Nonprofit Messaging and the 2020 Election: Findings from a Nonpartisan Get-Out-The-Vote (GOTV) Field Experiment

1 Introduction

While US voter turnout has increased in each of the past three national elections, voter participation poses an enduring problem for representative democracy. Voter participation in the U.S. reaches its highest point during presidential election years when on average, just a little more than half of the eligible electorate casts a ballot.^[1] Moreover, decades of studies in political science have demonstrated that participation is not equally distributed among the population, but rather rates of turnout increase linearly with income, education, and age (Brady, Verba, and Schlozman 1995; García Bedolla and Michelson 2012; Laurison, Brown, and Rastogi 2022). The 2020 turnout rate for citizens highlights this participation gap, indicating that the voting rate among those in the lowest income households was half the rate (39%) of those living in the highest income households of $150,000 or more (80%) (Census 2021).

This participation gap is continually exacerbated by political campaigns that use micro-targeting strategies to reach registered voters with a known voting history, factors which often correlate with age, education, and income. Results from a study of Ohio voters between 2004 and 2012 by Nickerson and Rogers (2014) showed that microtargeting led to disparities in contact between higher and lower propensity voters.^[2] By 2014, the rich-poor contact gap was the largest in history, leading researchers to label micro-targeting strategies as a form of passive voter suppression (Ross and Spencer 2019). These patterns in voter contact and resultant turnout have proven persistent over time, posing serious consequences for representation, as elites drive election outcomes and resulting policies of elected lawmakers are often inconsistent with the interests of the majority.

Scholars have increasingly pointed to community-based nonprofits as a solution to the turnout gap (Brown 2016; García Bedolla and Michelson 2012; Marwell 2004, 2007; LeRoux 2007; LeRoux and Krawczyk 2014). Often situated within low-income communities and/or serving underrepresented groups, nonprofits can act as civic intermediaries and are uniquely positioned to mobilize underrepresented citizens for political action (Brown 2016; LeRoux 2007).^[3],[4] Through their nonpartisan voter registration and get-out-the-vote (GOTV) efforts, community-based nonprofits can help to remedy the participation gap, promoting higher turnout among low-income citizens and marginalized groups. Previous experimental research has confirmed the theory that community-based nonprofits act as civic intermediaries for low-probability voters, increasing turnout in communities and among demographic groups where turnout is consistently lower than average (LeRoux and Krawczyk 2014; Nonprofit Vote 2011).

While there exists some evidence about the importance and effectiveness of nonprofits’ voter engagement efforts, there remains a need to better understand which types of messages help compel people to the polls, especially given that lower-information voters espouse low levels of political efficacy and are more likely to hold perceptions that their vote does not matter (Brady, Verba, and Schlozman 1995). One of the challenges for community-based nonprofits in undertaking voter engagement activities is that their messages are required by federal law to be nonpartisan. There is mixed evidence in the academic literature regarding whether nonpartisan messages – and which ones – are most effective in motivating voters to the polls. Moreover, there is a distinct gap in the literature regarding the effectiveness of new methods of reaching voters, such as peer-to-peer text messages. This study aims to build on what we know about nonprofits’ roles in elections through a field experiment testing the effects of nonprofits’ GOTV messaging and methods of message delivery.

In this study, we examine the following questions: When nonprofits encourage their clients to vote, does it increase turnout? When nonprofits mobilize their clients for electoral participation, which type of nonpartisan message might be most effective for increasing turnout? If the nonprofit is a known entity by the prospective voter, does the method of message delivery matter? In answering these questions, our study contributes not only to the growing body of research on nonprofits’ political roles, but our results also offer some practical insights for nonprofit leaders planning nonpartisan GOTV efforts, and for foundations and other entities that invest in this work.

