TRENDS: Following the Money? How Donor Information Affects Public Opinion about Initiatives

Abstract

Citizens are typically uninformed about politics and know little about issues at stake in direct democracy elections. Government efforts to inform electorates include requiring donors to initiative campaigns to report their activities and then circulating such donor information to citizens. What effects does donor information have on citizens’ opinions? We conduct a survey experiment where respondents express opinions about initiatives in a real-world election. We manipulate whether they receive donor information, party cues, policy information from a nonpartisan expert, or no additional information. We find that donor information influences citizens’ opinions in the aggregate, with effects comparable to those of party cues and policy information. However, donor information has negligible effects on uninformed citizens, who have difficulty inferring donors’ policy interests and connecting them to their own. These results underscore the potential benefits of efforts to inform electorates via disclosure laws and highlight disparities in their effectiveness for informed and uninformed citizens.

Keywords

donors party cues political knowledge campaign finance disclosure policy information direct democracy

In states with direct democracy, citizens make policy directly as opposed to relying exclusively on government officials. The initiative and referendum entrust citizens with crucial responsibilities, including decisions about social issues (e.g., abortion, same-sex marriage, and the death penalty), spending and taxation levels (e.g., income and sales tax hikes), and institutional reforms (e.g., term limits, primary and redistricting rules). Previous research offers reason to question whether citizens can perform these responsibilities competently. Citizens are uninformed about politics in general (Delli Carpini and Keeter 1996), and they typically know little about the issues at stake in direct democracy elections (Bowler and Donovan 1998; Cronin 1989). Nonetheless, the process remains popular and no state electorate has voted to terminate direct democracy institutions once they have been implemented.

Due to the salience of many initiatives, the difficulty of changing policies implemented via direct democracy, and the absence of campaign spending limits, elite efforts to influence citizens’ opinions about initiatives are substantial. For example, fifteen groups in California spent nearly $660 million on state and local ballot measures over a ten-year period—more than they contributed directly to candidates and party committees, and spent lobbying the state legislature combined (California Fair Political Practices Commission [FPPC] 2010). Between 2000 and 2012, total spending on initiatives in the state exceeded $2 billion with three initiatives garnering more than $130 million apiece (Public Policy Institute of California 2013). In 2016, $473 million was spent on seventeen statewide ballot measures in California, more than twice what Republican presidential candidate Donald Trump reported spending on his campaign (Bollag 2016).

Concerns about the effects of such unfettered spending have led state and local governments to devote significant resources to informing citizens about their choices in direct democracy settings. One strategy has been the enactment of disclosure laws, which require those who contribute to campaigns for and against initiatives to report their activities. For example, a recent law in California requires the state’s campaign finance watchdog, the Fair Political Practices Commission (FPPC), to publicize lists of the top donors to these campaigns. The legal justification for such disclosure laws has been that information about who is funding efforts to pass or defeat initiatives enables citizens to make informed decisions about them (Garrett and Smith 2005). In its decision in Citizens United v. FEC, for example, the U.S. Supreme Court reaffirmed that voter competence is a primary reason for upholding disclosure requirements. Writing for the majority, Justice Anthony Kennedy stated that “transparency [about donors] enables the electorate to make informed decisions and give proper weight to different speakers and messages” (Citizens United v. FEC, 130 S.Ct. 876, at 914).

What effects does the information elicited by disclosure laws have on citizens’ opinions about initiatives? How do these effects compare with those of other information—such as endorsements from political parties and policy information from nonpartisan experts—that citizens frequently have access to in direct democracy elections? We theorize that citizens can use information about contributions to initiative campaigns to infer where their policy interests lie. Rather than assume that all citizens have identical policy interests and/or that the choices of more informed citizens can proxy for these (Lupia 1994), we assert that citizens have heterogeneous policy interests and are uncertain about how various initiatives relate to them. Information about contributions from donors with reputations for supporting particular policies can reduce this uncertainty and help citizens determine where their own interests lie on particular initiatives. However, the efficacy of this information will depend on citizens’ ability to accurately discern donors’ policy interests.

We test several implications of our theory by conducting a survey experiment in which respondents express opinions about eight initiatives on the 2016 general election ballot in California.¹ We randomly assign respondents to receive either information about the top donors supporting and opposing the initiatives, the Democratic and Republican parties’ official positions on the initiatives (i.e., party cues), or policy information from a nonpartisan expert. We also include a control group in which no additional information is provided. We examine the effects that donor information has on respondents’ opinions and compare these effects with those of party cues and policy information. We also measure respondents’ reaction times, as well as their perceptions of the policy views of donors, the parties, and the nonpartisan expert, to shed light on whether citizens are willing to process different types of information and/or able to connect them to their own policy interests.

By randomly assigning respondents to receive authentic donor information, party cues, and policy information in a real-world election, we overcome several limitations of previous research. First, despite widespread efforts to disseminate donor information during initiative campaigns and theoretical reasons to expect such information to influence citizens, few studies examine whether and when citizens are willing and able to use it. Second, no experimental study directly tests whether different types of citizens (i.e., informed vs. uninformed) can distinguish the policy views of particular donors and connect this information to their opinions. Third, existing research has yet to compare the effects of donor information to the effects of party cues and policy information from nonpartisan experts. This makes it difficult to know which type of information holds greater value for citizens and, by implication, where scarce government resources might have their largest impact.

We find that donor information has large effects on citizens’ opinions, particularly on initiatives where citizens have yet to form strong attitudes. The effects of donor information are comparable to those of party cues and policy information. However, unlike party cues which work for all types of citizens, the effects of donor information are negligible for uninformed citizens. Our analyses indicate that uninformed citizens process donor information systematically but have difficulty recognizing differences in donors’ policy views. They, therefore, are unable to consistently connect information about donors’ contributions to their own policy interests. Together, these results demonstrate the benefits of government efforts to inform electorates via disclosure laws, as well as disparities in their effectiveness for informed and uninformed citizens.

Can Donor Information Promote Informed Decision Making about Initiatives?

