Examining Philly’s Voter Turnout

Local Elections and Turnout

The 2017 Philadelphia primaries are now mere months away. So far in this space, I’ve examined the role of randomized ballot position for the city as a whole and within wards. Today, let’s focus on election turnout.

Increasing turnout

A candidate’s total number of votes is the product of the total number of people voting and the percent of those voters who push that candidate’s button. Winning an election, then, can be decomposed into two tasks: (1) convincing voters in precincts with high turnout to vote for you, and (2) increasing turnout in precincts where voters are already most likely to choose you. Today, I’ll reflect on the second.

Of course, increasing turnout is easier said than done. How high can turnout go? What are the neighborhoods where turnout might be more malleable than others?

Overall turnout in Philadelphia’s Democratic primaries depends on the year and the type of election. Figure 1 shows the total votes in the primary for each election since 2003.[1] I’ve labeled years by the most prominent office in that 4-year cycle: Presidential, Mayoral, Gubernatorial, and District Attorney. The upcoming 2017 primary will be a D.A. election. First, you’ll notice that competitive elections receive much higher turnout than non-competitive ones. In 2008 and 2016, the presidential primary was competitive in Pennsylvania, while in 2004 and 2012 it wasn’t. In 2007 and 2015, the mayoral primary was competitive; in 2003 and 2011 it wasn’t. And the 2006 Gubernatorial primary wasn’t competitive—Ed Rendell was seeking reelection—while the 2010 and 2014 primaries were. The District Attorney shows a competitive election pattern too: 2005 and 2009 had competitive races, in 2013 Seth Williams ran unopposed. Interestingly, Senate races do not seem to affect turnout.[2]


If we limit ourselves to competitive elections, you’ll see a strict ranking of turnout by type of election: presidents have the highest turnout, followed by mayor, governor, and, lastly, D.A. The upcoming D.A. race projects to be competitive; as of early January, four challengers had declared for the primary. The best bet for turnout in the upcoming election is the 100,000 that came out in 2005 and 2009.


Where do people vote in D.A. elections? Figure 2 maps divisions’ average turnout in the 2005, 2009, and 2013 primaries, using the divisions number of registered Democrats in 2015 as the denominator.[3] People largely turn out for these elections in Center City, South Philly, and the Northwest, and particularly stay home in University City and North Philly.

These rates don’t indicate how much potential there is for movement. Is North Philly low because people don’t vote in general? Or are they disproportionately voting less in D.A. elections? Broadly, how malleable are these votes?

Answering this question requires an estimate of what is possible. Registered voters aren’t a good benchmark, especially for this low-attention election. Some neighborhoods might have residents who are less politically motivated or less able to afford the time costs of voting. Others have temporary or new Philadelphians who will vote in national elections, but will be hard to mobilize for local ones.

Instead, I’m going to use turnout in mayoral elections as a baseline for what is possible in the D.A. race. Mayoral elections generate a lot of attention in the city, and can draw out a large community of voters engaged in local politics. Because of this, they serve as a hard-to-reach goal for the low-attention D.A. years.

Returning to the wards with low D.A. turnout, is it that people in these wards never vote, or are they disproportionately voting less in D.A. elections? The answer: some of both. Figure 3 shows each district’s turnout for D.A. as a percent of its turnout for competitive mayoral primaries. In the entire city, competitive D.A. elections have 35% of the turnout of competitive Mayoral elections. But that varies across the city. West Center City, and particularly the divisions right on the Schuylkill River, have very high relative turnouts: voters here vote in D.A. races at over 75% the rate of mayoral races; a huge number for the city.

da_to_mayor Large regions of dark purples appear in North Philly and South West. These wards vote disproportionately less for D.A. races than for mayoral races. For this reason, they likely have the most room for turnout growth: they have the largest number of people who do vote for mayor, but who aren’t voting for  D.A.

As in last week’s ballot position post, readers familiar with Philadelphia might guess that class is driving these results. But it’s not as clear this time. The lower-income regions of West Philly actually seem to have city-average D.A. turnout, relative to mayoral turnout. This may speak to the power of the “get-out-the-vote” machine in West Philly. The purple North Philly regions are predominately Hispanic neighborhoods, and may have less-established machines.