In examining these questions, we draw upon the logic of the noticeable reminder theory (Dale and Strauss 2009) as well as evidence from the political science literature on nonpartisan voter mobilization campaigns, to formulate and test a series of hypotheses relating to both the substance of nonpartisan GOTV messages imparted by a nonprofit known to the prospective voter, as well as the message’s method of delivery. To test these hypotheses, we designed a nonpartisan GOTV field experiment carried out in partnership with two large nonprofit organizations in the Chicago region during the fall 2020 U.S. election. We test the effects of two different GOTV messages, one emphasizing the prospective voter’s political efficacy and another emphasizing an important policy issue at stake in the election (immigration). We further test whether delivering these messages by postcard or text message makes any difference. We find that compared to those nonprofit clients who received no GOTV messaging whatsoever, those who received either type of GOTV messages had an increased rate of turnout. While we find no statistically significant overall effect of either message type or mode of delivery on voter turnout, there is a small but statistically significant crossover effect between message and method. This means that the effect of message type on voter turnout is opposite, depending on which messaging method is used. In this case, text messages were most effective at boosting turnout when paired with an efficacy message, whereas postcards were most effective when paired with a policy message.

The remainder of the paper proceeds as follows. In the next section we briefly highlight the theory and literature on nonprofit roles in elections, followed by literature on nonpartisan GOTV messaging, and literature related to methods of GOTV message delivery. We next describe the details of our field experiment, carried out in the fall of 2020 in the weeks leading up to the election. We then present the results of analysis and conclude with a discussion of the implications of our findings for both research and practice.

2 The Role of Nonprofits in Mobilizing Low-Propensity Voters

Rates of turnout are not equal across all demographics in the U.S. Decades of census data predict that “likely” voters are more often white, wealthier, college educated homeowners; characteristics that differ significantly from “likely” voters in other parts of the world (Keefer and Khemani 2004). While income and education are long-standing factors linked to turnout, recent work has shown that racial minorities tend vote at lower rates, even after controlling for income, education, and other factors. Fraga (2018) finds a persistent turnout gap for racial minorities, which he attributes to a low level of self-efficacy among voters of color and a sense they cannot shape election outcomes, compounded by the failure of candidates to effectively target the minority electorate in campaigning. In his research, Fraga (2018) revealed that African American, Latino, and Asian American voters have lower turnout rates than white voters, and this gap has widened over time. Given the persistence of demographic trends, García Bedolla and Michelson (2012) assert that voter identity in the U.S. has long been gendered, classed, and racialized. As a result, there remains a large swath of U.S. citizens that have seldom been encouraged, socialized, or contacted to vote, which is highly problematic given that small changes in the distribution of the popular vote can have dramatic impacts on national politics (Cintrin et al. 2003). Through their engagement with marginalized communities, nonprofit organizations have the potential to rewrite this narrative by helping individuals that have largely been ignored by the American political system, to develop new identities as voters, rather than abstainers.

Research indicates that low-propensity voters are less inclined to trust government (Rainie, Keeter, and Perrin 2019), and much of election messaging tends to come from sources one might associate with government including politicians and political parties. Community-based nonprofits on the other hand, rarely issue election messages. Low-propensity voters may be especially receptive to election messages that encourage participation when the message comes from a known organization such as local nonprofit with whom they have an established connection. In one field experiment examining the impact of voter mobilization on the turnout rates of over 500 clients of nonprofit human service agencies, LeRoux and Krawczyk (2014) found that those who received a voter mobilization contact from agency staff voted at higher rates than clients who received no voter messaging, and further found the probability of voting increased by roughly 9.3 percentage points with each voting-related contact/message received by those in the treatment group. Similarly, in a series of field experiments, Michelson et al. (2008) report that voter mobilization efforts by more than 10 different nonprofits, significantly increased turnout among individuals in low-income, ethnically and racially diverse communities.