Decades of public opinion research demonstrate that citizens lack information about politics (Campbell et al. 1960; Delli Carpini and Keeter 1996). They are typically unaware of basic political facts and lack detailed information about the candidates and issues at stake in elections. Such widespread ignorance reflects the weak incentives most citizens have for acquiring information that would enable them to make informed political decisions (Downs 1957). The incentives for acquiring political information are, if anything, weaker in direct democracy settings. Initiatives typically require citizens to choose between an uncertain policy outcome and the status quo (Gerber and Lupia 1995; Lupia 1992). Citizens may see little benefit to investing in information about initiatives because they can simply allow a tolerable status quo to prevail. Furthermore, because the issues at stake in direct democracy elections are frequently complex, the costs of acquiring and assessing information may be high. Indeed, citizens are often confused about the substance of initiatives and, as a result, are less informed about initiatives than they are about candidates (Bowler and Donovan 1998; Cain and Miller 2001; Cronin 1989; Magleby 1984).

Deficiencies in citizens’ knowledge about politics in general, and initiatives in particular, have led scholars to consider whether effective substitutes for detailed political information might exist. Empirical studies show that information shortcuts like party cues (Arceneaux 2008; Bullock 2011; Kousser et al. 2015) and endorsements (Arceneaux and Kolodny 2009; Boudreau 2009; Lupia 1994) can enable citizens to make informed political decisions. Studies also find that policy information can influence citizens’ opinions even when party cues are also provided and have effects that are comparable to those of party cues (Boudreau and MacKenzie 2014; Bullock 2011; Nicholson 2011).

Despite mounting evidence of the effectiveness of information shortcuts and policy information, several unanswered questions remain. One question is whether information about contributions to initiative campaigns similarly helps citizens to make informed decisions. The Supreme Court has upheld federal and state disclosure laws even as it has struck down limits on campaign contributions. In its decision in Citizens United v. FEC, the Court emphasized that disclosure requirements are justified by the information they communicate to voters. Similarly, the Ninth Circuit in 2003 ruled that disclosure laws in direct democracy settings serve a compelling state interest by providing citizens with “a useful shorthand for evaluating the speaker behind the sound bite” (California Pro-Life Council, Inc. v. Getman; see Garrett and Smith 2005). As judges and others continue to weigh the benefits of disclosure laws against the burdens they impose on donors, evidence of the effects such information has on citizen decision making is likely to factor heavily into their deliberations.

To date, few studies empirically assess whether information about donors influences citizens’ opinions. Those that do so either focus on candidate elections or issue ads in federal elections (Dowling and Wichowsky 2013; Groenendyk and Valentino 2002; Weber, Dunaway, and Johnson 2012). We are not aware of any experimental study that examines the effects of information about contributions to initiative campaigns. Most existing studies of initiative campaigns focus on the effects of spending on the outcomes of initiative elections (e.g., vote shares, whether initiatives pass or fail; see de Figueiredo, Ji, and Kousser 2011; Rogers and Middleton 2015). Whether and when information about donors’ efforts in these campaigns will change individual citizens’ opinions remain open questions. The lack of attention to the effects of donor information in direct democracy settings is unfortunate given the large sums of money spent on initiatives and the significant resources governments have devoted to making such information available to citizens.

A second but related question is whether citizens can distinguish the policy views of various donors. Like political parties, the individuals and groups that contribute to political campaigns have reputations for supporting particular types of policies (Bonica 2013). However, citizens may be unaware of donors’ reputations and, as a result, unable to identify whether particular donors share their interests. Few studies examine whether citizens perceive differences in donors’ policy views in real-world elections. One exception is Lupia (1994), who shows that citizens can infer whether one donor (the insurance industry) shares their interests when voting on auto insurance initiatives. However, because information about the donor was not randomly assigned, it is possible that citizens differ in ways other than their knowledge of the insurance industry’s preferences (e.g., political interest, education) that could explain differences in their choices (see Arceneaux and Kolodny 2009). It is also unclear from Lupia’s study whether citizens can infer the interests of other donors on other types of initiatives. If citizens cannot consistently perceive differences among donors responsible for financing initiative campaigns, then it is not clear how information about campaign contributions will help citizens determine the proper weight that should be given to different speakers and messages, as Justice Kennedy envisioned in the Citizens United case.

Even if Justice Kennedy and others are correct about the benefits of disclosure laws, a third open question is how the effects of donor information compare to those of other types of political information—particularly party cues and policy information—studied by scholars and disseminated during campaigns. While governments devote extensive resources to collecting and publicizing information about campaign contributions, it is not clear whether doing so holds greater value for citizens than these other types of information. Empirical studies typically analyze particular types of information separately, which makes it difficult to compare their effects. While some studies directly compare the effects of party cues and policy information (Arceneaux 2008; Boudreau and MacKenzie 2014; Bullock 2011), no study examines how donor information compares with these two other types of information.

Theory and Hypotheses

Theoretical models of political decision making identify conditions under which citizens will base their decisions upon particular sources of information (Calvert 1985; Lupia 1992; Lupia and McCubbins 1998). Underlying each of these models is the reality that the costs of acquiring information about candidates or initiatives often outweigh the benefits. As a result, citizens typically forego obtaining detailed information and instead rely on the recommendations of others (Downs 1957). However, some information sources may not be credible, and citizens must figure out which sources to trust. These models predict that citizens will trust and benefit from information when two conditions are met. First, the information source must be knowledgeable about the choice, i.e., its likely consequences, and citizens must perceive the information source as such. Second, citizens must be able to identify the source’s interests and assess whether these interests are aligned with their own (Lupia and McCubbins 1998).

While endorsements from political parties and policy information from credible sources are widely viewed as satisfying these two conditions (e.g., Boudreau and MacKenzie 2014; Lupia and McCubbins 1998; Sniderman and Stiglitz 2012), it is unclear whether citizens will use and benefit from information about donors to initiative campaigns. On the one hand, information about donors lacks certain characteristics that make party cues and policy information effective. First, while many citizens have strong attachments to their political party (Campbell et al. 1960; Green, Palmquist, and Schickler 2002), attachments to donors are likely to be weak. As a result, donors’ expressions of support (as communicated by their contributions) might be perceived as less persuasive than party cues. Second, citizens may struggle to infer where their policy interests lie from lists of donors who are less well known than political parties. Third, unlike policy information, the information elicited by disclosure laws does not summarize the consequences of passing particular initiatives. Whereas information from nonpartisan experts might enable citizens to distinguish good and bad policies, information about campaign contributions says nothing about an initiative’s likely effects.

