THE STEALTH CAMPAIGN:

EXPERIMENTAL STUDIES OF SLATE MAIL IN CALIFORNIA

Shanto Iyengar, Stanford University

Daniel H. Lowenstein, University of California, Los Angeles

Seth Masket, University of California, Los Angeles

Prepared for Presentation at the Annual Meeting of the American Political Science Association, Atlanta, GA., September 2-5, 1999. This paper will be presented at Panel 34-16, Thursday, 1:30-3:15 p.m.

I. INTRODUCTION

Although scholars have devoted considerable study to campaign advertising in recent elections, their attention has been directed primarily to television (Ansolabehere and Iyengar 1995; Jamieson 1991; Diamond and Bates 1992). Much less attention has been given to mail as a campaign medium. Even the occasional study of mail in politics tends to focus on mail as a fundraising device used by candidates, parties, PACs and interest groups (Godwin 1988). Scholars have taken for granted that the high cost of mail makes it an ineffective means of mass communication (O'Shaughnessy and Peele 1985).

In fact, direct mail is often the most economical way for a candidate to communicate with voters. No doubt for that reason, mail is used extensively in congressional and many other campaigns, and its use has been increasing (Weaver-Lariscy and Tinkham 1996). It is true that the cost of sending mail to a household is much greater than the cost of broadcasting a television ad to that household. But many metropolitan areas feature multiple districts and political jurisdictions. Usually, these boundaries are not geographically congruent with the local television market. For example, Los Angeles County alone contains all or part of 17 congressional districts, 14 state senate districts, and 25 assembly districts (California Secretary of State 1998). A candidate for one of these offices who wishes to purchase television ads on stations operating within the Los AngelesADI media market would pay to reach the entire county and several adjoining counties. With rare exceptions, such candidates do not advertise at all on television. Electoral congestion also makes it difficult for candidates to obtain free media coverage, and the districts are much too large for a campaign to rely predominantly on candidate appearances. Campaigns at this level must rely primarily on mail.

If scholars have given too little attention to campaign mail, they have entirely overlooked the particular form of mail phenomenon known in California as "slate mail." Although political "slates" are known in other states, they usually consist simply of cards or similar materials containing a list of candidates nominated by a political party. Typically, they are produced and distributed by a political the party, on a relatively small scale. In California, although occasional slate mailers occasionally originate from parties, they are usually published by private entrepreneurs, some of which whom mail millions of pieces per election.

Although we believe slate mail of this sort is unique to California, it may not remain so. As we shall demonstrate, slate mail provides valuable benefits to campaigns as well as a modestly profitable business for the publishers, most of whom publish their slate mail as an adjunct to other campaign consulting work. Sooner or later, slate mail on the California model is likely to could easily spread to other large states.

Whether or not it does, there are at least three reasons for studying California slate mail. First, a campaign practice that receives substantial resources in the nation's largest state is worthy of study for its own sake by scholars who wish to understand American elections. Second, slate mail has at times been a controversial practice in California, attracting considerable legislative and judicial attention. The study of slate mail therefore may have direct policy consequences implications. Third, slate mail can be used as a vehicle for studying numerous questions of interest to political scientists. A single piece of slate mail urges a vote for candidates in races from the most visible to the most obscure, in partisan and nonpartisan races, and for or against both well known and obscure ballot propositions. Slate mail can provide clues to voter behavior in low salience elections, an aspect of electoral choice far less well understood that are perhaps more representative of electoral choice than the more widely researched presidential and statewide contests. Further, slate mail is especially amenable to experimental research; the content of slate mail can be manipulated in order to study the effects of any number of important voting cues, including endorsements from varying individuals and organizations, partisanship, the candidate’s ethnicity and gender, , of candidates, the content of ballot propositions, and various other types of symbols and appeals.

This paper is a preliminary consideration of slate mail. We provide some basic information about the nature and use of slate mail in California and about the policy debates that have surrounded it in the last decade or two. We then briefly describe a program of experimental research into the effects of slate mail and report and analyze in detail the results of one of these such experiments.

II. SLATE MAIL IN CALIFORNIA

A typical slate mailer consists of a 5.5-inch by 8.5-inch post card. The side with the voter's address is also likely to include small photographs of some of the candidates appearing on the slate, and short endorsements of candidates or propositions. There may also be information on the date of the election and the location of the voter's polling place, as well as other relevant logos or graphics. On the reverse side is the slate itself, which is a listing of all the endorsed candidates and yes or no recommendations on ballot measures. As is discussed in the next section of this paper, at the bottom of one of the sides is a box containing required notices. A typical slate fitting this description from the 1998 primary is reproduced on the following page.