In addition to studies looking specifically at nonprofit messaging agents, there is a robust set of findings from the field of political science demonstrating that social networks and interpersonal influences (that is exemplified by a nonprofit-client relationship) consistently increase the likelihood of voting. For example, recent studies examining peer-to-peer messaging find that mobile apps can increase voter turnout by almost 10 percentage points, when messages are exchanged by trusted individuals or friends (Schein et al. 2020). Similarly, neighbor-to-neighbor mobilization has also been linked strong turnout at the polls (Middleton and Green 2008). Michelson et al. (2008) reported on a small experiment undertaken by a community organization (PICO) in California aimed at mobilizing members of PICO-affiliated churches. The results showed that voter turnout was higher when individuals in the community were contacted by friends or fellow (unknown) congregants than when they were contacted by strangers with no ties to the organization.

Social networks and interpersonal GOTV communication can also have positive spillover effects. In other words, GOTV efforts can boost turnout among voters that share a social tie or relationship by someone who received a voting contact, even if they themselves did not receive the initial contact. For example, Nickerson (2008) found that 60% of the effect of door-to-door GOTV messages delivered from messengers associated with the fictitious “Center for Environmental Citizenship” spread to those living in the same household. Similarly, Sinclair, McConnell, and Green (2012) found social pressure messages delivered from a fictitious consulting group, to be significantly associated with voter spillover effects within households. In another door-to-door campaign carried out by the former community organizing nonprofit, ACORN, findings indicated that voter turnout increased by 12 points for those in one person households and by nearly 18 percentage points for those in two-voter households (Michelson 2006). A study of Michigan residents mobilized to vote by human service organizations arrived at the same conclusion. This study found that turnout not only increased among those who received a voting contact from the service agency, but also that the likelihood of encouraging family and friends to vote increased by 4.5 percentage points for each voting-related contact the client received from their service agency (Nonprofit Vote 2011).

Taken together, the evidence presented here suggests that both individual cognitive and sociocultural processes interact and impact voter behavior. More specifically, there is evidence to suggest that voting messages delivered by personal contacts are more effective than messages delivered by people outside a voter’s social network. Based on this evidence, we hypothesize that among those receiving services from a nonprofit organization, voter turnout will be higher for those who receive a nonpartisan GOTV message, than those who do not receive a GOTV message from their service agency.

3 Methodology

Field experiments are considered the gold standard when it comes to measuring the effects voter mobilization techniques (see Dale and Strauss 2009; Gerber and Green 2000a; Michelson 2006; Nickerson 2007 as examples). The reason for this is that this method relies on the random assignment of subjects to either a control group or one or more treatment groups. Random assignment to control and treatment groups ensures that subjects do not vary in some systematic way, which allows researchers to isolate the treatment (GOTV messages in this case) as the cause of variation in the subjects’ behavior (voting in this case). In other words, experiments with random assignment and use of control groups are superior methods for ensuring external validity, because they are the strongest form of design ruling out rival explanations in the outcome variable. This field experiment was designed to test the effectiveness of nonprofit voter mobilization contact, as well as the message content and mode of content delivery on voter turnout in the 2020 U.S. election. The first phase of this experiment involved recruiting and selecting organizational participants for this study, providing them with training and data collection materials, and having them carry out the first level of intervention. The project was carried out in partnership with two large nonprofit organizations, one in Chicago and one in the suburbs. Both provide a mix of services ranging from neighborhood outreach and community planning to education and job placement and serve low-income, ethnically diverse populations. The organizations were recruited into this project through a call for proposals that was distributed widely via the researchers’ professional networks and through social media. The project was framed as an opportunity to receive a grant for carrying out a civic engagement initiative. The research team received nine applications. The two organizations were selected on the basis of their organizational and staffing capacity, size of client population, and willingness to carry out the elements of the research design and data collection.

