Polit Behav (2014) 36:359-376 DOI 10.1007/sl 1109-013-9230-8
ORIGINAL PAPER
Campaign Targets and Messages in Direct Mail Fundraising
Hans J. G. Hassell • J. Quin Monson
Published online: 28 April 2013
© Springer Science+Business Media New York 2013
Abstract Political campaigns raise millions of dollars each election cycle. While past research provides valuable insight into who these donors are and why they are motivated to give, little research takes into account the actions of political campaigns. This paper examines why and how campaigns target habitual donors for political donations. Using the 2004 Campaign Communication Survey, a national survey of registered voters who were asked to collect and send in all campaign mail they received during the last 3 weeks of a campaign, we show that campaigns send donation solicitations predominantly to individuals who have previously donated to a campaign. We also show that campaigns match targeting fundraising appeals to the potential motivations for giving: campaigns target the type of fundraising appeal they use, whether ideological, solidary, or material, to match the socioeconomic and partisan characteristics of the potential donor. The implication of effective targeting is that the "unequal" voice of participation in campaign contributions is not onesided and simply resource based, but rather that campaigns also contribute to the situation with targeted messages to potential donors.
Keywords Campaign fundraising • Political communication • Political campaigns • Campaign donors
H. J. G. Hassell (El)
Department of Politics, Cornell College, 600 First St. SW, Mount Vernon, IA 52314, USA e-mail: hhassell @ cornellcollege.edu
J. Q. Monson
Department of Political Science, Brigham Young University, 745 Kimball Tower, Provo, UT 25545, USA
e-mail: quin.monson@byu.edu
Springer
360
Polit Behav (2014) 36:359-376
Although interest groups and party organizations play a major role in financing a candidate's campaign, individual donors continue to contribute more than half the money raised by all federal candidates (Herrnson 2012) and are a sign of a healthy and viable campaign (Butler 2005; Magleby and Patterson 2008; Malbin 2006; Wilcox 2008). Campaign cash, however, is not freely available to all those who want it. Instead campaigns must front the costs of fundraising, such as printing, postage, consultant fees, and the salaries of personnel to run phone banks and mail rooms (Godwin 1984; Hart 1992; Kanfer 1991), as donors are unlikely to give when campaigns or campaign surrogates do not ask them to do so (Grant and Rudolph 2002; Brady et al. 1999; Rosenstone and Hansen 1993; Cummings and Cummings 2004; Shea and Burton 2006). As such, in order to limit the costs of fundraising and maximize the returns, campaigns target those individuals they believe are most likely to give (Francia et al. 2003; Jones and Miller 1985; Sorauf 1992).
While previous research has examined the question of whom campaigns solicit, these studies have relied upon respondents to report if they were asked to give money to a political party, political group, or individual candidate (Brady et al. 1999; Grant and Rudolph 2002). The problem with such a measure is not only that it is prone to error (Goldstein and Freedman 2002; Price and Zaller 1993), but that it also provides scholars with an incomplete view of campaign targeting which leaves several important questions unaddressed.
The most obvious problem is that a dichotomous variable does not provide information about the number of times campaigns contact potential donors. Previous studies of participation and voting demonstrate that that the probability of political participation increases with additional contacts (Arceneaux and Nickerson 2009; Gerber and Green 2000; Gerber et al. 2003; Green and Shachar 2000). However, unlike with voting, even after an individual has donated to one campaign, that campaign and other campaigns may continue to contact that individual in hopes that he or she will give again. While previous work has theorized that campaigns send a greater volume of requests to individuals with a greater propensity to give, these studies have been unable to explicitly examine that relationship (Grant and Rudolph 2002).
Another glaring omission in the campaign finance literature is the absence of information about the types of requests campaigns send different types of potential donors in their efforts to raise money. The message that a campaign sends an individual has a substantial effect on the likelihood of a resulting donation (Han 2008; Levine and Monson 2011; Levine 2009; Miller and Krosnick 2004; Miller et al. 2000). Donors do not all have the same motivations for donating (Francia et al. 2003; Dawson and Zinser 1976; Wilson 1974) and do not respond to the same types of solicitations (Hart 1992; Kanfer 1991; Shea and Burton 2006). Microtargeting and the use of voter lists has become more important in all aspects of the political campaign to match mail and other persuasive communications with individuals who will be responsive to its message (Gerber et al. 2011; Hillygus and Shields 2008; Lovett and Peress 2010; Monson and Oliphant 2007). Campaigns use donor lists extensively to identify potential contributors and use complicated algorithms to identify non-donors who look like donors but have not yet contributed. Once campaigns have identified these prospective donors, they target them with fundraising appeals designed to match a donor's characteristics (Cho and Gimpel
Springer
Polit Behav (2014) 36:359-376
361
2007; Culnan and Regan 1995; Ginn 2005; Kanfer 1991). Without more than anecdotal information about the content of fundraising appeals we are missing a substantial portion of the full picture of campaign donations.
Those who donate to political campaigns are on the whole wealthier, more educated (Francia et al. 2003; Verba et al. 1995; Rosenstone and Hansen 1993), older than the average American, predominantly male (Francia et al. 2003), and more extreme in their ideological views (Panagopoulos and Bergan 2006). While previous studies have shown that income is the strongest predictor that campaigns will solicit an individual for a financial contribution (Brady et al. 1999; Grant and Rudolph 2002), these studies have not controlled for an individual's contribution history. Because of the positive correlation between an individual's contribution history and income, the effects of income in these studies are inflated due to omitted variable bias. While previous literature has theorized about the impact of past donations on a campaign's decision to solicit individuals it has not yet been able to formally test its impact (Grant and Rudolph 2002).