While these difficulties are real, there is also reason to believe that donor information might satisfy the two conditions above. Like political parties, the donors responsible for financing initiative campaigns often have reputations for supporting particular types of policies. The National Rifle Association (NRA), for example, is a well-known supporter of Republican candidates and causes, including efforts to reduce restrictions on firearms. The Service Employees International Union (SEIU) typically supports Democratic candidates and causes. Donors such as these, by virtue of their regular participation in policy debates, are knowledgeable about particular issues (e.g., the NRA and firearms policy). Moreover, the financial investments donors make in initiative campaigns offer costly signals that enhance their credibility. The frequent affiliation of many donors with one party combined with their investments in campaigns offer clues as to their policy views and interests in particular outcomes (e.g., the NRA prefers no ban on semiautomatic weapons). If citizens can identify donors’ interests, then information about their contributions may help citizens to determine where their own interests lie on particular initiatives.² These considerations lead to our first prediction:

Hypothesis 1: Citizens exposed to donor information will be more likely to support initiatives that donors affiliated with their own party or its causes support and oppose initiatives that donors affiliated with their own party or its causes oppose than citizens who are not exposed to donor information.

In making this prediction, we are asserting that (1) different citizens (e.g., Democrats and Republicans) may draw different conclusions from the same information and (2) donor information functions similarly to political party endorsements. We argue that the first claim is more realistic than assuming that all citizens will react the same way to donor information. Lupia’s (1994) influential observational study, for example, assumes that all citizens have identical interests (i.e., pro-consumer, anti-industry) on five California insurance initiatives.³ The second claim follows from the nature of the information elicited by disclosure laws, which identifies supporters and opponents but communicates little about the consequences of initiatives. Donor information is analogous to lists of endorsements (with amounts attached) that citizens may use to infer which position (yea or nay) better accords with their own policy interests.

The effects of donor information can depend on the nature of the policy that an initiative seeks to change. Some initiatives address policies that are highly salient, either because they implicate personal self-interest or citizens’ identifications with social groups and values (Howe and Krosnick 2017; Visser, Bizer, and Krosnick 2006). Gun control, which states and the national government have debated for decades, is an issue that invokes citizens’ partisanship and core values. Citizens have considerable information about and strong attitudes toward gun control that are resistant to change. In contrast, many initiatives address the technical details of particular state policies and programs or the institutional rules governing them. Citizens are likely to have less information about and weaker attitudes toward initiatives proposing, for example, to float bonds for public works projects or to change vote requirements for particular state policies. Research indicates that citizens are more responsive to information when their prior attitudes toward a policy are weak (Boudreau and MacKenzie 2014; Chong and Druckman 2010; Druckman and Leeper 2012). Thus, we expect the effects of donor information to vary depending on the strength of citizens’ prior attitudes about the issue that an initiative addresses.

Hypothesis 2: Donor information will have larger (smaller) effects on initiatives addressing policies about which citizens have weak (strong) prior attitudes.

The effects of donor information might also vary depending on citizens’ level of political knowledge. Like political parties, many donors have reputations for supporting particular types of policies. However, donors’ reputations are typically less well known than those of parties. Indeed, many citizens know that the Democratic (Republican) Party typically supports liberal (conservative) policies (Sniderman and Stiglitz 2012). Thus, both informed and uninformed citizens ought to be able to identify the parties’ interests and relate them to their own. In contrast, uninformed citizens may not know, for example, that labor unions like the SEIU tend to support Democratic candidates and causes or that police organizations often support Republican candidates and causes. Identifying donors’ policy views likely requires a level of political knowledge and interest that uninformed citizens lack. Furthermore, citizens (both informed and uninformed) lack strong attachments to donors. Strong party attachments give uninformed citizens a reason to follow their party’s recommendation even if they are unaware of the party’s reputation for supporting particular types of policies. Lacking such attachments to donors, uninformed citizens might be unable to use or simply ignore information about campaign contributions from donors they do not recognize. This yields our third prediction:

Hypothesis 3: Donor information will increase informed citizens’ propensity to support initiatives that donors affiliated with their own party or its causes support and oppose initiatives that donors affiliated with their own party or its causes oppose, relative to those not exposed to the information. These effects of donor information will be weaker among uninformed citizens.

In the same way that uninformed citizens might have difficulty connecting donor information to their own policy interests, we would also expect them to have trouble using policy information from nonpartisan experts. Unlike party cues and donor information, policy information from nonpartisan experts is unbiased in that the experts do not have an identifiable political or financial interest in the outcome. Citizens may trust such information because the experts are perceived as knowledgeable and interested in achieving better policy outcomes—an interest that citizens share. However, such policy information seldom provides an explicit recommendation about whether to support or oppose an initiative. Rather, it typically provides extensive detail about initiatives and clarifies their likely consequences, from which citizens may infer whether they should support or oppose them. Making such an inference requires citizens to process and comprehend sometimes complex information, assess its importance, and connect it to their own feelings about a particular outcome. Informed citizens, by virtue of their superior political knowledge and greater interest in politics, are better positioned to perform these tasks (Arceneaux 2008; Kam 2005).

Given that uninformed citizens may have difficulty using donor and policy information effectively, we would not expect large differences in the relative effects of these types of information. In contrast, the effects of donor and policy information are likely to be different for informed citizens. By virtue of their superior knowledge of and interest in politics, informed citizens are better positioned to accurately perceive donors’ policy interests. They might also be more motivated to process policy information systematically and connect it to their feelings about the likely consequences it reveals. Nonetheless, because nonpartisan experts lack a political or financial interest in the outcomes of initiative elections, informed citizens might be less likely to base their opinions on the policy information they provide. Indeed, theoretical models support the counterintuitive claim that an information source with an identifiable political or financial interest in the outcome of a choice will be more useful than advice from an unbiased expert, as long as citizens can determine how the source’s interests relate to their own (Calvert 1985). By clarifying donors’ positions on initiatives, information about their contributions enables citizens who can identify whether donors share their interests to infer whether they should support or oppose particular initiatives. This yields our fourth prediction:

Hypothesis 4: Donor information will have larger effects on informed citizens’ opinions than policy information from nonpartisan experts.