Slate mailers may depart from this standard form in various ways. Some are on larger post cards and some are folded over, so that there are two outside pages each 5.5 by 8.5 inches and one inside page twice as large. Occasionally, slates are contained inside envelopes. Slates also differ from each other in additional ways including their "glossiness." The largest slate mailings usually occur in conjunction with the statewide primary and general elections in even-numbered years. Slates in these elections often have a distinct partisan orientation, so that they may contain expressions such as "Attention Democrats!" or "Attention RepublicansRepublican Voter Guide,!" and they may be ornamented with donkeys or elephants and contain pictures of party heroes such as Franklin Roosevelt or Ronald Reagan. Others, including many slates sent in conjunction with local elections, are avowedly nonpartisan or independent. Still others may emphasize an single issue emphasis, such as pro-choice, environmentalist, or law and order.

It is apparent from the foregoing that slate mailers typically contain relatively little information about any particular candidate or ballot proposition. Why, then, are campaigns willing and often anxious to spend money to be included in slate mailers? There are several reasons, but first and foremost is that slate mail is by far the cheapest form of large-scale advertising. We have already seen that for candidates running in districts substantially smaller than the media markets in which they are located, mail generally is the only economically feasible medium. For statewide candidates and propositions, and for others whose districts more or less coincide with a media market, television will be a cheaper medium than mail that is dedicated only to a single campaign. But the costs of slate mail are shared by many campaigns. Accordingly, it is the cheapest form of mass communication. For many other campaigns -- for example, judicial candidates in large counties, candidates for many local offices, and ballot measure campaigns that attract only modest funding -- slate mail is the only feasible advertising medium. For still others, such as many ballot propositions and many candidates for statewide office below the level of governor, a minimal amount of television advertising may be possible, but this must be supplemented by slate mail advertising.

Low cost is the foremost advantage of slate mail for campaigns, but by no means the only one. Probably the next most important benefit is the particular ability of slate mail to convey a campaign message grounded in partisan, coalitional politics. Although some of the causes advocated on a slate mailer -- presidential, senatorial and gubernatorial candidates and an occasional well-publicized proposition -- are typically known to voters, most of the candidates and propositions on the slate are not. By means of the prominent causes that are advocated and the symbols and messages contained on the slate, the slate identifies itself with a broader cause or point of view. Most commonly, it is the cause of Republicans or Democrats, but it can also be as general as "good government" or as specific as a particular position on a particular issue. The implied message to the voter is that if you are the kind of person, for instance, who likes mainstream Democrats like Bill Clinton and Gray Davis, then you will probably like the other, less familiar candidates and causes advocated in this mailer. If the appeal is a partisan one, then this message is not necessary for partisan candidates in the general election. They will be identified as Democrats or Republicans merely by their presence on the ballot itself. But for candidates in primaries and in nonpartisan elections and for ballot measure campaigns, the partisan identification can help to compensate for the lack of a partisan cue on the ballot. This ability of slate mail to identify a particular cause with a broader collectivity or "in group" explains why even well-funded initiative campaigns that can and do rely heavily on television advertising also typically spend substantial amounts on slate mailers.

A corollary of the collective-partisan or coalitional approach used by slate mailers as well as the compactness of the message is that slate mail lends itself to actual use by the voter at the polling place. Although some smaller slates make recommendations covering only a fraction of the candidates and propositions on the ballot, the largest and most sophisticated seek to match as completely as possible the choices that will appear on the ballot for each household. Doing so presents a technical challenge, because of the uneven overlap of congressional, senate and assembly districts, as well as myriad boundaries for county supervisors, city councils, and governing boards of special purpose districts ranging from schools to mosquito abatement. But publishers regard it as well worth the effort, for the closer the fit between the slate mailer and the ballot, the more likely that some voters will find it useful enough to take with them to the polling place. There is no way of determining the number of voters who do so, but there is anecdotal evidence based on the a "garbage can test" that at least some voters do. This test assumes that if numerous copies of a slate mailer appear in garbage cans near polling places, the mailer probably was quite effective (Smith, 1998).

A third benefit provided by slate mailers is the targeting of messages to voters. To date, however, this has been more of a potential benefit than a real one. The only slate mail program that has used relatively sophisticated targeting has been was the one published by Michael Berman and Carl D'Agostino, described below. Their mailings typically contained an innocuous endorsement message, such as "Jones is the best Democrat for the Assembly", -- that might be signed by an especially influential source -- a prominent Jewish congressman, a woman, an environmental organization, the county sheriff, , or a generic Democrat, --depending on the location and demographics of the household. No other slate mail organization has developed such fine-grained targeting. However, technology has made it possible for marketers of commercial products and services to target their mailings with sophistication that far surpasses the methods that were used by Berman and D'Agostino, and it may be a matter of time before similar techniques are used by slate mail publishers.