The call for proposals specified that selected applicants would be required to make a voter engagement contact with at least 200 eligible voters served by their organization and keep a spreadsheet with the name, address, phone number, and date of birth for each person with whom they made such a contact. A voter engagement contact was defined as asking the client if they are registered to vote; if not, helping them to register online, and if so, helping them check their registration status online and asking the client to commit to a ‘pledge-to-vote’. In this experiment 58 of the 397 voter engagement contacts resulted in a nonprofit helping their client to register to vote online, the remaining 339 resulted in a voting pledge. If an organization contacted an individual not eligible to vote, that data were not reported to us. Additionally, participant agencies were required to select at random, 50 people who receive services from their organization in the past year who did not receive any contact from the organization about voting, and supply the research team with the name, address, cell phone number, and date of birth for those clients as well. In order to be selected, organizations also had to consent to allow our research team to send a nonpartisan voting reminder to those people in their mobilization list, using the agency’s name and logo the week before the election. Each organization received a $1000 ‘civic engagement grant’ for their participation. Our research team provided training and data collection protocols for each organization, and had weekly check-ins with the point person at the organization designated to lead the project. Our project kicked off in early September 2020 and participants were given a deadline of 30 days before the election (the deadline to register for the election in Illinois) to make their mobilization contacts.

By the deadline, our participants had made mobilization contacts with 397 clients, very close to the projected goal of 200 per organization, and provided 163 names of clients with whom no voting contact was made. Of those mobilized by the nonprofits, approximately 43% identified as male and 56% identified as female. Asian, African American, and Hispanic or Latino individuals comprised approximately 87% of those mobilized. Over half (53%) of those mobilized by the nonprofits did not vote in the previous 2016 election. The characteristics of the control group, which was comprised of individuals served by the nonprofits but that did not receive any voting related contact, resembles those who were mobilized relatively well. Full sample statistics for both groups can be found in Appendix A.

The second phase involved sending a voting reminder message the week before the election to all of the 397 clients our nonprofit partners mobilized. Clients were randomly assigned to receive one of two messages via text message or postcard. Sample postcards can be found in Appendix B. Within each message delivery group, clients were again randomly assigned to receive one of two possible messages.^[6] Message one focused on the political efficacy of the voter. This message was chosen because historically low-propensity voters have been underrepresented in the polity and thus form common conceptions of what voters “look like”. According to García Bedolla and Michelson (2012) for low-propensity voters a “change in the voters’ feelings of self-worth and therefore political efficacy [is] necessary to move them toward the belief that they can and should act politically,” (p. 370). In other words, these populations need to first understand that their voice matters, in order to “see” themselves as voters and take action. Given the populations served by the nonprofit partners in this study, a message geared toward reinforcing the message that their vote matters, was chosen over other well-tested GOTV messages sent to the general population.

Message two focused on a politically salient national policy issue, emphasizing that immigration policy was at stake in the election. This choice of this policy issue was intentional. Immigration was identified by Pew (2020) to be a top 10 issue of importance to Americans in making their decision on who to vote for in the national election. Further, given the increasing attention to and action on the issue by the Trump administration in the years leading up to the election, the issue was of broad interest to individuals across the ideological spectrum, thus making even a nonpartisan reference to it, potentially appealing to a diverse array of individuals (Box 2017; Goodsell 2019):

Message 1: “Dear (subject name), All of us at [Nonprofit Name] would like to remind you to vote on Nov. 3^rd. Every vote counts and YOURS could decide the election!”

Message 2: “Dear (subject name), All of us at [Nonprofit Name] would like to remind you to vote on Nov. 3^rd. The next four years will be critical for immigration policy in the US, make your voice heard!”

With the subjects randomly assigned to these message groups, we split each messaging group again by random assignment to test the effects of two different methods of message delivery. Figure 1 shows the distribution of cases between control and treatment groups.

Figure 1:

Between-subjects factors.

This portion of our study represents a between-subjects two-by-two factorial design, with random assignment of clients to one of four treatment conditions: 1) political efficacy message via text, 2) political efficacy message via postcard, 3) policy message (immigration) via text, 4) policy message (immigration) via postcard. Figure 2 provide an illustration of the four treatment groups.

Figure 2:

Treatment groups.