To the extent that donor information has different effects on informed and uninformed citizens, there are two possible explanations for such differences. First, informed and uninformed citizens might differ in their willingness to process donor information systematically (Eagly and Chaiken 1993).⁴ If uninformed citizens are unwilling to exert the effort to process donor information systematically, then it will not increase the time these citizens spend expressing their opinions about initiatives, relative to those not exposed to the information. In contrast, informed citizens may be more willing to engage in this effortful form of processing and, thus, will take longer to express their opinions when exposed to donor information.

Alternatively, as we emphasize above, informed and uninformed citizens might differ in their ability to use donor information effectively. Uninformed citizens might be unable to connect information about donors’ contributions to their own policy interests because they do not recognize differences in donors’ policy views. If this is the case, then we should observe small differences in uninformed citizens’ perceptions of the ideological positions of donors affiliated with the Democratic and Republican parties or their causes. In contrast, informed citizens will perceive donors affiliated with the Republican Party and its causes as more conservative than donors affiliated with the Democratic Party and its causes. Our last prediction reflects these expectations about the relative accuracy of informed and uninformed citizens’ perceptions:

Hypothesis 5: Informed citizens will exhibit more accurate perceptions of the ideological reputations of donors supporting and opposing initiatives than uninformed citizens.

Experimental Design

The 2016 general election in California provides a useful opportunity to test our hypotheses about the effects of donor information, relative to party cues and policy information, in a real-world context. In that election, the initiatives on the ballot addressed a range of policy subjects. A variety of interest groups, politicians, and private citizens contributed large sums of money seeking to secure the passage or defeat of particular initiatives. Furthermore, the Democratic and Republican parties took opposing positions on nearly all of the initiatives. California’s nonpartisan Legislative Analyst’s Office (LAO) produced detailed policy information about the likely consequences of passing each initiative. We take advantage of the different issues, donors, party positions, and policy information to assess how these types of information affect citizens’ opinions.⁵

In the survey experiment, all respondents receive short descriptions of eight initiatives presented in random order. These initiatives asked citizens to decide whether to (1) uphold a law prohibiting grocery stores from providing plastic bags to customers (Proposition 67), (2) require a public vote before the state can issue more than $2 billion in revenue bonds (Proposition 53), (3) allow inmates convicted of nonviolent crimes to be given early parole consideration (Proposition 57), (4) increase the vote requirement in the state legislature for changing how the fees that hospitals pay to Medi-Cal (California’s heath care program for low-income patients) are used (Proposition 52), (5) legalize marijuana for recreational use (Proposition 64), (6) require background checks before individuals can purchase ammunition (Proposition 63), (7) repeal the death penalty (Proposition 62), and (8) increase the cigarette tax by $2 per pack (Proposition 56). We chose these initiatives because they represent a range of policy issues and because they attracted campaign contributions from a variety of political actors, thereby enabling us to test our hypotheses about the effects of donor information.

In the control group, respondents receive only the short descriptions of the initiatives. For example, on Proposition 57, these respondents read the following:

This November, Californians will be asked to vote on a ballot measure that would allow inmates convicted of nonviolent crimes to be given parole consideration upon completion of their primary sentence. Currently, many prisoners receive both a primary sentence for a crime and “enhancements” or extra time if there are multiple victims or if they previously were in prison. This measure would allow prison officials to award credits toward early release to prisoners who demonstrate good behavior, efforts to rehabilitate themselves, or participate in prison education programs.

Respondents are then asked whether they strongly support, somewhat support, somewhat oppose, or strongly oppose the initiative, or whether they “don’t know” (see the online appendix [OA], pp. 1–4, for descriptions of the other initiatives).

In the “donor information” treatment group, respondents also receive information about up to two top donors (in terms of contributions) to the “yes” and “no” campaigns for each initiative. Such information is available on the California FPPC’s website, and a link to this information is required by law to be included in the official ballot pamphlet.⁶ The donor information that respondents receive on Proposition 57 is shown in Table 1. The donor information for the other initiatives is structured similarly, with the official names of the donors, descriptions of them,⁷ and the amounts they contributed shown in a table.⁸

Table 1.

Donor Information.

Donor	Description	Amount
Donors supporting additional parole consideration for nonviolent inmates
Governor Brown’s Ballot Measure Committee	Organization established by Governor Brown to support select initiatives	$4,138,764
Service Employees International Union (SEIU)—California State Council	California Union of state and local employees, nurses, and other workers with more than 700,000 members	$164,846
Donors opposing additional parole consideration for nonviolent inmates
Association of Deputy District Attorneys	Association representing deputy district attorneys in Los Angeles County with approximately 1,000 members	$60,000
San Diego Police Officers Association	Association representing police officers in the City of San Diego	$5,000

In the “party cues” treatment group, respondents receive the official positions that the Democratic and Republican parties took on each initiative. The two parties regularly advertise their positions on initiatives, and this election was no exception. On Proposition 57, respondents read that the Democratic Party supports allowing inmates convicted of nonviolent crimes to be given early parole consideration, while the Republican Party opposes this. The passages for the other initiatives similarly link the parties to their official positions.⁹

In the “policy information” treatment group, respondents receive information drawn from reports by California’s nonpartisan LAO about the likely consequences of passing each initiative. The LAO is an independent, nonpartisan government agency that provides fiscal and policy advice to the California state legislature. It is responsible for analyzing the effects of proposed initiatives; a summary of the LAO’s analysis of the fiscal effects of each initiative is required by law to appear on the ballot, as well as in the official ballot pamphlet. Thus, respondents receive policy information that resembles what they would actually encounter in the election. For example, on Proposition 57, respondents read the following:

This initiative would help reduce significant overcrowding problems in state prisons by increasing the number of non-violent inmates eligible for parole consideration. California’s nonpartisan Legislative Analyst’s Office estimates that this initiative could save the state tens of millions of dollars each year in correctional and other costs.

On this initiative, the policy information provides a reason for supporting it. The information for the other initiatives is structured similarly and is linked to the LAO, although whether it provides a reason for support or opposition varies across the eight initiatives.

After expressing their opinions about the eight initiatives, respondents rate three of the groups (selected at random) that we manipulate in our study on a seven-point liberal–conservative scale. These groups included the two state political parties, the LAO, and twenty-nine donors to the campaigns for and against the eight initiatives. These ratings measure the extent to which respondents perceive differences in these groups’ policy views, a necessary condition if they are to relate these groups’ interests to their own.