This paper offers a new and unique dataset of campaign mailings sent to individuals during the last 3 weeks of the 2004 election cycle that corrects previous flaws and addresses previously unanswered questions, allowing for more appropriate tests of theories about whom campaigns' contact, the frequency with which they contact potential donors, and what types of appeals they use when they solicit funds. Although the dynamics of fundraising are different during the last 3 weeks of an election, campaigns are just as active in soliciting funds during this time period as they are in the early stages of the campaign and use the same techniques (Godwin 1988; Hassell 2011). We find that an individual's previous contribution history strongly influences the quantity of solicitations that campaigns send that individual. By including this measure, we find weaker influence of other important factors such as income and education.
We also find that campaigns vary the type of fundraising appeal they use to match the demographic and partisan characteristics of the potential donor. In the last 3 weeks of a campaign political fundraisers target past contributors disproportionately and generally use ideological message in doing so. However when campaigns do not use ideological messages they rely less on a donor's past donation history and more on an individual's income level and political affiliations in deciding the volume and content of mail they will send to that individual.
Donors' Motivations for Giving and Campaign Fundraising Appeals
While an individual's propensity to participate financially in a political campaign depends largely upon his or her financial resources (Verba et al. 1995; Grant and Rudolph 2002; Rosenstone and Hansen 1993), these individuals do not spontaneously decide to donate to a campaign (Olson 1965). Campaigns must persuade potential donors that a contribution to the campaign is a worthy investment of their resources (Grant and Rudolph 2002; Brady et al. 1999; Rosenstone and Hansen 1993; Cummings and Cummings 2004; Shea and Burton 2006).
Springer
362
Polit Behav (2014) 36:359-376
Previous studies have identified three main motivations behind an individual's donation: material motivations, ideological motivations, and solidary motivations (Francia et al. 2003; Wilson 1974). While the traditional view of direct mail is that it is the tool of partisan ideologues (Fritz and Morris 1992; Godwin 1988), more recent accounts suggest that campaigns use a wide variety of different types of appeals to target different types of donors (Sabato 1981; Shea and Burton 2006; Verba et al. 1995) and craft campaign messages with appeals that match the wide variety of motivations for giving (Hassell 2011).
Individuals who donate on the basis of solidary incentives are donors who enjoy being associated with the campaign and the politicians for primarily social or status reasons. Donors who donate for solidary motives often give because they were asked to by a close friend or associate or relish the feeling of being associated with the campaign, and is often associated with wealth and social connections (Francia et al. 2003; Sinclair 2012). As such we believe that campaigns will target solidary fundraising appeals disproportionately at the wealthiest individuals.
Individuals who donate on the basis of material incentive give to a campaign because they wish to influence legislation and access for purposes of material gain or because their donation allows them to receive something of worth in return. In recent years the incentives to join political groups have moved from the benefits of social camaraderie to material incentives (Putnam 2000; Skocpol 2003). As such, we believe campaigns will target material appeals to middle income individuals who may not necessarily have the financial ability to justify a donation on the basis of solidary or ideological motives, but have joined other political groups in the past, perhaps in response to material incentives.
Lastly, ideologically motivated donors are donors who give for purposive reasons. They become part of the campaign because they see a particular candidate as an advocate for a cause they wish to forward. This motivation, also called a consumption motive, comes from the intrinsic benefits that the individual receives for the act of contributing to a cause or an organization which he or she believes is important (Dawson and Zinser 1976). We believe that campaigns will target ideological appeals towards more partisan individuals. In addition, because past studies have suggested that campaigns rely primarily on ideological fundraising appeals at the end of the campaign (Hassell 2011), we believe that campaigns will target past contributors disproportionately with ideological appeals, while relying on solidary and material appeals more when targeting new donors.
Description of the Data
This paper draws upon the 2004 Campaign Communications Survey (CCS) conducted by the Center for the Study of Elections and Democracy (CSED) at Brigham Young University. The CCS is a multi-mode national survey of registered voters who were contacted prior to the 2004 election and asked to keep a log of all of their campaign communications during the last 3 weeks of the presidential election including phone calls and personal contacts. Respondents also collected and sent in the political mail and email they received during that time.
Springer
Polit Behav (2014) 36:359-376
363
The field work was conducted by the Social and Economic Sciences Research Center (SESRC) at Washington State University. A similar survey was conducted on a smaller scale in 2002 (Monson and Curtis 2004, 293-294) and both the 2004 and 2002 surveys rely heavily on the method for mail and mixed mode surveys developed by Dillman (2000).
The sample was obtained from the "DataMart" national database of registered voters maintained by the Democratic National Committee leading up to the 2004 election. Because the survey design required some telephone contact, the population was defined as registered voters living in households with telephones. The sample design was intended to maximize the participation of respondents most likely to receive campaign communications. To this end the sample was stratified by factors that contribute to a higher probability of receiving all types of political mail: past turnout behavior and residence in a state that was likely to have competitive US Senate races. Past turnout behavior is a common variable used by campaign consultants to target campaign mail and other communications (e.g. Malchow 2003) and competitive races generate more campaign activity of all types than non-competitive races (Goldstein and Ridout 2002; Herrnson 2012; Jones et al. 2007; Magleby et al. 2006).
