A “Terrific Symbol”: Physical Personalization of Pandemic Relief Enhances Presidential Support

ABSTRACT The COVID-19 pandemic has forced governments worldwide—many that previously prioritized austerity—to approve large relief packages. Political economy tells us that politicians will try to profit from this electorally, but much remains unknown about precisely how pandemic relief might influence voting intentions. Then-President Donald Trump foregrounded this question early in the pandemic by becoming the first US president to physically place his name on Internal Revenue Service relief checks mailed to citizens. By leveraging a nationally representative survey whose timing achieved quasi-experimental variation in the receipt of payments both with and without Trump’s name physically on them, this study asks: Can a president successfully win support through physical personalization of the payments? Yes, the study finds. Receiving a physically personalized check in the mail is associated with a much greater self-reported likelihood of voting for the president, with gains mainly from partisan outgroups. No clear effect is found for unpersonalized electronic transfers. These findings withstand multiple robustness checks.

T he coronavirus pandemic has forced governments across the globe-including many that had previously prioritized fiscal austerity-to adopt massive economic relief packages, often including direct payments to citizens. Understanding the politics of such payments is important. Not only can they help citizens make ends meet in difficult times and significantly impact how the economy responds to the crisis; they also potentially can reduce citizens' financial need to continue or assume work that might further expose them to the virus, thereby extending the pandemic (Glum 2020). What we know about political economy tells us that politicians are likely to seek ways to profit politically from the crisis and to take those actions that benefit them most (Campbell 2012). Pandemic relief is not ordinary policy making, however, and the rarity and typically sudden onset of a high-profile pandemic leaves us without a solid research basis for understanding the political incentives surrounding it.
By addressing these larger questions of political economy and public health, this study asks: Can a president appropriate public-opinion dividends by physically personalizing federal aid to citizens who are suffering the economic consequences of a pandemic? Putting a finer point on it, we already have strong reason to believe that timely relief payments can benefit presidents (Kriner and Reeves 2015, 82-100;Liu and Kirwan 2020;Reeves 2011), but does the way in which they are delivered matter? Do relief payments provided in one way enable more effective presidential credit claiming than the same relief payments delivered in a different way? Then-President Trump put this question at the center of American politics when it was announced in April 2020 that his name would appear on hardcopy federal "stimulus" checks being mailed to citizens to help them cope with the material fallout of the COVID-19 health crisis. This was the first time a president physically attached his name to checks paid out by the Internal Revenue Service (IRS) (Rein 2020). Many ascribed this move to the President's reelection effort. His own treasury secretary called it "a terrific symbol to the American public" (Mansoor 2020). Some opponents challenged it as a criminal abuse of federal resources for political purposes, and Senate Minority Leader Chuck Schumer initiated legislation that would explicitly bar such moves in the future (Davidson 2020).
There is reason to think that the physical personalization of relief payments might have electoral consequences. It is something of an axiom that "as go voters' economic fortunes, so go incumbents' reelection prospects" (Erikson, MacKuen, and Stimson 2002). We also know from different political contexts that doling out cash can lead recipients to return the favor at the ballot box even without enforcement, reinforcing a sense that the giver is powerful and/or caring (Hicken and Nathan 2020;Kramon 2017). Although this effect has yet to be studied systematically for pandemic-related aid, it has been shown to hold for disasterrelief payments of different types (Kriner and Reeves 2015, 82-100;Liu and Kirwan 2020;Reeves 2011). Politicians are thus attentive to ensuring that they get credit for major "pork barrel" and other benefits they supply (Balla et al. 2002;Bickers and Stein 1996;Bueno 2018;Cain, Ferejohn, and Fiorina 1990).
Getting credit is far from automatic (Mayhew 2004, 60;Stein and Bickers 1994), leading some scholars to question whether voters are actually directly rewarding politicians at all (Klingensmith 2019;Samuels 2002). In short, voters' incomplete information about who is responsible can create attribution problems that could reduce officials' incentive to supply goods that otherwise would seem to benefit them as well as the public (Keefer and Khemani 2005). One way that these attribution problems can be solved is by broadcasting credit-claiming messages through the media (Grimmer, Messing, and Westwood 2012;Rogowski and Stone 2020). Less attention has been given to the mechanism of interest in this article: physical personalization, or the literal attachment of politicians' names to the very goods they supply to voters. Perhaps the most prominent examples of this practice are the seemingly countless buildings, highways, and other structures named after US Senator Robert C. Byrd in West Virgina (Hagen 2007;Kriner and Reeves 2015, 102). What we do not know, however, is whether physical personalization can move the political needle for crisis relief generally (i.e., "economic stimulus payments") and for pandemic aid more specifically. It also remains unknown whether physical personalization can effectively claim for a president electoral benefits from congressionally approved transfers supported by both major parties that other methods do not obtain.