To test our hypotheses, we matched those people mobilized by our nonprofit project partners with voter turnout data from the Illinois state voter file obtained via the election data vendor Catalist. With this information, the citizens mobilized by our nonprofit field partners were matched by name, address, and date of birth, with the state voter files to determine whether they had voted. This information allows us to conduct two sets of analyses to test our hypotheses, 1) comparing the rates of turnout among those who were mobilized, versus the control group of nonprofit clients who were not provided with any GOTV messaging, and 2) comparing rates of turnout among the four treatment groups. Table 1 provides descriptive statistics for each group. Independent samples difference of means tests, and univariate analysis of variance (ANOVA), were used to determine whether there are statistically significant differences between the treatment and control groups on the outcome variable (voted 2020), and between subjects within the treatment groups.

Bivariate logistic regression was used to determine the odds of voting based on the treatment, all else being equal. Given that voted is a binary measure (1 = voted, 0 = did not vote) the means presented in Table 1 can be interpreted as percentages. We now turn to the results of this analysis.

4 Results

Our first hypothesis was that voter turnout would be higher for those who received a nonpartisan GOTV contact from a nonprofit, than those who do not receive any voter messaging. Independent samples differences of means tests confirm this hypothesis. The mean turnout of those who received no GOTV contact is 0.42 (n = 110), while the mean turnout rate of all treatment groups combined is 0.59 (n = 387). An independent samples difference of means test indicates a statistically significant different of 0.168 (p = 0.002) between these groups, meaning the rate of turnout among those who received a GOTV contact from a nonprofit was 16.8 percentage points higher than those received no voting contact. Lastly, we ran a logistic regression model to determine the odds of voting based on organizational contact, controlling for factors known to impact voting behavior, including city residency, gender, age, race, income, and previous voting history, using data obtained from Catalist (Full descriptive statistics and variable measures can be found in Appendix C). As presented in Table 2, the odds of voting increase by a factor of 2.4 when a person is mobilized by a known nonprofit service provider, all else being equal.

In line with extant theory, prior voting history is the single best predictor of voting in this model, increasing the odds of voting by a factor of 5.6.

Our next analysis examines differences in turnout among the four treatment groups. Our second hypothesis predicted that a nonpartisan message emphasizing a salient national policy issue would be more effective at increasing turnout than a generic political efficacy message, but the results fail to confirm this hypothesis. We report the results of a two-way ANOVA in Table 3, and rely on Cohen’s (1988, p. 283) rules for interpreting η² as being 0.01 = small effect, 0.06 = medium effect, 0.14 = large effect. As the findings in Table 3 indicate, there is no significant overall effect of message type on voter turnout, F(1, 382) = 0.40, p = 0.528. Hypothesis three predicted that messages delivered via text message would be more effective than those delivered via postcard, but again we find no statistically significant overall effect of the method of message delivery on voter turnout, F(1, 382) = 0.016, p = 0.698.

We do however find a small (η² = 0.02) statistically significant crossover effect between message and method F(1, 382) = 6.26, p = 0.013, revealing that the effect of message type on voting is opposite, depending on which method is used to deliver it. Figure 3 illustrates the crossover effect using the estimated marginal means. More specifically, this figure shows that text messages were most effective at bolstering turnout when paired with an efficacy message (67%), whereas postcards were most effective when paired with a policy message (62%).

Figure 3:

Estimated marginal means of voter turnout by message and method.

Also of note in the ANOVA model is the control variable representing whether the individual contacted by the nonprofit was provided help with registering to vote, as opposed to making a pledge to vote. When making the initial voter contact, our nonprofit partners helped each person check their registration status and if they were already registered, the client was asked to make a voting pledge; if not, the nonprofit helped them register online. In our model, the effect of this variable on voter turnout is larger than the message/method interaction F(1, 382) = 22.16, p = 0.00, (η² = 0.06). The largest effect in the model (η² = 0.13) is associated with prior voting history which comports with previous studies showing that voting behavior is habit-forming.