Following the model for mail surveys outlined by Dillman (2000), respondents were contacted seven times from October 4, 2004 through the December 22, 2004 to elicit their cooperation in the project. In the national sample, the response rate for the post-election telephone survey was 50 % while the response rate for the questionnaire and log booklet portion of the survey was 37 %.
After completing the data collection, SERSC entered the data from the questionnaire and log booklet as well as the initial coding of the mail pieces. The 1,071 respondents that returned the questionnaire/log booklet along with their political mail sent a total of 19,297 pieces of political mail (including local races which we do not include in our analysis). Coders sorted the mail and identified 2,466 unique mail pieces from federal races received by the study participants during the 3-week field period many of which were sent to multiple recipients.
Within the 2,466 unique mail pieces we identified 293 pieces of mail sent to 162 individuals that asked the recipient for a monetary donation. Because some of these pieces were sent to more than one individual, the group of 162 individuals received 494 total pieces of mail containing fundraising appeals. We then linked the 1,071 survey respondents to the mail that they received allowing us to analyze the factors that contribute to the volume of fundraising mail and type of requests that each individual receives.
We acknowledge that certain selection effects could be at work in our sample and that the people willing to take the time to send in their mail are more likely to be interested and engaged in politics. We cannot observe those people who chose not to participate, nor can we differentiate those participants who did not send in mail because they did not receive any, versus those participants who received mail but did not comply with our instructions. However, without the ability to observe the actions of a campaign directly, this new survey gives us the best indication of campaign actions currently available. Among those participants who did send in mail, we observe people both who contributed in the past and those who did not. There is also no correlation between the total number of pieces of political mail
Springer
364
Polit Behav (2014) 36:359-376
(including advocacy pieces) an individual sent in and the likelihood that an individual had donated in the past.
Types of Campaign Fundraising Messages
Our data allows us to go beyond just showing that campaigns are more likely to contact individuals with certain characteristics. A campaign's ability to secure a donation is not just a matter of contacting the right individual about giving a donation but also includes sending the right message in order to maximize donations at a minimal cost (Hart 1992; Kanfer 1991; Shea and Burton 2006).
To identify the nature of campaign fundraising targeting we first analyzed each individual piece of fundraising mail to identify the types of appeals that each unique piece of mail used to invite the recipient to give money. We coded each mail piece to identify whether it used ideological appeals, solidary appeals, material appeals or a mixture of the three to solicit potential donors.
We coded mail as containing solidary appeals if the mail contained language that appealed to the sense of being a part of a team or implied that a donation linked the donor to a larger group of supporters or famous individuals. Such mail included phrases such as "will you stand beside me as one of my core supporters?" or "join the thousands of others who have already donated to the campaign." We also coded mail as having a solidary appeal if the fundraising message came from a nonpartisan celebrity such as a sports or entertainment star.
We identified mail as having material appeals if it promised quantifiable benefits—such as merchandise, items with a monetary value, or increased material benefits that might come as a result of the candidate winning—in exchange for a donation. This included appeals which would have a direct impact on the recipient's pocketbook such as highlighting a member of Congress's work on issues such as "eliminating the marriage penalty" or efforts to "reduce medical liability" in a message sent to a physician. In addition we also coded mail as having a material appeal if it promised campaign paraphernalia in exchange for donations.
Lastly, we coded appeals as being ideological if they focused on partisan politics. These appeals frequently invoked ideology or used language highlighting a candidate's party as a reason for donating. A fundraising mailer could have all of these types of appeals in it or none of them, or any combination. 98 % of mailers contained at least one appeal, 35 % contained at least two types of appeals, and only 4 % contained all three types. Those appeals that did not contain any of the three different types of appeals were simple messages such as "We need to do everything we can to win!" without mention of party or issues.
Figure 1 shows the frequency of types of appeals that appear in fundraising mail and how these appeals are used together. During the final 3 weeks of the election cycle that the study covered campaigns rely more heavily on ideological appeals in their direct mail fundraising. Over 90 % of all fundraising solicitations contained an ideological appeal, and 57 % contained solely an ideological appeal. However, campaigns do not rely on ideological messages alone. Just fewer than 30 % of the 293 unique pieces of fundraising contained a solidary appeal and 18 % contained a
Springer
Polit Behav (2014) 36:359-376
365
material appeal. Solidary and material appeals are often used in combination with ideological appeals. Campaigns combined 86 % of solidary appeals and 82 % of material appeals with an ideological appeal. Thus, while campaigns use all types of appeals during these last weeks before an election, they rely heavily on ideological appeals during this time period.
Campaign Donor Targets and the Volume of Fundraising Appeals
Turning to the sample we find a significant disparity in the contribution history of solicited and non-solicited individuals. Campaigns sent 494 pieces of mail to 162 individuals (out of the 1,071 person sample) asking them to make a contribution. Of those 162 individuals, 80 (48 %) of them indicated that they had donated to a political organization or campaign in the previous 8 years. Of those individuals who did not receive a donation solicitation, only 19 % (171 of 909) % of them responded that they had previously given a campaign donation. In total, campaigns contacted over 30 % of individuals in the sample who had given previously compared with just fewer than 10 % of non-contributors, a difference significant at the 99.9 % level.
