Leading up to the 2008 elections, the state of California required some voters to vote by mail while others could continue to vote in person. It turns out that voter mobilization drives are less effective when people vote by mail, see the full study to explore why.
Face-to-face voter mobilization drives, aka “get out the vote” (GOTV) canvassing, affects each voter differently depending on why it is that individual votes in the first place. Arceneaux, Kousser, and Mullin (2012), examine whether responses to these drives are affected by the processes by which people vote, such as by mail or in person.
An individual’s incentives to vote are shaped by the probability of affecting the outcome of the election, the gains from their party’s winning, the costs associated with voting, and the fulfillment of their civic duty to vote. Civic duty motivations include the personal enjoyment that some derive from voting, but also the social pressure to conform to the norm that one should vote. Those who get higher personal enjoyment from voting are high-propensity voters – that is, they are likely voters – while those for whom the personal enjoyment is small are low-propensity voters. Voting by mail, which can be done with more anonymity than voting in person, reduces the costs associated with voting, but it also removes the social rewards one can get from standing in line and casting a vote. Previous studies have shown that GOTV canvassing has the greatest effects on the low-propensity voters and the effects are only minimal on high-propensity voters, most of whom have already decided to vote before being contacted through GOTV canvassing. On the other hand, medium-propensity voters exhibit a moderate boost in turnout.
Arceneaux and colleagues examine the effect of GOTV canvassing on people who vote by mail versus in person. They test the extent to which personal enjoyment as opposed to social rewards—which are lower in voting by mail than when voting in person—shape whether high, medium, and low-propensity voters actually vote. They expected to find that social pressures have the strongest effects on medium-propensity voters – that is, those who pick and choose when to vote – because low-propensity voters will require more direct pushes (e.g., GOTV canvassing) and not just subtle social pressures.
To figure out the effects of different voting systems, the research team randomly 1 assigned both polling place and voting by mail precincts into a treatment group that received GOTV canvassing for the 2008 general election and a control group that did not.
Table 1:Experimental Design:
Natural experiment carried out by registrar
Vote in Person Precincts
Vote by Mail Precincts
Randomized Field Experiment
Control (no GOTV contact)
22,769 voters in 51 precincts
5,955 voters in 51 precincts
Treatment (GOTV contact)
22,892 voters in 50 precincts
6,825 voters in 50 precincts
45,661 voters in 101 precincts
12,780 voters in 101 precincts
Note: Treatment conditions for cluster randomized design. This is a factorial design in which one dimension was assigned “as-if” randomly by government and a second was assigned truly randomly by the research team.
The team matched voting by mail precincts with the adjacent traditional voting precincts that are most demographically similar in order to ensure that differences they find are more likely due to the voting methods employed as opposed to being driven by demographic differences. This matching technique helps to improve the precision of the effects being studied. From the 101 matched pairs of districts in California, they randomly selected 50 pairs to receive GOTV canvassing in which voters answering at the door were read a GOTV script, while postcards were left at the homes of those who did not answer (see Table 1).
GOTV canvassing activities have a larger effect on those who vote in person than on those who vote by mail, but only among those for whom voting provides stronger social rewards. Meaning GOTV canvassing may not be as worthwhile for political campaigns in settings in which individuals receive fewer social rewards from voting.
Overall, voters living in in-person voting precincts that were canvassed through GOTV efforts were 1.7% more likely to vote than those voters in in-person voting precincts that did not receive GOTV canvassing, while the effect of GOTV canvassing was somewhat weaker in voting by mail precincts. The estimated effects on the treated — those that received the messages — is even stronger (see Fig 1 below). However, neither of these overall effects is statistically significant.
Figure 1: Treatment Effects (ATT: Treatment on the treated)
GOTV canvassing had a strong positive effect on low-propensity voters however, who were 5.5% more likely than those in the control group to turn out in voting-in-person areas. GOTV also had a strong effect on low propensity voters in mail areas, and there was not a significant difference in the intent-to-treat effect 2 between traditional and vote-by-mail precincts. It had little effect on high-propensity voters in either type of voting system, whether voting by mail or voting in person. Things were quite different for medium-propensity voters however: medium propensity voters who voted in person and received GOTV canvassing were 3.1% more likely to vote than those who voted in person that were not canvassed, whereas those in voting by mail precincts were actually 4.2% less likely than their control group counterparts to turn out (see Figure 2 below).
Figure 2: Differential Effect of GOTV for low and medium propensity voters in each condition
1. Randomized field experiments have been used extensively in recent political science studies of the effects of grassroots mobilization on voter turnout, and have the advantage of assigning a treatment (e.g., being canvassed through GOTV), and random assignment to treatment and control groups helps to reduce bias in estimating the causal effect of GOTV on voting likelihood.
2. Intent-to-treat (ITT) estimates the effect of random assignment – that is those who the researchers intended to treat through GOTV canvassing – on probability of voting.
GOTV canvassing, a very commonly used political campaign tool, had limited effectiveness on people who voted by mail. The fact that GOTV canvassing decreased turnout for medium-propensity voters who voted by mail and had almost no effect on high-propensity voters will come as a surprise to political campaigns ad suggests a need to recalibrate mobilization strategies. An implication for American policymakers who want to increase political participation is that making voting more convenient may also make mobilization harder for some voters. In high-salience elections, GOTV can boost turnout among low-propensity voters in both vote-by-mail and traditional voting systems, though GOTV may be less effective for medium-propensity voters. Policymakers must consider more than the factors affecting timing and convenience of casting the ballot. They must also consider:
The context of the election, such as the campaign activity and salience, which conditions the impact of voting by mail systems, which are aimed at reducing the costs associated with turning out.
The fact that while enabling individuals to vote by mail reduces certain barriers to voting, it also reduces the social rewards that individuals gain when they are less intrinsically motivated to vote.
The fact that the type of voting system employed may be sufficient to swing an election.
Employing electoral strategies that increase the extrinsic social rewards that voters reap from turning out.
Looking beyond the United States to the developing world, in which voting may be more difficult (thus suggesting a greater need for new opportunities in place to facilitate easier voting) but also fraught with intimidation, the voting (whether by mail or in person) and mobilization tools that policymakers choose to adopt will have important consequences with respect to who ultimately turns out to the polls and whose voices are left unheard.