Speaking to these larger theoretical issues of political economy, this study addresses whether Trump's physical personalization of direct coronavirus economic stimulus payments to citizens in Spring 2020 rendered recipients more likely to express an intention to vote for him. The study used a quasi-experimental design in an original nationally representative survey of US adults in early May 2020, after many stimulus checks had been mailed but not all had been received. Several elements of the payment process and research design facilitate causal identification. These elements include the crucial fact that some people received stimulus payments by physically personalized check and others received them through nonpersonalized direct deposit, as well as the near-randomness involved in the timing of the receipt of the checks during the period chosen for the survey. With numerous robustness checks, the study reveals that people who received a pandemic relief check with Trump's name physically on it tended to be more willing to state that they would vote for him than were other people, including those who received the same sum by direct deposit. The presidential gains came primarily from partisan outgroups, including people who identified as independents and some Democrats. Physical personalization thus is established as an important mechanism for politician credit claiming-a finding with potential implications extending beyond pandemic-relief aid to broader questions of the role of symbols in patronage politics and clientelism.

RESEARCH DESIGN
Shortly after bipartisan majorities approved it in both houses of Congress, President Trump signed the Coronavirus Aid, Relief, and Economic Security (CARES) Act into law on March 27, 2020. It provided for financial transfers to individuals earning up to $99,000, to heads of household earning up to $136,500, and to couples earning up to $198,000 per year. Eligible people already registered to receive IRS payments electronically or who had registered by May 13 would receive their payment transferred directly to their bank account. Other eligible citizens would receive a check in their home mailbox. In mid-April 2020, the White House revealed that the president's name would be printed on the lower left side of the checks (Davidson 2020;Mansoor 2020). People who received electronic payments would not see Trump's name. The IRS began electronic transfers on April 11 and started mailing checks on April 20, expecting to continue through the summer (Internal Revenue Service 2020; Vann 2020).
After many but far from all citizens had received their checks, a nationally representative sample of 2,000 US residents was surveyed online from May 4 to May 11, 2020, by YouGov America for this study (Hale 2021). 1 Survey participants were asked if they had received either a stimulus payment by direct deposit, a check in the mail, or no payment. Of all respondents, 51% had received stimulus payments electronically and 9% by mail, with the remainder (40%) reporting no payment for a wide range of reasons, including not checking their bank account, delivery delays, and ineligibility. 2 This variation reflects the nearrandom assignment of check receipt within a subset of the population; the sources of nonrandomness are systematically addressed in the discussion of causal identification. From these data, binary variables were created measuring whether respondents (1) received a stimulus payment electronically; or (2) received a check by mail. The survey also asked respondents to rate their likelihood of voting for Trump, which yielded the main dependent variable: a six-point scale from "impossible" (the lowest value) to "certain" (the highest). 3 Using data that YouGov America obtained from respondents prior to their selection for this study (which eliminates endogeneity concerns), binary control variables were constructed for race, Republican In mid-April 2020, the White House revealed that the president's name would be printed on the lower left side of the checks. People who received electronic payments would not see Trump's name. and Democratic partisanship, higher education, and female gender, as well as scales for age and income categories. 4 The dependent variable (i.e., subjective likelihood of voting for Trump) was regressed on the key independent variable (i.e., receiving a stimulus check in the mail with Trump's signature on it) and on these controls, including receiving the same stimulus payment by direct deposit. To maximally isolate any effect of check receipt from any effect of income-based eligibility, the analysis was conducted only on those who reported family income below $100,000 for nonmarried and $200,000 for married respondents. For transparency, the simplest possible results are reported first (i.e., from an OLS model with no weighting or adjusted standard errors). The robustness of the main findings to modeling choices was assessed by adding controls, estimating an ordinal logistic model, 5 applying income fixed effects, 6 and weighting calculations to the US national population as a whole using propensity scores supplied by YouGov America. 7 Table 1 presents the main results. As shown in the first column, a simple bivariate model found that people receiving a stimulus check in the mail reported being 0.44 points more likely to vote for Trump on the six-point scale. This represents more than a 20% increase from the average propensity to vote for him (i.e., 3.12 on the scale from 1 to 6), clearly a substantial difference. This general finding holds up to multiple model specifications.