Political campaigns also are more likely to target individuals with higher incomes and more education. Of the 254 individuals in the survey who responded that they made more than $80,000 a year, campaigns sent 17 % of them at least one piece of mail requesting a donation. Of the 305 respondents who made less than $40,000 a year, only 12 % of them received a request for a donation. The differences are similar for those with no college education (9 % contacted) and those with a college degree (18 %
Ideological (92%)
6
(2%)
Fig. 1 Venn diagram of use of appeals in direct mail fundraising
Springer
366
Polit Behav (2014) 36:359-376
contacted). Although smaller than the differences between past contributors and non-contributors, both of these differences are significant at the 95 % levels.
However, our data allows us to explain more than just whom campaigns are more likely to contact, but how often they contact those individuals. Indeed, not only are political campaigns more likely to target past contributors, individuals with higher incomes, and individuals with more education, they send these individuals many more requests for money. Past contributors received just over seven times as many fundraising requests as non-contributors. Individuals who made more than $80,000 a year were sent 3.75 times as many fundraising requests on average than those individuals making less than $40,000 a year. Likewise, individuals with a college education received almost 4.5 times as many campaign requests for money than those individuals who had not attended college. All of these differences are significant at the 99.9 % levels.
This unique data set allows us to model the total number of fundraising appeals and total number of appeals of a certain type that an individual receives as a function of an individual's past contribution history, partisanship, ideology, group membership, income and education. To measure these variables we rely on the survey responses of the individuals. A respondent's ideology and partisanship were measured on a seven-point scale, and those who labeled themselves as very conservative/liberal and strong Republican/Democrat were considered extreme ideologues and partisans respectively. As part of the post-election survey respondents were also asked to list what organizations or groups they belonged to. In the model we include the total number of groups to which the individual belonged under the premise that campaigns will also have similar information that would allow them to better gauge an individual's interests and to target messages more specifically towards individuals about which they had more information (Ginn 2005).
Table 1 contains the results from a zero-inflated negative binomial regression with the number of pieces of mail that an individual received that contained a donation solicitation as the dependent variable. Models 1 and 2 look at all fundraising mail, while models 3, 4, and 5 look at the number of appeals that an individual received of a certain type. Because our dependent variable does not allow us to differentiate between those who received no appeals because they were not a fundraising target and those who were a campaign target but did not receive mail during the time period we use a zero-inflated negative binomial. This lets counts of zero occur in two distinct ways; either as a result of campaigns not targeting an individual at all, or a campaign targeting an individual but choosing not to send an appeal at the time (Cameron and Trivedi 2005).1 From this we can tell not only who
1 Because the dependent variable is a discrete count, a traditional linear regression model (LRM) may produce estimates that are inefficient, inconsistent, and biased and is thus inappropriate (Long 1997, 217). The LRM also does not constrain the number of events to be positive (King 1989, 123). In this case reasonable modeling choices include the Poisson regression model (PRM) or the negative binomial regression model (NBRM). When using the PRM one assumes that the events are uncorrected and that the rate of event occurrence is homogeneous. When this is the case, the conditional mean of the outcome is equal to its variance. In practice these assumptions are rarely met—event counts are correlated with each other or the rate of event occurrence is heterogeneous resulting in overdispersion. Using the PRM with data that are overdispersed results in estimates that are inefficient and standard errors that are biased downward. However, there are two possible sources for the over-dispersion in our data. First an individual
Springer
Polit Behav (2014) 36:359-376
367
campaigns are more likely to target, but also how much more fundraising mail they send to certain types of individuals.
The results in the first and second models of Table 1 from the logit stage of the model confirm our expectations about the importance of an individual's past contribution history on the likelihood of being a fundraising target. The negative value on the model is an indication that an increase in the value of that variable makes it less likely that the individual is a "certain zero" or not a campaign target (Cameron and Trivedi 2005). We also find significant effects of age and education on the likelihood of a campaign targeting an individual.
More surprising is the effect of the income and partisanship variables. While an individual's past contribution history and group membership all increase the likelihood of a campaign targeting an individual, the strength of partisan identification and annual income have no effect on a campaign's targeting decision. This finding is contrasted with a significant and positive relationship between an individual's partisanship and income and the total number of donation solicitations received from political campaigns evident in stage two of the ZINB in models one and two. Thus while income and partisanship affect the number of solicitations received, once we control for an individual's past contribution history, they are not a significant predictor of the likelihood of being a fundraising target.
The model also allows us to see the combined effects of these variables in both stages of the model on the number of fundraising mail campaigns send to certain types of individuals over the 3 week period (Bartus 2005). As indicated in Table 2, holding all other variables at their means, campaigns were likely to send just under 0.8 more pieces of mail over the 3 week period to individuals who had previously
Footnote 1 continued
could receive no mail because that individual is not a campaign target. Alternatively, a large number of individuals could be a campaign target but received no fundraising mail during the three week period when the survey was conducted. To account for these two distinct sources of overdispersion we use a zero-inflated model which simultaneously estimates the likelihood of an individual receiving a certain number of fundraising appeals using a logit model and a negative binomial model (Cameron and Trivedi 2005). The logit model accounts for the first possible source of overdispersion (not being a campaign target), while the negative binomial predicts the expected number of fundraising mailers given the likelihood of being a campaign target and controls for the second source of overdispersion (heterogeneity). For all models presented here the ZINB passes the Vuong test which indicates that the zero-inflated model is more appropriate than the normal NBRM (Vuong 1989). To choose between a zero-inflated PRM (ZIP) and zero-inflated NBRM (ZINB) requires estimating the NBRM and calculating the dispersion (a). If a — 0 then the NBRM reduces to the PRM. If p < 0.05 that a does not equal 0 then the NBRM is the more appropriate choice because it is likely that the data are overdispersed due to heterogeneity. An alternative test is a likelihood ratio test where G2 — 2(ln LNBRM—In LPRM) is a test of HO: oe — 0 (see Long 1997, 237). In the all models presented in Table 1 with the exception of the model of the number of material appeals received, the alpha is significant atp < 0.01, thus we can reject the null hypothesis that a — 0 and conclude that the NBRM is the more appropriate choice because it is likely that the data are overdispersed. Alternatively, using the likelihood ratio test the null hypothesis a — 0 is also rejected (p < 0.01). In the case of the model using material appeals, while the Vuong test is significant, the likelihood ratio tests are not significant and the results presented revert to a ZIP.
