Mark S. Tenenbaum

What share of US voters are “independent?” There are two primary methods used by scholars to answer this question.

The first method involves looking at voters’ party registration. Most states make voter registration data publicly available, so one can simply calculate the share voters who are registered as independents. This method, which is employed by the Current Population Survey and the UVA Center for Politics, finds that the share of registered independent voters has remained around 30% since the mid-2000s.

The second method involves simply asking voters how they identify, regardless of their registration status. This method, which is employed by Gallup and Pew, shows a gradual increase in share of voters who describe themselves as independent, increasing from 30% in the mid-2000s to nearly 40% today.

Both methods have costs and benefits. Voter registration data is not nearly as elastic to changes in political attitudes as polling responses. This is partly because most states have either fully or partially open primaries; in the former, voters can cast their ballots for either party regardless of their registration status; in the latter, independent or unaffiliated voters are free to vote in either primary. Hence, since voter registration is not a barrier to political participation for a significant subset of US voters, it is reasonable to expect “lags” between changes in voters’ political ideologies vs. when voters finally get around to updating their registration statuses.

On the other hand, some data indicates that voters consider themselves far more “independent” than their voting patterns suggest. For example, Michigan State University political scientist Corwin Smidt found that today’s “independent” voters vote more reliably for one political party than “strong partisans” did in the mid-1970s. Given this fact, there is good reason to be skeptical that there could be a simultaneous increase in both the share of the electorate that identifies as independent and in the share that votes reliably for one political party. (Note: if this is ultimately the case, one plausible explanation could be an increase in negative partisanship — but such a discussion is worthy of its own post.)

Although both methods differ in their “top line” estimates of the share of independent voters, their findings converge on a central fact: most independent voters are not very “independent.” That is to say, most independent voters “lean” towards one political party, while only a subset of them — likely around 10% – 12% of the US electorate — can be considered genuine “swing” voters.

I made the following data visualizations using CCES common content data from 2006 – 2019, which calculates survey weights based on voter registration status.

This first image displays the share of voters who are registered with the Democratic party, the Republican party, and as independents, as well the partisan lean of independent voters.

In regards to party registration, we can see that the gap between Democratic and Republican registration widened around the 2008 general election, narrowed in the aftermath of the 2010 midterms, and then has remained relatively stable since the 2012 election.

Besides a nearly five-point dip before the 2008 election, and a nearly seven-point dip in 2018, the share of the electorate registered registered as independent has largely fluctuated around 30%. However, the share of “true” independent voters — voters that do not reliably vote for one party — increased from under 30% of independent voters in 2006 to about 40% in 2019.

Notably, these figures are not equally distributed across states.

There is no obvious partisan skew (based on states’ 2020 electoral college outcomes) in the share of a state’s electorate that identifies as independent, nor in the proportion of independents that “lean” towards either major party.

Where significant skews can be seen, however, is when comparing independent affiliation between religious groups.

While most religious traditions follow the same pattern of independent registration — a modest increase before 2012 and a modest decrease after — there are sharp differences in the magnitude of independent affiliations across groups.

These differences become more obvious when ploting them together for a single year:

Religious groups more likely to support the Republican party — Protestants, Roman Catholics, and Mormons — are the least likely to identify as independent. Conversely, non-religious voters — atheists, agnostics, and those who identify as “nothing in particular” — and voters who identify as “something else” are the most likely to be registered as independents.

These patterns mostly hold when categorizing respondents by race. The follow chart looks at white, black, and Hispanic registered voters who identify as Protestant, Roman Catholic, Atheist, Agnostic, or “nothing in particular.” Data for other racial-denominational intersections (e.g., Hispanic-Hindu) were too thinly sampled to include. 2018 data was used as CCES samples are larger in election years.

Notably: (1) Black Catholics are about 8 percentage points more likely to identify as independent than white and Hispanic Catholics; (2) Hispanic Protestants are about 5 percentage points more likely to identify as independent than white or black Protestants — an intuitive finding given that Hispanic Protestants are not nearly as homogeneous in their voting behavior relative to white or black Protestants; (3) among voters who identify as “nothing in particular,” Hispanics are about 10 percentage points more likely to identify as independents than blacks and are about 8 percentage points more likely than whites; (4) white atheists are about 8 percentage points more likely to identify as independent than black or Hispanic atheists; (5) Black agnostics are about 10 percentage points more likely to identify as independent than Hispanic or white agnostics; (6) while the proportion of “true” independents remains relatively level across races, one exception is among Protestants — Hispanic Protestant independents are more likely to be genuine swing voters than black or white Protestants.

Here is my key takeaway from the breakdowns by religion: religious communities can reinforce partisan attachments — e.g., GOP affiliation among white Evangelicals or Mormons and Democratic affiliation among black Evangelicals — which is why religious Americans appear to identify with the Democratic and Republican parties at higher rates than non-religious Americans. Obviously, this takeaway requires far more unpacking. Perhaps I will explore it further in a future post.