Republican and Democratic voters has fluctuated back and forth: in some elections Republicans form a relatively greater share of those voting, at the expense of the Democratic share, and sometimes the situation is reversed.

Presumably Democratic and Republican turnout (and hence relative share of those voting) is influenced by party members’ feelings about their party’s candidates on the ballot, their general level of enthusiasm about the election, and other factors. Why don’t these factors likewise affect “independents” and cause their share of the voting population to fluctuate as much as Democrats and Republicans?

There are at least two possible answers I can think of. First, it’s possible that any fluctuations in the share of unaffiliated and other voters are not that apparent because their share is much smaller than that of the two parties (about half that of Republicans, and just over a third that of Democrats), and also because fluctuations from election to election are masked by the strong secular growth in the number of unaffiliated and other voters.

The second possibility is more interesting: If most “independent” voters are actually closet partisans, then their propensity to turn out for a given election will likely mirror that of members of the particular party they prefer. In other words, unaffiliated voters who lean Democratic will tend to turn out at the same rates as actual Democrats, and unaffiliated voters who lean Republican will tend to turn out at the same rate as Republicans.

Thus, for example, if Republicans were more enthusiastic than Democrats in a given election cycle and turned out in greater relative numbers, then Republican-leaning unaffiliated voters would turn out in greater relative numbers as well. However the increased turnout on the part of those voters would be offset by decreased turnout on the part of unaffiliated voters who lean Democratic. The result would then be that the overall reported share of all voters who were unaffiliated would remain relatively constant, while “under the covers” (so to speak) the actual composition of that group of voter (i.e., Republican leaners vs. Democratic leaners) would change from election to election.

It’s a nice hypothesis, but how to prove it? I’m not really sure, in the absence of detailed data about the partisan preferences of unaffiliated voters. However one suggestive piece of data is that turnout of unaffiliated and other voters is slightly more closely correlated with Democratic turnout than with Republican turnout.2 This is exactly what we’d expect to see if the partisan preferences of unaffiliated voters mirrors that of Howard County voters overall.

But of course “correlation is not causation,” as the saying goes, so for now this is just an interesting idea. If anyone else is interested in this question feel free to take the data and do your own blog post.


Sarah - 2010-10-21 11:08

A post with footnotes! Be still my heart. I think your hypotheses are good ones, and it could even be both. They both sound reasonable, but you’re right– how to prove it? Sometimes it’s maddening what available data cannot tell us.

hecker - 2010-10-21 12:45

Sarah, thanks for stopping by. On the footnote thing, based on a scientific education and my open source experience I strongly believe in the principle of “showing your work”. It helps me convince myself when I think I’m right, and if I’m wrong someone can spot my errors and I’ll learn something in the process.


  1. For the general election turnout I used reports from the Maryland State Board of Elections for 1988, 1990, 1992 (for all voters, the Democratic, Republican, Libertarian, and Alliance parties, and unaffiliated voters), 1994, 1996 (for all voters, the Democratic, Republican, Libertarian, Reform, Natural-Law, and Taxpayers parties, and unaffiliated voters), 1998, 2000 (for all voters, the Democratic, Republican, Libertarian, Reform, Green, and Constitution parties and unaffliated voters), 2002, 2004 (for all voters and broken down by party), 2006, and 2008.

    For the primary turnout I used the Maryland Board of Elections primary election turnout reports for 1988, 1990, 1992 (for all voters, for the Democratic, Republican, Libertarian, and Alliance parties, and for unaffiliated voters), 1994, 1998, 2000 (for all voters, for the Democratic, Republican, and Libertarian parties, and unaffliated voters), 2002 (for all voters and broken down by party), 2004, 2006, and 2008. Note that the Maryland Board of Elections web site does not include turnout figures for the 1996 primary election. ↩︎

  2. Using the R-compatible text file for general election turnout linked to above, cor(PctTurnoutD, PctTurnoutOther) is 0.982 while cor(PctTurnoutR, PctTurnoutOther) is 0.939. ↩︎