2 For model simplicity and precision we remove income and strength of party identification from future iterations of the first stage of the model and education and age from future iterations of the second stage of the model (Agresti and Finlay 2007). The removal of these variables does not significantly affect the goodness of fit of the model and the inclusion of these additional variables has no effect on the substantial effects of other variables in the model.
Springer
368
Polit Behav (2014) 36:359-376
given a campaign contribution versus to those who had not previously donated. Comparatively, campaigns only sent just under 0.5 more pieces of mail during that time to an individual making more than $80,000 a year compared to an individual who made less than $40,000 a year and about 0.3 more pieces of mail to extreme partisans as opposed to moderates. Figure 2 shows the same results as in Table 2, but with 95 % confidence intervals included.
While campaigns do send more mail to individuals who are older, wealthier, better educated, and more partisan, these factors are no more important and possibly less important than an individual's past contribution history.
How Campaigns Ask
Successful campaign fundraising also involves targeting the right people with the right message (Hart 1992; Kanfer 1991; Shea and Burton 2006). Models 3, 4, and 5 in Table 1 model the number of fundraising solicitations an individual receives of a certain type. As in models 1 and 2, we use a zero-inflated negative binomial to account for the two possible processes that lead to an observation of an individual receiving no fundraising mail of a certain type.
As we should expect, models 3, 4, and 5 in Table 1 show that the decision to make an individual a fundraising target is similar to the overall model shown in model 2. A campaign bases its decision to include an individual on its target list largely on that individual's past contribution history. There are, however, small variations in a campaign's decision to target individuals with material and solidary messages. While individual's with a college education and more political group memberships are more likely to be targeted with ideological messages, these attributes have no effect on the likelihood that a campaign will identify an individual as a target for solidary or material messages. Instead, the campaigns appear to be aiming these appeals towards a wider population in attempts to draw in new donors.
The second stage of the model shows how campaigns use different characteristics to distinguish the quantity of appeals they choose to send to certain individuals once they have identified potential targets. While campaigns are more likely to identify previous contributors as potential targets, the effect of having donated previously does not have an effect on the quantity of solidary and material appeals that they receive conditional on being a target. While campaigns send more ideological appeals to reactivate donors who have participated in the past, campaigns send non-donors the same amount of solidary and material appeals as past donors in an attempt to include them in the campaign through less polarizing methods.
Likewise, an individual's partisanship does play a role in the number of ideological and solidary appeals that campaigns send to that individual, however, it has no effect on the total number of material appeals. Campaigns appear to be using
3 Again, for model simplicity and precision we remove income and strength of party identification from the first stage of the model and education and age the second stage of the models predicting the volume of appeals sent to an individual of a certain type (Agresti and Finlay 2007). We did run models with the inclusion of these variables and their addition does not significantly affect the goodness of fit of the model and has no effect on the substantial effects of other variables in the model.
Springer
Table 1 Total number of fundraising solicitations received
Stage 1 (logit model)
	(1)	(2)	(3)	(4)	(5)
	Fundraising target	Fundraising target	Fundraising target	Fundraising target	Fundraising target
Past contributor	-1.171*	-0.921**	-1.053**	-1.107**	-1.231*
	(0.685)	(0.420)	(0.417)	(0.474)	(0.724)
Pol. group membership	-0.415*	-0.320**	-0.342**	-0.069	0.164
	(0.220)	(0.160)	(0.161)	(0.117)	(0.123)
College education	-1.121	-1.010**	-0.953**	-0.722	-0.100
	(0.818)	(0.452)	(0.454)	(0.557)	(0.490)
Age	-0.053**	-0.056***	-0.062***	-0.052***	-0.068***
	(0.022)	(0.014)	(0.015)	(0.015)	(0.016)
Extreme partisan	0.468 (0.585)				
Extreme ideology	-0.558 (0.884)				
Middle income (40-80K)	0.535 (0.787)				
Wealthy income (80K+)	1.195 (1.014)				
Constant	4iig***	5.072***	5.615***	6.311***	5.818***
	(1.525)	(0.954)	(0.987)	(1.078)	(1.118)
Table 1 continued
Stage 2 (negative binomial model)
	Total appeals	Total appeals	Ideological appeals	Solidary appeals	Material appeals
Past contributor	0.661	0.648**	0.595**	0.349	0.603
	(0.403)	(0.301)	(0.301)	(0.449)	(0.587)
Pol. group membership	0.078	0.078	0.074	0.047	0.182***
	(0.064)	(0.062)	(0.061)	(0.077)	(0.056)
Extreme partisan	0.842***	0.724***	0.812***	0.769**	0.303
	(0.312)	(0.237)	(0.239)	(0.342)	(0.285)
Extreme ideology	0.133	0.260	0.223	-0.199	0.401
	(0.445)	(0.402)	(0.392)	(0.470)	(0.413)
Middle income (40-80K)	0.703*	0.647**	0.702**	0.547	0.777**
	(0.382)	(0.285)	(0.290)	(0.465)	(0.331)
Wealthy income (80K+)	1.287**	1.049***	1.165***	1.698***	0.638*
	(0.592)	(0.303)	(0.307)	(0.458)	(0.374)
College education	0.370 (0.501)				
Age	0.001 (0.016)				
Constant	-1.947*	-1.301***	— 1.416***	-1.486***	-1.890***
	(1.109)	(0.322)	(0.331)	(0.542)	(0.563)
Observations	735	735	735	735	735
Log-likelihood	-489.3	-492.4	-459.2	-200.3	-218.4
Cragg and Uhler's R-squared	0.193	0.182	0.204	0.178	0.224
Vuong	2.990	3.102	3.483	2.427	2.944
Standard errors in parentheses
*** p < 0.01, ** p < 0.05, * p < 0.1
Polit Behav (2014) 36:359-376
371
solidary and ideological appeals to entice individuals who already feel an association with the party to become more involved.