RESULTS
Who were the people being swayed by receipt of a physically personalized check? Table 2 reports the results among individuals with different priors. Model 1 shows that physically personalized check receipt had no significant effect among Republicans but a large and highly significant effect (i.e., 0.7 on the six-point scale) among non-Republicans (model 2). Delving deeper, the study found evidence of Democrats being swayed (model 3) along with a much larger and more statistically pronounced effect among people who identified as neither Republican nor Democrat (model 4

CAUSAL IDENTIFICATION
Several concerns must be addressed before concluding that this correlation reflects cause. Perhaps most obvious, the model 1 result in table 1 could be capturing the effect of simply receiving money, whether personalized or not. If so, we would expect the receipt of an equivalent payment by another method to have the same effect.
However, model 4 in table 1 (i.e., simple OLS with full controls) shows that the effect was specific to receiving a mailed check.
There was no consistently statistically significant relationship between receiving a direct deposit (without Trump's name visibly linked to it) and the likelihood of voting for him, and the coefficient is much smaller. This calls attention to whom the check recipients are compared in the analysis. For models 1-3, the comparison group is everyone who was eligible but did not report receiving a mailed check-that is, the 55% of eligible people who reported receiving a direct deposit and the 36% of those eligible who did not report receiving any payment at the time of the survey. For models 4-8, because a variable was included for direct-deposit recipients, the comparison group was effectively the 36% of eligible people who did not report receiving any payment. An additional way to test whether the method of payment delivery mattered is to examine only those people who received payments: Were those receiving physically personalized checks more supportive of Trump than those who received a direct deposit? Table 3 shows that signed-check recipients systematically stood out for more Trump support with at least 90% confidence, with significance levels of more than 99% yielded by the most restrictive models (i.e., those with full controls and income fixed effects). Moreover, table SM7 in Section 3.5 of the online supplementary materials shows that across multiple categories of non-Republicans and non-Trump supporters, including Democrats, the effect was large and consistently statistically significant.
Another concern is whether an omitted variable may have produced the observed correlation by influencing both Trump Notes: Priors include Republican "R," not Republican "NoR," Democrat "D," not Republican or Democrat "NoR-D," Trump voter "Tvot," not Trump voter "NoTvot," Clinton voter "Cvot," and not Clinton or Trump voter "NoC-T." Data collected May 4-11, 2020. Results were estimated using OLS model, unweighted. A full set of controls is included but not reported here. Standard errors are in parentheses. *p ≤ 0.05, **p ≤ 0.01.
Overall, Trump was gaining primarily new supporters (i.e., non-Republicans and non-Trump voters) by stamping his name on the checks.  support and variation in mailed-check receipt. Fortunately, the factors determining (1) whether a citizen was eligible to receive any payment, (2) whether a citizen received a check instead of an electronic transfer, and (3) the timing of the mailing of checks are generally known, which helps us to isolate the physically personalized, mailed-check effect itself.
To address factors that may be associated with (1), the analysis was restricted to aid-eligible populations, as described previously. As for possible confounders linked to (2), eligible people were to receive a direct deposit when they either (a) had previously given their bank information to the IRS for tax refunds, or (b) used the portal that the IRS set up to obtain this information for the CARES payments. Variation in (a) hinges largely on income and minority status. Clearly, people who did not earn enough to receive a tax refund would have no need to set up direct deposit (WABC-TV 2020). However, and more fundamentally, poor and minority communities are systematically less likely to have access to electronic financial services (Birkenmaier and Tyuse 2005). These concerns were addressed by controlling for income and being nonwhite (see models 4-8 in table 1). In addition, the main models were estimated with income fixed effects (see models 7 and 8 in table 1), a method that effectively analyzes variation only within income brackets. 9 We can expect variation in (b) to have been linked to digital competency, comfort, and access, which also are linked to income and minority status and therefore were addressed in the ways discussed previously. Moreover, in the context of these inequalities, this study's online sample provides an important methodological advantage, effectively ensuring that all respondents were reasonably digitally competent. Among those people, variation in comfort level with digital technology was likely related to age, minority status, and education level, all of which were controlled for in models 4-8 of table 1.