While we predicted that messages emphasizing an at-stake salient national policy issue might compel people to the polls, this prediction does not bear out. In fact, our data show higher rates of turnout among those who received the political efficacy message and suggests that voters may be responding to something entirely unrelated to policies (or the candidates’ positions on issues) but rather may be responding to messaging that reinforces positive self-identity of the voter. Indeed, others have suggested this (Bryan et al. 2011; Jenke and Huettel 2016) and future research should endeavor more tests of this theory, as it has extraordinary implications for rational choice theories of voting and for the way that candidates and parties currently campaign.

5 Conclusions

Future studies should aim to replicate these findings on a larger scale. However, we must acknowledge that carrying out a field experiment in the natural setting of a community-based nonprofit presents challenges. While we encourage more studies using these designs, we share some observations in hopes of helping others who are considering partnering with nonprofit organization to carry out field experiments. First, many nonprofits lack the staffing resources and capacity to carry out elements of an experiment with fidelity, and to follow through with the necessary data collection, even with the added incentive of a small grant. Researchers must take these capacity features into consideration and ensure that partner organizations have a full understanding of what is expected of them. We suggest putting this in writing in the form of a partner agreement, or MOU as we did in this case. Moreover, many nonprofits may be unwilling to randomly select a group of clients from whom to withhold treatment or the intervention (in this case a voter mobilization contact) for the sake of an experiment. Many nonprofits that applied to be part of our study preferred to mobilize as many clients as possible and did not see the value in excluding a set of clients from their messaging efforts in order to preserve a control group. Researchers can help nonprofits understand the value of this by explaining how the research design can generate new evidence about what works, and once this is known the elements deemed effective can be applied to all clients. While we emphasized the importance of random selection to our project partners and discussed with them their method of selection for control group cases, we are unable to verify for ourselves whether those cases are truly random. Privacy laws and client confidentiality considerations precluded us from having full access to the nonprofits’ client lists to randomize the cases ourselves. This remains a limitation of our study, and field experiments in general, which is why some have suggested the use of A/B testing as an alternative strategy (Jilke et al. 2019).

Researchers considering partnering with nonprofits to carry out experiments should also know that unexpected contingencies may also arise. One of these in our case was the fact that COVID case rates were surging in the community at the time the experiment was being carried out, limiting the face-to-face contact nonprofits were having with their clients altogether and placing increased demands and pressures on these organizations as they adapted to rapidly changing conditions and pivoted to new forms of service delivery. For example, many nonprofits who had done voter engagement work in the past would host a big event such as a picnic or community rally for their clients and members, often to coincide with National Voter Registration Day. State and local regulations limiting the size of in-person events meant an additional burden for nonprofits to have to reach clients by phone or Zoom for these contacts.

Despite these challenges, our process of careful selection of nonprofit project partners and close work with site leaders to oversee the implementation of the interventions and data collection enabled us to execute a successful experiment. Our findings add to the accumulation of experimental evidence that nonprofits act as civic intermediaries during elections, helping to increase turnout among politically underrepresented populations and groups. By engaging in nonpartisan GOTV efforts, nonprofits can act as civic intermediaries, helping to promote civic literacy and bridge the gap between disaffected citizens and government by increasing political participation in the form of voting. The findings here underscore the potential of nonprofits in expanding electoral participation and making democracy more inclusive. In terms of practical implications, our findings suggest the message itself is less important than the act of issuing a message. Any messaging that encourages voting is better than none (as long as it is nonpartisan). For nonprofits deciding whether or not to get involved in elections, they can be assured that any level of messaging effort will yield an increase in turnout, and any organizational resources expended will yield a civic return on investment.