In contrast, campaigns send more material appeals to individuals who do not necessarily have stronger partisan affiliations but who belong to more political groups. Political group membership has no effect on the number of solidary appeals that campaigns send to an individual. Given the rise of political groups oriented around material and ideological benefits and the decline of membership organizations that provide social benefits (Putnam 2000; Skocpol 2003) it should not be
Table 2 Marginal effects of total number of fundraising solicitations received due to changes in significant independent variables
Variable	Change in IV		Increase in solicitations received
Past contributor	0-1		0.7935**
Income	<$40,000 to	between $40 and 80K	0.2367*
Income	<$40,000 to	>$80,000	0.4827**
Extreme partisan	0-1		0.3397**
Age	10 years		0.1107**
College education	0-1		0.1951**
Political group membership	1 additional	group membership	0.1016**
** p < 0.01, * p < 0.05
Extreme Partisan i-•-
Political Group Membership >—•—i
College Education i-•-1
Income i-•■
Past Contributor i-
-0.1 0.1 0.3 0.5 0.7 0.9 1.1 1.3
Change in Number of Fundraising Appeals Received
Fig. 2 Effect on the total number of fundraising appeals an individual receives
Springer
372
Polit Behav (2014) 36:359-376
surprising that campaigns use information about an individual's political affiliations to better target material appeals rather than solidary appeals.
The combined effects of both stages of the model are displayed in Table 3. We find that the strongest predictor of the quantity of fundraising pieces an individual receives is that individual's past contribution history. Campaigns overwhelmingly target fundraising appeals of all types towards those who have previously given. We also find that campaigns send significantly more solidary and ideological appeals to respondents in the highest income bracket. Consistent with research indicating the importance of social connections among wealthy donors, campaigns also appear send more material and ideological appeals to those in the middle income bracket while targeting solidary appeals disproportionately towards those in the highest income bracket (Sinclair 2012). Although the difference is not significant, campaigns send a slightly larger number of material appeals to individuals in the middle income bracket than those in the highest income bracket confirming our theory that campaigns attempt to motivate lower income individuals who do not have the financial resources to feel capable of donating for solidary or ideological reasons, but may be responsive to material incentives.
All of this information suggests that campaigns send more material appeals and solidary appeals in an attempt to broaden their fundraising base, while campaigns send more ideological appeals to donors with a previous donation history. When trying to appeal to those who have not donated in the past, campaigns target wealthy and partisan individuals with more solidary appeals, while targeting middle income individuals and individuals with a number of group memberships with material appeals. Previous research has suggested that campaigns use ideological appeals as their primary form of solicitation during the waning moments of the campaign (Godwin 1988; Hassell 2011). Thus variations from this norm appear be geared towards attracting different sets of potential donors.
Discussion
This unique dataset of mail that individuals received 3 weeks prior to the election in 2004 has allowed us to measure not just the increase in the probability that a campaign contacts a potential donor, but the change in the volume of mail the individual receives. Because of this, we have a clearer understanding of who campaigns target and how often these campaigns contact certain types of potential donors during the last 3 weeks of the election. While past research has only been able to theorize about the influence of past donations on the likelihood of donor mobilization, we show that a donor's past contribution history is the strongest predictor of future solicitations. We find that an individual's income and strength of partisanship play less of a role in the decision to send an individual a solicitation and more of a role in the number of solicitations that an individual receives.
The implication of the campaign targeting is that the "unequal" voice of participation in campaign contributions is not simply resource based, but rather that campaigns contribute to the situation by targeting past donors and by sending more solicitations to those with higher levels of income and stronger partisan affiliations.