Could the timing of the IRS in mailing out checks somehow have been driven by a factor correlated with support for Trump? Fortunately, we know the IRS's mailing policy, which was to send checks earliest to people in the lowest income brackets and latest to those in the highest. Moreover, the IRS made known its mailing schedule: the week ending April 24, 2020, for those with an income of less than $10,000, the week ending May 1 for those with an income of $10,001-$20,000, the week ending May 8 for $20,001-$30,000, and the week ending May 15 for $30,001-$40,000 (Friedman 2020). Thus, models 4-8 in table 1 controlled not only for income but also for people with a family income of less than $40,000 (a binary variable)-people who should have been mailed their check by the end of the last week covered by the survey. 10 Beyond this, variation in the actual receipt of mailed checks should have arisen mainly from the length of time the US Postal Service took to deliver them; therefore, the study also controlled for rural residence, where mail delivery can take longer. Analysis presented in Section 3.1 of the online supplementary materials further shows that the check effect did not accrue to Republicans generally, ruling out omitted variables that may be correlated with both check receipt and Republican support. On these grounds, the distribution of the "treatment" (i.e., receipt of a mailed check with the president's name on it) can be cautiously treated as if random.
As shown in table 1, even explicitly modeling for the factors known to have driven both selection into the category of designated check recipients and the timing of the checks' mailing and receipt, we continue to find a large, statistically significant correlation between electoral support for Trump and receiving mailed, presidentially personalized stimulus checks. In fact, under the strictest conditions, with income fixed effects and a full set of controls, the analysis yielded even stronger findings: effects of more than 0.5 on the six-point scale, with higher levels of statistical significance.

CONCLUSION
Overall, this study reveals that receiving COVID-19 relief checks from the federal government in the mail with Trump's name on them led people to be substantially more likely to express intent to vote for his reelection, reflecting gains primarily among non-Republicans (including Democrats). Similar effects generally were not detected-or were found to be much smaller and less statistically significant-for the same payments when they were not emblazoned with a presidential signature. This indicates that incumbents can appropriate federal-aid efforts for their own political gain in a democracy through actions that leverage their status as a national symbol. Moreover, findings show that this is possible even during a major pandemic health crisis, including in highly politically polarized situations in which the incumbent president is under intense fire from the opposition party and major media.
This study naturally has limitations. It cannot establish how long the physically personalized check effect lasted, including whether the effect identified in this article later influenced actual voting decisions. These are questions to which the current findings would direct future research. The most confident claims also pertain only to lower-income Americans because only they had received checks as part of the CARES package at the time of this study. Of course, lower-income citizens comprise a substantively large and important population to understand, but future research should explore whether similar effects also can be detected among the higher-income population. It also has not been established whether check personalization would have the same effect if it were applied to other forms of disaster relief or transfers generally; however, it is unclear why people would be more attentive to check signatures during a pandemic than in other times of need.
Despite these limitations, the findings contribute to theories of credit claiming and the personal vote by expanding our understanding of the mechanisms through which elected officials might profit electorally through the distribution of government resources. The findings highlight the need for new research into not only narrative messaging (Grimmer, Messing, and Westwood 2012;Rogowski and Stone 2020) but also the types of symbols that elected officials may have available to link themselves to benefits in voters' minds. This avenue for research will yield further insight into clientelistic practices and political economy more generally.
… incumbents can appropriate federal-aid efforts for their own political gain in a democracy through actions that leverage their status as a national symbol. 5. Ordinal logistic models are often recommended for ordered categorical dependent variables like the one in this study because they do not assume a continuous underlying scale. 6. For reasons that are discussed in the causal identification section.
7. For readers interested in tighter application to known population parameters. 8. Respondents too young to have voted in 2016 were excluded from the analysis relating to that election.
9. The first 10 brackets categorize respondents by family income in $10,000 increments; the remaining brackets were $100,000-$119,999; $120,000-$149,999; and $150,000-$199,999. Finer increments were not available. 10. Only family income brackets were available; therefore, some people in higher family brackets may have been mailed checks by this time if they filed taxes individually and their own income was less than $40,000.