When completing these ratings, respondents were given a short description of the group in question, and for the donors, information about which initiative campaign the organization or individual contributed to. For example, when rating Governor Brown’s Ballot Measure Committee, respondents are told that this is an organization established by Governor Brown to support selected initiatives. They are also told that it is one of the top donors supporting an initiative that would allow prison officials to give additional parole consideration to inmates convicted of nonviolent crimes who have demonstrated good behavior. This gives respondents context for evaluating each donor and exposes them to the same information respondents in our donor information treatment group received before expressing their opinions about the initiatives. It also resembles information citizens might encounter in the real world (e.g., if they consulted the FPPC’s list of top donors and/or obtained descriptions of donors via Google).

Methods

Our analyses use a sample of 1,409 Californians from the Survey Sampling International (SSI) panel, 928 of which were assigned to our treatment and control groups.¹⁰ SSI is a survey research firm that recruits samples of adults via the Internet. We administered our survey experiment online using Qualtrics software from October 18 to October 23, 2016, approximately three weeks before Election Day (November 8, 2016).¹¹

To test our hypotheses, we first estimate a probit model that pools the eight initiatives.¹² Our dependent variable is a dummy variable that indicates whether a respondent supports a particular initiative. This variable takes the value 1 for respondents who “strongly support” or “somewhat support” an initiative and 0 for respondents who “somewhat oppose” or “strongly oppose” the initiative. This enables us to assess whether respondents are on the same side of an issue as their party and donors affiliated with their party or its causes.¹³

To examine how donor information, party cues, and policy information affect support for the initiatives, we create independent variables that reflect the nature of the signal that Democratic and Republican respondents receive on each initiative in each group.¹⁴ The variable Donor Info takes the value 1 for respondents in the donor information treatment group on initiatives where donors affiliated with their own party or its causes are financing the support side and ‒1 for respondents on initiatives where donors affiliated with their own party or its causes are bankrolling the opposition. The variable Party Cues similarly takes the value 1 for respondents in the party cues treatment group whose own party supports an initiative and ‒1 for respondents whose own party opposes an initiative. In all but one case, the coding of this variable is exactly the same as for the Donor Info variable.¹⁵ The variable Policy Info takes the value 1 for respondents in the policy information treatment group on initiatives for which the information provides a reason to support and ‒1 on those for which the information offers a reason to oppose.¹⁶ To simplify the presentation of our results, we convert the coefficients in our models to predicted probabilities and first differences. We test our hypotheses by comparing probabilities of support in our treatment groups with the control group and assessing differences in the first differences (i.e., in the effects of donor information, party cues, and policy information).

We then test our second prediction by estimating the same model separately for initiatives addressing issues about which respondents likely have weak prior attitudes and those about which they have strong prior attitudes. Four initiatives in our study involve salient issues that have been debated at the national and/or state level: legalizing marijuana for recreational use, repealing the death penalty, imposing ammunition limits, and raising (cigarette) taxes. These are policy issues about which respondents likely have considerable information and strong prior attitudes. The other four initiatives involve more esoteric state policies and programs where prior attitudes are likely to be weak: requiring a public vote before the state can issue more than $2 billion in revenue bonds, increasing the vote requirement in the state legislature for changing Medi-Cal fees, granting parole credits to nonviolent offenders, and upholding a law prohibiting grocery stores from providing plastic bags. Our classification of these initiatives is supported by evidence about the strength of attitudes toward the initiatives among control group respondents (who do not receive additional information). In the OA (Table A10, p. 60), we show that these respondents’ opinions are less extreme and exhibit a higher proportion of don’t know responses on the four initiatives that we classify as involving weak prior attitudes, relative to those involving strong prior attitudes.

To test our predictions about the effects of information on informed and uninformed citizens, we estimate an additional model that pools the eight initiatives and includes interactions between the treatment variables and an indicator of respondents’ level of knowledge about state politics. We classify respondents based on their answers to five fact-based questions about California politics. Respondents who correctly answered three or fewer questions (the median) are defined as low knowledge, or politically uninformed, while respondents who correctly answered more than this are high knowledge, or politically informed.¹⁷ We convert the coefficients in our models to predicted first differences and compare the effects of donor information, party cues, and policy information on low- and high-knowledge respondents.

Finally, we conduct difference-of-means tests of low- and high-knowledge respondents’ reaction times and their ideological ratings of the donors responsible for financing the campaigns for and against the initiatives. This enables us to assess whether differences in these respondents’ reactions to donor information reflect differences in their willingness to process the information and/or ability to connect it to their own policy interests.

Results

Our results support our predictions about the effects of donor information on citizens’ opinions about initiatives. Respondents shift their opinions in the direction recommended by donors associated with their own party or its causes, and the effects are particularly pronounced on initiatives addressing issues where respondents’ prior attitudes are weak. For high-knowledge respondents, the effects of donor information are comparable with those of party cues and larger than the effects of policy information. Unlike party cues, however, we find that donor information has minimal effects on low-knowledge respondents. These minimal effects reflect low-knowledge respondents’ inability to recognize differences in donors’ policy views, which prevents them from connecting information about donors’ contributions to their own policy interests.

The Effects of Information on Opinions about Initiatives

Consistent with Hypothesis 1, respondents in the aggregate are more likely to support initiatives that donors affiliated with their own party or its causes support and oppose initiatives that these same donors oppose. Figure 1 displays predicted probabilities of support for initiatives that donors affiliated with a respondent’s party or its causes support, a respondent’s own party supports, and the policy information provides a reason for supporting. These predicted probabilities are generated from the probit model that pools the eight initiatives (Table A1 in the OA, p. 44). In the control group, the probability of supporting these initiatives is .61. In the donor information treatment group, the probability of support (.65) is significantly greater. This positive effect of donor information provides evidence of the potential benefits of providing information about contributions to initiative campaigns via disclosure laws.

Figure 1.

The effect of information on support for initiatives.

The effects of party cues and policy information are also statistically significant. These results accord with our expectations that both types of information will influence citizens’ opinions about initiatives. In comparing the three types of information, party cues have the largest impact on citizens’ opinions. The probability of supporting the initiatives in the party cues treatment group is .71, which is significantly greater than in the donor (.65) or policy information (.64) treatment groups. Given the parties’ well-known ideological reputations and citizens’ strong attachments to them, the large effects of party cues are not surprising. The similarly large effects of donor information are more striking. Despite the absence of strong attachments to particular donors, donor information works much like party cues.