Springer
Polit Behav (2014) 36:359-376
373
Table 3 Marginal effects of total number of ideological, solidary, and material fundraising solicitations received due to changes in significant independent variables
Variable Change in IV
Ideological appeals
Past contributor
Income
Income
Extreme partisan Age
College education Political group membership
Past contributor
Income
Income
Extreme partisan Age
College education Political group membership
Past contributor
Income
Income
Extreme partisan Age
College education Political group membership
0-1
<$40,000 to between $40 and $80K
<$40,000 to >$80,000
0-1
10 years 0-1
1 additional group membership
0-1
<$40,000 to between $40 and $80K
<$40,000 to >$80,000
0-1
10 years 0-1
1 additional group membership
0-1
<$40,000 to between $40 and $80K
<$40,000 to >$80,000
0-1
10 years 0-1
1 additional group membership
Avg. solicitations received: .425 Increase in solicitations received
0.7221**
0.2263*
0.4906**
0.3441**
0.1146**
0.1699**
0.0963**
Avg. solicitations received: .131 Increase in solicitations received
0.2669*
0.0403
0.2468**
0.0979*
0.0529**
0.0626
0.0132
Avg. solicitations received: .110 Increase in solicitations received
0.2806**
0.0829*
0.0631
0.0346
0.0476**
0.0069
0.0095
Solidary appeals
Variable Change in IV
Material appeals
Variable Change in IV
** p < 0.01, * p < 0.05
Campaign decisions to target individuals who have donated in the past with direct mail requests for donations perpetuates and limits the pool of individuals who give to a campaign. Although the rise of internet fundraising may allow campaigns to broaden that pool of donors at a little cost, present day campaigns still rely heavily on direct mail as a key source of campaign money. Because campaigns give wealthier and more partisan individuals more opportunities to contribute they should be more likely to donate. As a result, they should also be more likely to stay in the pool of donors that campaigns will target in future election cycles than moderates and those with lower income levels.
Springer
374
Polit Behav (2014) 36:359-376
We have also shown how campaigns target their message to potential donors based upon the characteristics of individual donors. Although campaigns primarily use ideological appeals in their direct mail fundraising towards the end of the campaign, they also use solidary and material appeals to target specific potential donors. Our findings paint a picture of savvy political organizations who mobilize and target donors not on the basis of resources alone, but also on the basis of the potential donor's characteristics and past donation history, using messages that resonate with those particular individuals.
Acknowledgments The data collection was funded in part by a grant from the Pew Charitable Trusts to the Center for the Study of Elections and Democracy (CSED) at Brigham Young University (BYU). Additional funding for data acquisition and research assistance was provided by CSED and the College of Family, Home, and Social Sciences at BYU. We thank members of the Center for the Study of Elections and Democracy at BYU and the Human Nature Group at UC-San Diego for helpful comments. All shortcomings remain our own.
Conflict of interest   The authors declare that they have no conflicts of interest.
References
Agresti, A., & Finlay, B. (2007). Statistical methods for the social sciences. New York: Pearson. Arceneaux, K., & Nickerson, D. W. (2009). Who is mobilized to vote? A re-analysis of seven randomized
field experiments. American Journal of Political Science, 53(1), 1-16. Bartus, T. (2005). Estimation of marginal effects using margeff. Stata Journal, 5(3), 309-329. Brady, H. E., Schlozman, K. L., & Verba, S. (1999). Prospecting for participants: Rational expectations
and the recruitment of political activists. American Political Science Review, 93(1), 153-168. Butler, L. W. (2005). Claiming the mantle: How presidential nominations are won and lost before the
votes are cast. Cambridge, MA: Westview Press. Cameron, A. C, & Trivedi, P. K. (2005). Microeconometrics: Methods and applications. New York:
Cambridge University Press. Cho, W. K. T., & Gimpel, J. G. (2007). Prospecting for (campaign) gold. American Journal of Political
Science, 51(2), 255-268.
Culnan, M. J., & Regan, P. (1995). Privacy issues and the creation of campaign mailing lists. The
Information Society, 11(2), 85-100. Cummings, N., & Cummings, G. (2004). Strategy and tactics for campaign fundraising. In J. Thurber &
C. Nelson (Eds.), Campaigns and elections: American style (pp. 67-82). Boulder, CO: Westview
Press.
Dawson, P. A., & Zinser, J. E. (1976). Political finance and participation in congressional elections.
ANNALS of the American Academy of Political and Social Science, 425(1), 59-73. Dillman, D. A. (2000). Mail and internet surveys. New York: Wiley.
Francia, P. L., Green, J. C, Herrnson, P. S., Powell, L. W., & Wilcox, C. (2003). The financiers of congressional elections: Investors, ideologues and intimates. New York: Columbia University Press.
Fritz, S., & Morris, D. (1992). Gold plated politics: Running for congress in the 1990s. Washington, DC: CQ Press.
Gerber, A. S., Gimpel, J. G., Green, D. P., & Shaw, D. R. (2011). How large and long-lasting are the
persuasive effects of televised campaign ads? Results from a randomized field experiment.
American Political Science Review, 105(01), 135-150. Gerber, A. S., & Green, D. P. (2000). The effects of canvassing, telephone calls, and direct mail on voter
turnout: A field experiment. American Political Science Review, 94(3), 653-663. Gerber, A. S., Green, D. P., & Shachar, R. (2003). Voting may be habit-forming: Evidence from a
randomized field experiment. American Journal of Political Science, 47(3), 540-550.
Springer
Polit Behav (2014) 36:359-376
375
Ginn, M. H. (2005). Prospecting for gold: Finding financial resources for your organization. Public Manager, 34(1), 11-13.
Godwin, R. K. (1984). The implications of direct mail for political organizations. Social Science
Quarterly, 65(3), 829-839. Godwin, R. K. (1988). The structure, content, and use of political direct mail. Polity, 20(3), 527-538. Goldstein, K. M., & Freedman, P. (2002). Lessons learned: Campaign advertising in the 2000 elections.