More specifically, our findings suggest that nonprofits might yield the most benefits when focusing their efforts on mobilizing new or first-time voters and registering them to vote. Prior voting history and whether or not the known nonprofit helped register a client to vote are the strongest predictors of voting in this study. Studies have suggested that registration is the biggest hurdle to electoral engagement (Powell 1986; Squire, Wolfinger, and Glass 1987; Wolfinger and Rosenstone 1980). However, overcoming this hurdle can alter an individual’s self-identity, helping them to think of themselves as participators rather than abstainers, engendering future voting behavior (Gerber, Green, and Shachar 2003). The importance of nonprofits in the voter registration process will vary from state to state based on whether or not voter access has been restricted or expanded by state laws. For example, nonprofits may play a less important overall role in turnout in states where voter registration is automatic, or in states where same day registration laws exist. However, given the important roles nonprofits may play in shaping the identity of their clients as civic participants, their efforts remain important for representative democracy. Voting is habit forming and nonprofits may play a critical role in establishing the initial thought process that enables this behavior. Future studies can examine this process to better understand the role that nonprofit play in helping to shape the civic identities of the clients.

In finding no main effects of message type, our results contribute to the literature on impact of message substance, aligning with previous experimental evidence showing that message content is not statistically linked to turnout rates. That said, the evidence remains mixed on whether the message matters, and newer evidence summarized by Searle and Abrajano (2019) shows that the substance of GOTV messages do indeed matter. Our study provides limited support for the noticeable reminder theory, as messages received by text only appear to compel voters to the polls when paired with a specific type of message, one where the nonprofit emphasizes the importance of the individual voter. The pairing of this message and method might add to the growing body of work emphasizing “relational organizing” which suggests that when messengers reach out to their known networks and let them know their voice matters (political efficacy) using tools which can cut through daily noise (texting), voting turnout increases (Green and McClellan 2020). Additional studies and further analyses are needed to test relational organizing models and noticeable reminder theory, as it is possible that the effectiveness of texting may differ among particular demographics or may fade over time as more players enter into and saturate this arena. The findings with regard to our crossover interaction merit further investigation on their own. Our results point to an important line of inquiry for future GOTV messaging experiments, which is to test the match between method and message to ascertain whether there are ideal combinations for reaching particular demographics.

Future studies may also examine these theories in different nonprofit contexts that are highly influential in U.S. electoral politics. One of these contexts might be among churches and their congregations. According to García Bedolla and Michelson (2012), “appeals from pastors or their representatives can be expected to cue targeted voter’s social identification as parishioners in addition to their cognitive schemas as voters, potentially making those appeals more effective,” (p. 737). Another avenue of importance to consider is the impact of GOTV efforts in the world of 501c4s, where there are highly partisan pushes to influence election outcomes.

There is a growing body of evidence which suggests that ethno-racial voters in the U.S. are not socialized in a way that allows them to develop a positive self-identity as a voter or civic participant. As a result, these citizens vote less and are therefore contacted less by political campaigns (Brown 2016), thus compounding the problem of voter representation over time. Through their voter registration efforts nonprofits and community-based organizations may not only be helping democracy by lessening the participation gap, but also the individual, by prompting a cognitive shift which allows them to see themselves as valuable civic participants. The key to this shift may in part, be due to the authentic relationship between the nonprofit and the individual which permits them to be more receptive to voter mobilization efforts such as registration outreach or noticeable reminders in the first place. While text messages sent from a known service provider highlighting the positive self-identity of the voter may reinforce the developing cognitive schema of the individual as a “voter” in a way that’s hard to ignore, and therefore increase turnout, it’s quite possible that the authentic connection between the sender and receiver mediates this relationship. That which is authentic or inauthentic has been shown to impact attitudes and behaviors across diverse domains (Newman 2019), and it’s likely that voter mobilization is no exception.

Our study results, placed in the context of the existing evidence suggest a continued need for testing of ‘what works’ in the way of GOTV messaging, as the American electorate evolves and changes alongside shifting policies, priorities, and technologies. Our study offers some novel findings with regard to the interaction of GOTV message method and substance that can be tested further in future experiments. We especially encourage further experiments carried out in partnership with nonprofits with the goal of detecting the most effective strategies for mobilizing low-income voters and underrepresented groups.