The effects of donor information are particularly pronounced on initiatives addressing issues where respondents have weak prior attitudes. Figure 2 plots predicted first differences from the two probit models of support for the subsets of initiatives where respondents have weak versus strong prior attitudes (Table A1 in the OA, p. 44). For both models, we generated the first differences for initiatives that donors affiliated with a respondent’s party or its causes support, a respondent’s own party supports, and the policy information provides a reason for supporting. As Figure 2A shows, on initiatives where respondents have weak prior attitudes, donor information increases the probability of support for initiatives that donors affiliated with a respondent’s own party or its causes support by .06. This is a significant increase relative to the control group. Party cues produce a similar, albeit larger, increase in the probability of supporting these initiatives (.11). Policy information also significantly increases the probability of supporting the initiatives (.06), relative to the control group. The positive effect of donor information offers evidence of its efficacy in shaping citizens’ views about initiatives addressing issues about which they have yet to form strong opinions.

Figure 2.

The effect of information on support for initiatives by strength of prior attitudes: (A) weak prior attitudes and (B) strong prior attitudes.

In contrast, donor information has a negative effect on initiatives addressing issues about which respondents have already formed strong attitudes. As shown in Figure 2B, donor information significantly decreases the probability of supporting the initiatives by .06, even though the donors affiliated with a respondent’s own party or its causes support these initiatives. This negative effect of donor information could arise if respondents react against the recommendations of donors on these salient issues or if some respondents are confused about the donors’ policy views.¹⁸ Party cues and policy information also have minimal effects on respondents’ opinions about these initiatives.

How Information Affects Low- and High-Knowledge Respondents

Our analyses of low- and high-knowledge respondents reveal differences in these respondents’ reactions to donor and other types of political information. Figure 3 displays predicted first differences from the probit model that pools the eight initiatives and includes interactions with respondents’ level of knowledge about state politics (Table A2 in the OA, p. 45). As with Figures 1 and 2, we generated these first differences for initiatives that donors affiliated with a respondent’s party or its causes support, a respondent’s own party supports, and the policy information provides a reason for supporting. As Figure 3A indicates, donor information has no effect on low-knowledge respondents’ opinions about the initiatives. The effects of policy information are also not significant. In contrast, party cues have meaningful and significant effects on low-knowledge respondents’ opinions.

Figure 3.

The effect of information on support for initiatives by knowledge: (A) low-knowledge respondents and (B) high-knowledge respondents.

The effects of information on high-knowledge respondents’ opinions are much stronger. As Figure 3B shows, donor information increases the probability of support for initiatives that donors affiliated with a respondent’s own party or its causes support by .10, a significant effect relative to the control group. The size of this effect is comparable to that of party cues, which increase support by .14. Consistent with Hypothesis 3, donor information has larger effects on high- than low-knowledge respondents.¹⁹ As Figure 3B indicates, the difference in the first differences for low- and high-knowledge respondents is statistically significant in the donor information group. We find a similar gap in the effects of party cues on the opinions of these two types of respondents. However, whereas the significant effect of party cues for low-knowledge respondents suggests an ability to use the information, the null effect of donor information implies that information about campaign contributions is ineffective for these respondents.

We observe more modest effects of policy information on high-knowledge respondents’ opinions. As Figure 3B shows, policy information that provides a reason for supporting the initiatives increases high-knowledge respondents’ probability of support by .05. This is a significant increase relative to the control group. Relative to donor information and party cues, however, the effect of policy information on high-knowledge respondents’ opinions is significantly smaller. Consistent with Hypothesis 4, high-knowledge respondents are more influenced by information sources with identifiable political or financial interests in the outcome of a choice than advice from an unbiased expert.

Are Uninformed Citizens Unwilling or Unable to Use Donor Information?

What explains these large differences in low- and high-knowledge respondents’ reactions to donor information? We find little evidence that low-knowledge respondents are unwilling to process donor information systematically. Low-knowledge respondents, like their more informed counterparts, take more time to express their opinions when they receive donor information. In the control group, for example, low-knowledge respondents take 18.29 seconds to express their opinions. In the donor information treatment group, they take 22.96 seconds, a significant increase (Table A7 in the OA, p. 56).²⁰

Instead, our results indicate that low-knowledge respondents are unable to perceive meaningful differences in the policy views of donors responsible for financing initiative campaigns. Figure 4 plots the average ideological ratings (with 95% confidence intervals) that low- and high-knowledge respondents assign to the political parties, LAO, and donors responsible for financing the initiative campaigns. The Democratic Party and donors affiliated with it or its causes are arrayed on the left-hand side of the figure with their ratings indicated by circles. The Republican Party and donors affiliated with it or its causes are arrayed on the right with their ratings indicated by triangles. As Figure 4A shows, there is substantial overlap in low-knowledge respondents’ ratings of left- and right-leaning donors. In a few cases, a donor affiliated with the Republican Party or its causes is actually rated as more liberal than a donor affiliated with the Democratic Party or its causes. Difference-of-means tests indicate few significant differences in the ratings of each left–right pair among the twenty-nine donors we examined (Table A4 in the OA, pp. 49–50).

Figure 4.

Ideological ratings of parties, the LAO, and donors by knowledge: (A) low-knowledge respondents and (B) high-knowledge respondents.

In contrast, Figure 4B shows that high-knowledge respondents accurately perceive large differences in these donors’ policy views.²¹ For these respondents, there is virtually no overlap in the ratings of the left- and right-leaning donors that financed the initiative campaigns. All donors affiliated with the Democratic Party or its causes have average ratings of 3.72 or less. All donors affiliated with the Republican Party or its causes have average ratings of 3.60 or more. The differences in the ratings of these donors are statistically significant in every pairwise comparison (Table A5 in the OA, pp. 51–52). As we show in the OA (Figure A2, p. 53), the ratings of high-knowledge respondents more accurately reflect reality than those of low-knowledge respondents, as indicated by a plot of these donors’ ideological positions based on their actual political contributions (see Bonica 2013).