Political Communication, 19(1), 5-28. Goldstein, K. M., & Ridout, T. N. (2002). The politics of participation: Mobilization and turnout over
time. Political Behavior, 24(1), 3-29. Grant, J. T., & Rudolph, T. J. (2002). To give or not to give: Modeling individuals' contribution
decisions. Political Behavior, 24(1), 31-54. Green, D. P., & Shachar, R. (2000). Habit-formation and political behaviour: Evidence of consuetude in
voter turnout. British Journal of Political Science, 30(4), 561-573. Han, H. C. (2008). Does the content of political appeals matter in motivating participation? A field
experiment on self-disclosure in political appeals. Political Behavior, 31(1), 103-116. Hart, N. (1992). Buddy, can you spare a grand? Campaigns & Elections, 14(9).
Hassell, H. J. G. (2011). Looking beyond the voting constituency: A study of campaign donation solicitations in the 2008 presidential primary and general election. Journal of Political Marketing, 10(1), 27-42.
Herrnson, P. S. (2012). Congressional elections: Campaigning at home and in Washington (6th ed.).
Washington, DC: CQ Press. Hillygus, D. S., & Shields, T. G. (2008). The persuadable voter: Wedge issues in presidential campaigns.
Princeton, NJ: Princeton University Press. Jones, E. T., Kropf, M., Wiedlocher, L., & Battles, L. (2007). The candidates depart but the dance goes
on: The 2004 presidential campaign in Missouri. In D. B. Magleby, J. Q. Monson, & K. D. Patterson
(Eds.), Dancing without partners: How candidates, parties, and interest groups interact in the
presidential election (pp. 97-112). Lanham: Rowman and Littlefield. Jones, R. S., & Miller, W. E. (1985). Macro level innovation and micro level response. Western Political
Quarterly, 38(2), 187-210. Kanfer, R. (1991). Direct to the bank. Campaigns & Elections, 13(7).
King, G. (1989). Unifying political methodology: the likelihood theory of statistical inference. New York:
Cambridge University Press. Levine, A. S. (2009). Perceptions of relative income and the decision to contribute, in Meetings of the
midwest political science association, Chicago. Levine, A. S., and Monson, J. Q. (2011). Campaign microtargeting and its unintended consequences on
participation decisions, in Meetings of the midwest political science association, Chicago. Long, J. S. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks,
CA: Sage Publications, Inc. Lovett, M., and Peress, M., (2010). Targeting political advertising on television. University of Rochester
Working Paper.
Magleby, D. B., Monson J. Q., and Patterson K. D. (2006). Mail communications in political campaigns: The 2004 campaign communications survey, in Meetings of the midwest political science association, Chicago.
Magleby, D. B., & Patterson, K. D. (2008). The battle for congress: Iraq, Scandal, and campaign finance
in the 2006 election. Boulder, CO: Paradigm Publishers. Malbin, M. J. (2006). A public funding system in Jeopardy: Lessons from the presidential nomination
contest of 2004. Election Law Journal, 5(1), 2-22. Malchow, H. (2003). The new political targeting. Washington, DC: Campaigns and Elections Magazine. Miller, J. M., Krosnick, J. A. and Lowe L. (2000). The impact of policy change threat on financial
contributions to interest groups. Ohio State University Working Paper. Miller, J. M., & Krosnick, J. A. (2004). Threat as a motivator of political activism: A field experiment.
Political Psychology, 25(4), 507-523. Monson, J. Q., & Curtis, S. P. (2004). Studying the noncandidate campaign: Case study and survey
methodology. In D. B. Magleby & J. Q. Monson (Eds.), The last hurrah? Soft money and issue
advocacy in the 2002 congressional elections (pp. 281-299). Washington, DC: Brookings Institution
Press.
Springer
376
Polit Behav (2014) 36:359-376
Monson, J. Q., & Oliphant, J. B. (2007). Microtargeting and the instrumental mobilization of religious conservatives. In D. E. Campbell (Ed.), A matter of faith: Religion in the 2004 presidential election (pp. 95-119). Washington, DC: Brookings Institution Press.
Olson, M. (1965). The logic of collective action: Public goods and the theory of groups. Cambridge: Harvard University Press.
Panagopoulos, C, & Bergan, D. (2006). Contributions and contributors in the 2004 presidential election
cycle. Presidential Studies Quarterly, 36(2), 155-172. Price, V., & Zaller, J. (1993). Who gets the news? Alternative measures of news reception and their
implications for research. Public Opinion Quarterly, 57, 133-164. Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community. New York:
Simon and Schuster.
Rosenstone, S. J., & Hansen, J. M. (1993). Mobilization, participation, and American democracy. New York: Macmillan.
Sabato, L. (1981). The rise of political consultants: New ways of winning elections. New York: Basic Books.
Shea, D. M., & Burton, M. J. (2006). Campaign craft: The strategies, tactics, and art of political
campaign management. Westport, CT: Praeger. Sinclair, B. (2012). The social citizen. Chicago: University of Chicago Press.
Skocpol, T. (2003). Diminished democracy: From membership to management in American civic life.
Norman: University of Oklahoma Press. Sorauf, F. J. (1992). Inside campaign finance. New Haven: Yale University Press. Verba, S., Schlozman, K. L., & Brady, H. E. (1995). Voice and equality: Civic voluntarism in American
politics. Cambridge: Harvard University Press. Vuong, Q. H.  (1989). Likelihood ratio tests for model selection and non-nested hypotheses.
Econometrica, 57(2), 307-333. Wilcox, C. (2008). Internet fundraising in 2008: A new model? The Forum, 6(1), 1-13. Wilson, J. Q. (1974). Political organizations. New York: Basic Books.
Springer