Differential turnout appears to break down along racial lines.[4] Figure 4 plots each division’s number of votes in mayoral primaries on the x-axis, and its number of votes in the D.A. primaries on the y-axis. Divisions are assigned to a facet based on which ethnoracial group—Asian, Black, Hispanic, or non-Hispanic White—has the highest representation. The lines represent the points where the D.A. votes would be 35% of the mayoral votes, the city-wide average rate. Points below the line don’t just have fewer voters in D.A. primaries than in mayoral primaries, but disproportionately fewer than even the other divisions in the city.


Notice that predominately-Hispanic divisions (including the purple divisions in North Philly on the map) have both low numbers of voters to begin with—an average of 89 voters in mayoral elections—and lower turnout in D.A. elections than even their mayoral turnout would suggest—only 24% of those mayoral voters come out for the D.A. primary. Voters in predominately-Black and predominately-White divisions are centered about the line, meaning they vote at about the city-wide rate, with the exception that voters in White, high-turnout divisions vote in D.A. primaries at vastly higher rates. Turnout efforts in these divisions wouldn’t be fruitful, as there are only a small percentage of mayoral voters who aren’t already coming out on D.A. primary day. Instead, a candidate should focus on convincing the voters in these divisions, who are already voting, to vote for her.

A final way to think about turnout is in actual voter counts. Even if predominately-Hispanic wards in North Philly started voting in D.A. races at the same rates they do in mayoral races, they may not be a strong force because their turnout is low even for the mayor (they are at the left of the plot in Figure 4). Figure 5 maps the difference in raw vote counts between mayoral and D.A. primaries. This shows the potential gained number of voters if D.A. turnout increased to match the mayoral turnout, and is a combination of the relative D.A.-to-mayoral turnout and the overall voters in mayoral races to begin with. Center City and South Philly, Northwest Philly, and West Philly all have the largest counts of voters who show up to vote for the mayor, but don’t in the D.A. years.


What should a candidate do?

A 2017 candidate wants to increase turnout the most in neighborhoods where voters are particularly likely to vote for them. As a candidate, you want to target your turnout efforts to divisions with three traits: (1) voters who will vote for you if they do vote, and (2) a relatively high baseline number of voters, but for whom (3) mayoral election turnout is disproportionately higher than D.A. turnout. With regards to Figure 4, that would be divisions to the right and bottom of the plot.  While I’m calling this a “D.A. election,” this applies to candidates up and down the ballot.

Figure 5 identifies the districts where increasing turnout is most plausible: large voter increases are more likely in yellow districts that have a large gap, and less likely in purple districts that are already voting at similar rates in mayoral and D.A. elections, or have few voters to begin with. A candidate should take her personal map of support and overlay it with the map in Figure 5 to identify districts where she has high support and where turnout can be most increased. A candidate who can expect large support in West Philly, for example, would be well-served by pushing for turnout there; large gains in voters seems plausible. These are the districts to focus get-out-the-vote efforts on.


[1] As in previous posts, I only have the vote totals for Wards & Divisions, not individual voting data. I estimate the number of voters as the total number of votes for the office with the highest vote counts, and for which voters can only select one candidate. For example, the 2016 number is the total number of votes for President. The office is usually as expected: e.g. the total votes for President, for Governor, for Mayor, for D.A., etc.

[2] It is possible that this competitive-election effect is partly an artifact of my estimating turnout by votes in the largest race. It could be that people are still voting, but skipping the race with only one candidate, so overall turnout is higher, but this exact race is lower. This wouldn’t affect the rest of the analysis, as I just focus on competitive races anyway.

[3] I’ve excluded Division 63-07 in the Northeast, which has a turnout of 83%—almost twice that of any other division—and ruins the color scale.

[4] A separate regression analysis (not presented here) shows that the income of a division does not help predict the D.A. to Mayoral ratio, once you’ve accounted for race.


tannen photo

Jonathan Tannen, Ph.D., was previously a Director at Econsult Solutions, Inc (ESI). Jonathan’s dissertation research used GIS and large-scale computational techniques to develop a Bayesian method to measure the movement of neighborhood boundaries.

Share This