These results provide strong evidence for Hypothesis 5 and reveal an important benefit of political knowledge: the ability to infer the policy interests of political actors. This disparity in low- and high-knowledge respondents’ ability to recognize differences in donors’ policy views explains their different responses to the donor information. Given that low-knowledge respondents cannot distinguish the interests of donors, giving them information about campaign contributions is unlikely to aid their decision making. In contrast, these respondents can distinguish the Democratic and Republican parties (the difference in their mean ratings is statistically significant), which explains their greater responsiveness to party cues than to donors.

Conclusion

Fears that citizens will not understand the choices they make in direct democracy elections and about the outsized influence of money have motivated governments to pass laws requiring campaign finance disclosures. Courts have been called upon to weigh the benefits of such laws against any burdens they impose. As the Supreme Court’s Citizens United decision illustrates, practitioners have often assumed (with little empirical evidence to guide them) that the benefits to citizen competence are large.

Our results offer concrete evidence of the effects of donor information in a real-world direct democracy election. They demonstrate that donor information can help citizens infer where their policy interests lie, with effects comparable to those of party cues. Donor information can be particularly informative on initiatives addressing issues about which citizens have yet to form strong attitudes. Nonetheless, we find disparities in informed and uninformed citizens’ ability to use donor information effectively. Unlike party cues, which work for all types of citizens, the effects of donor information are negligible for the uninformed. The reason for this disparity is that uninformed citizens have difficulty distinguishing donors’ policy views. For these citizens, a necessary condition for realizing the benefits of this information shortcut—the ability to identify groups with common interests (Lupia 1994; Lupia and McCubbins 1998)—might not be satisfied in many initiative contests.

Our results offer lessons for political scientists studying citizen competence and practitioners seeking to better inform electorates. For scholars, our study reveals the benefits of manipulating multiple types of information in a real-world election. We were able to compare the effects of donor information against party cues and policy information, as well as a “no information” baseline. Including party cues sheds light on the similarities between donor information and a widely disseminated shortcut. Both help citizens with heterogeneous policy interests determine how various initiatives relate to them. Our study also uncovers an important difference: donor information is ineffective for uninformed citizens. Finally, including policy information enables us to empirically demonstrate a counterintuitive result: an information source with identifiable political or financial interests (e.g., donors) can be more useful than advice from an unbiased expert (e.g., the LAO), as long as citizens can determine how the source’s interests relate to their own.

For practitioners, our study offers evidence of the effectiveness of some disclosure laws and circulating the donor information they provide, while drawing attention to disparities between informed and uninformed citizens. Specifically, in states that require the disclosure of donors’ actual identities, this information can help citizens to, in Justice Kennedy’s words, give “proper weight” to the efforts of major donors. Unfortunately, many states require only the disclosure of registered campaign names, which are selected to have broad appeal and disguise contributors’ true identities (Dowling and Wichowsky 2013; Lesenyie 2020; Weber, Dunaway, and Johnson 2012). Thus, while our results illustrate how citizens use donor information when donors’ identities are transparent, they offer less insight into contexts where donors can veil their identities (Garrett and Smith 2005). In these settings, concerns about “dark money” influencing election outcomes and the effectiveness of campaign finance laws are likely well founded (Oklobdzija 2019; Wood 2018).

Future efforts to increase the efficacy of campaign finance disclosures should address this practical challenge: how can the disparities in citizens’ ability to use donor information be reduced? The donor information in our experiments resembles what is available to citizens living in states, like California, that have laws requiring transparency about donor’s identities. Other legal variants should be tested. In addition, our experiments provide brief descriptions of donors that are readily obtained from a Google search. Our results make clear that even under these favorable circumstances, uninformed citizens are unable to identify donors’ interests and connect them to their own. It is possible such descriptions help politically informed citizens. While our analyses suggest otherwise (see the OA, pp. 62–68), an ideal test would be to randomize whether brief descriptions like ours or more extensive information are needed for informed and uninformed citizens to use donor information effectively. One possibility is to communicate information about campaign contributions together with information about the ideological positions of donors—measured using contributions, expert ratings, or positions on overlapping sets of policy proposals.

Whether these or other interventions will result in more informed electorates and, thereby, policies that better reflect citizens’ preferences is difficult to say. What we can say is that the extensive resources expended to inform citizens about their choices in direct democracy settings have, to date, proven to be useful, but partial, solutions. Our study, the first to examine donor information alongside party cues and policy information, identifies differences in how such information affects informed and uninformed citizens. It also offers guidance as to how we might improve their effectiveness and study them in the future.

Supplemental Material

sj-docx-1-prq-10.1177_1065912921990744 – Supplemental material for Following the Money? How Donor Information Affects Public Opinion about Initiatives

Supplemental material, sj-docx-1-prq-10.1177_1065912921990744 for Following the Money? How Donor Information Affects Public Opinion about Initiatives by Cheryl Boudreau and Scott A. MacKenzie in Political Research Quarterly

Footnotes

Acknowledgements

We thank Sharif Amlani for valuable research assistance. We also thank participants in the Experimental Political Science Workshop at the University of California, Davis and at the Center for Behavioral Political Economy at Stony Brook University for valuable feedback.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was generously funded by grants from the Institute for Social Sciences, the Public Opinion Workshop, and the Committee on Research at the University of California, Davis.

ORCID iD

Cheryl Boudreau

Notes

Supplemental Material

Replication materials for this paper can be viewed at . The online appendix for this article is available at the Political Research Quarterly website.

References

Arceneaux

Kevin

. 2008. “Can Partisan Cues Diminish Democratic Accountability?” Political Behavior 30 (2): 139–60.

Arceneaux

Kevin

Kolodny

Robin

. 2009. “Educating the Least Informed.” American Journal of Political Science 53 (4): 755–70.

Barber

Michael J.

2016. “Ideological Donors, Contribution Limits, and the Polarization of American Legislatures.” Journal of Politics 78 (1): 296–310.

Bollag

Sophia

. 2016. “California Ballot Propositions Rake in a Record $473 Million in Campaign Cash.” Los Angeles Times, November 7. http://www.latimes.com/politics/la-pol-ca-november-proposition-money-20161107-story.html.

Bolsen

Toby

Druckman

James N.

Cook

Fay Lomax

. 2014. “The Influence of Partisan Motivated Reasoning on Public Opinion.” Political Behavior 36:235–62.

Bonica

Adam

. 2013. “Ideology and Interests in the Political Marketplace.” American Journal of Political Science 57 (2): 294–311.

Boudreau

Cheryl

. 2009. “Closing the Gap.” Journal of Politics 71 (3): 964–76.

Boudreau

Cheryl

MacKenzie

Scott A.

2014. “Informing the Electorate?” American Journal of Political Science 58 (1): 48–62.

Bowler

Shaun

Donovan

Todd

. 1998. Demanding Choices. Ann Arbor: University of Michigan Press.

10.

Bullock

John

. 2011. “Elite Influence on Public Opinion in an Informed Electorate.” American Political Science Review 105 (3): 496–515.

11.

Cain

Bruce E.

Miller

Kenneth P.

2001. “The Populist Legacy.” In Dangerous Democracy: The Battle over Ballot Initiatives in America, edited by Sabato

Larry J.

Ernst

Howard R.

Larson

Bruce A.

, 33–62. Lanham: Rowman & Littlefield.

12.

California Fair Political Practices Commission. 2010. Big Money Talks. Sacramento: Fair Political Practices Commission.

13.

California Pro-Life Council, Inc. v. Getman, 328 F. 3d 1088 (9th Cir. 2003).

14.

Calvert

Randall

. 1985. “The Value of Biased Information.” Journal of Politics 47:530–55.

15.

Campbell

Angus

Converse

Philip

Miller

Warren

Stokes

Donald

. 1960. The American Voter. New York: John Wiley.

16.

Chong

Dennis

Druckman

James N.

2010. “Dynamic Public Opinion.” American Political Science Review 104 (4): 663–80.

17.

Citizens United v. Federal Election Commission, 558 S. Ct. 310 (2010).

18.

Cronin

Thomas E.

1989. Direct Democracy. Cambridge: Harvard University Press.

19.

de Figueiredo

John M

Chang Ho

Kousser

Thad

. 2011. “Financing Direct Democracy.” Journal of Law, Economics, & Organization 27:485–514.

20.

Delli Carpini

Michael X

Keeter

Scott

. 1996. What Americans Know about Politics and Why It Matters. New Haven: Yale University Press.

21.

Dowling

Conor M.

Wichowsky

Amber

. 2013. “Does It Matter Who’s behind the Curtain?” American Politics Research 41 (6): 965–96.

22.

Downs

Anthony

. 1957. An Economic Theory of Democracy. New York: HarperCollins.

23.

Druckman

James N.

2004. “Political Preference Formation.” American Political Science Review 98:671–86.

24.

Druckman

James N.

Leeper

Thomas J.

2012. “Learning More from Political Communication Experiments.” American Journal of Political Science 56 (4): 875–96.

25.

Eagly

Alice H.

Chaiken

Shelly

. 1993. The Psychology of Attitudes. Fort Worth: Harcourt.

26.

Francia

Peter L.

Green

John C.

Herrnson

Paul S.

Powell

Lynda W.

Wilcox

Clyde

. 2003. The Financiers of Congressional Elections: Investors, Ideologues, and Intimates. New York: Columbia University Press.

27.

Garrett

Elizabeth

Smith

Daniel A.

2005. “Veiled Political Actors and Campaign Disclosure Laws in Direct Democracy.” Election Law Journal 4 (4): 295–328.

28.

Gerber

Elisabeth R.

Lupia

Arthur

. 1995. “Campaign Competition and Policy Responsiveness in Direct Legislation Elections.” Political Behavior 17:287–306.

29.

Green

Donald

Palmquist

Bradley

Schickler

Eric

. 2002. Partisan Hearts and Minds. New Haven: Yale University Press.

30.

Groenendyk

Eric W.

Valentino

Nicholas A.

2002. “Of Dark Clouds and Silver Linings.” Communication Research 29 (3): 295–319.

31.

Howe

Lauren C.

Krosnick

Jon A.

2017. “Attitude Strength.” Annual Review of Psychology 68:327–51.

32.

Kam

Cindy D.

2005. “Who Toes the Party Line?” Political Behavior 27 (2): 163–82.

33.

Kousser

Thad

Lucas

Scott

Masket

Seth

McGhee

Eric

. 2015. “Kingmakers or Cheerleaders?” Political Research Quarterly 68 (3): 443–56.

34.

Lesenyie

Matthew

. 2020. “Reading the Fine Print.” American Politics Research 48 (1): 155–74.

35.

Lupia

Arthur

. 1992. “Busy Voters, Agenda Control, and the Power of Information.” American Political Science Review 86 (2): 390–403.

36.

Lupia

Arthur

. 1994. “Shortcuts versus Encyclopedias.” American Political Science Review 88:63–76.

37.

Lupia

Arthur

McCubbins

Mathew D.

1998. The Democratic Dilemma. Cambridge: Cambridge University Press.

38.

Magleby

David B.

1984. Direct Legislation. Baltimore: Johns Hopkins University Press.

39.

Nicholson

Stephen P.

2011. “Dominating Cues and the Limits of Elite Influence.” Journal of Politics 73 (4): 1165–77.

40.

Oklobdzija

Stan

. 2019. “Public Positions, Private Giving.” Research and Politics (January – March): 1–8.

41.

Public Policy Institute of California. 2013. “The Initiative Process in California.” Just the Facts, October 2013. San Francisco: Public Policy Institute of California.

42.

Rogers

Todd

Middleton

Joel

. 2015. “Are Ballot Initiative Outcomes Influenced by the Campaigns of Independent Groups?” Political Behavior 37 (3): 567–93.

43.

Sniderman

Paul M.

Stiglitz

Edward H.

2012. The Reputational Premium. Princeton: Princeton University Press.

44.

Visser

Penny S.

Bizer

George Y.

Krosnick

Jon A.

2006. “Exploring the Latent Structure of Strength-Related Attitude Attributes.” Advances in Experimental Social Psychology 38 (1): 1–67.

45.

Weber

Christopher

Dunaway

Johanna

Johnson

Tyler

. 2012. “It’s All in the Name: Source Cue Ambiguity and the Persuasive Appeal of Campaign Ads.” Political Behavior 34 (3): 561–84.

46.

Wood

Abby K.

2018. “Campaign Finance Disclosure.” Annual Review of Law and Social Science 14:11–27.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

2.43 MB