These benefits of slate mail are sufficient to attract substantial campaign resources. According to one account, slate mailers were originated in the 1950s by the late -Congressman from San Francisco, Phil Burton, when he was beginning his political career. Burton could not win the endorsement of the San Francisco County Democratic Central Committee in his early primary contests for state assembly, and so he invented the slate mailer as a means of associating himself with other Democrats and thereby attracting partisan votes (Jacobs 1994). Thereafter, slate mailers were produced sporadically, including several in the 1970s by the two Democratic political operatives mentioned above, Michael Berman and Carl D'Agostino. In 1981, they formed a political consulting firm, Berman and D'Agostino Campaigns -- -universally referred to as BAD Campaigns -- and published slate mail in every primary and general election through the 1994 primary. During the firm's heyday, the BAD Campaigns slates were by a substantial margin the largest in the state. In 1990, which may have been their peak year, they published and distributed 8 million slate mailers in the primary and 9 million in the general. These pieces mailers included hundreds of thousands of variations. The candidates and ballot measure campaigns that participated in these slate mailers paid a total of $5 million in fees (Miller 1992).

Despite the disappearance of BAD Campaigns from the ranks of slate mail publishers, the rise of numerous competitors in the late 1980s and early 1990s has prevented any diminution in slate mail activity as a whole. Beginning in 1988, slate mail organizations have been required to register and file financial statements with the Secretary of State. 35 Thirty-five suchsuch organizations reported activity in 1988 and 67 in 1992 (Murphy 1992). In 1994, 78 slate mail organizations were registered (Hayward 1994). However, not all the organizations that were registered or reported activity participated in the state primary or general elections. By our count, 31 slate mail organizations reported spending in connection with either or both of the 1994 primary and general elections, and the same number reported spending in 1996. In 1998, the number increased to 59. slate mail organizations reported spending in connection with either or both of the 1998 primary and general elections. The figures from these years are not fully comparable. The 1994 figure probably included some organizations that were registered but not active, and the 1988 and 1992 may have included a few that were active in local elections but not the statewide primary or general. Despite the problems in comparing these figures, it appears fairly clear that the number of slate mail publishers increased rapidly in the late 1980s and early 1990s and has continued to grow, albeit perhaps irregularly,either flattened or grown more slowly since then during the present decade.

Slate mail publishers reported spending a total of $7,818,339.85 in connection with the 1998 primary and $11,405,201.45 in the 1998 general, for a 1998 total of $18.8 million,752,893.30 in 1998. These expenditures reflected a significant increase over the $7.3 million spent in 1996 and the $8.9 million in 1994. The decline from 1994 to 1996 is largely attributable to the fact that BAD Campaigns was still active in the 1994 primary. In addition, expenditures on slate mail and, to a lesser extent, the number of slate mail publishers active, are likely to fluctuate from election to election depending on factors such as the presence of ballot propositions generating large campaign expenditures.

Despite such transient factors, the more general picture is one of growth in spending over the last several elections. Clearly, California campaigns are devoting significant resources to slate mail. We turn next to the policy and legal controversies that have surrounded the practice. [SHANTO: There will follow some more analysis of the expenditures for 1994, 96, and 98, a better estimate of the trend in number of publishers than is contained in the preceding paragraph, and a rough estimate of the number of pieces sent in those years, with an explanation of how we derived it. I hope to discuss these a bit with more with Cisca Mok, who prepared the spreadsheet with which I'm working.]

It is clear, then, that California campaigns are devoting significant resources to slate mail. We turn next to the policy and legal controversies that have surrounded the use of slate mail.

I11. LEGISLATION AND LITIGATION

"(Slate] mailers have become big business in California" (Federighi 1992, 569). So wrote one critic, but the claim is a common one (Hayward 1994). It is also false. As we have seen, the revenues of the largest state mail publisher in a good year came to five million dollars, and the combined expenditures of 59 slate mail publishers in the most recent election year were less than nineteen million dollars. Far from being "big business," slate mail is an activity made up of small, entrepreneurial firms, most ran by one or a small number of individuals.

A related claim is that slate mail is "commercial" (Walters 1994) or published merely for profit (Feldman 1992). It is true that slate mail publishers hope to get paid for their time and effort, as do the journalists and spokespersons for public interest organizations who make such charges. However, most slate mail publishers are, actively engaged in politics, particularly as campaign consultants. Like most people in politics, they pursue a variety of goals: ideological, partisan, and personal, as well as pecuniary. No critic has ever attempted to show that the publishers receive excessive profits (however that term might be defined) relative to the time and skill they dedicate to the task. Nor is it likely that they do receive excessive profits, given the modest amounts -- from a business perspective, though not from a political perspective -- expended on slates and the intense competition within the industry.

Aside from charges of this sort that consist primarily of rhetoric and name-calling rather than substantive analysis, critics' most important claim is that slate mail is deceptive. Of course, any campaign medium can be used to convey false or misleading information, and the fact that slate mail is no exception is no basis for condemning it or regulating it. But critics claim that there are two ways in which slate mail is pervasively and characteristically deceptive. First, they claim that slate mail publishers go to considerable lengths to make it appear that their mailings originate from the official political parties (Federighi 1992; Heinicke 1997). Second, they claim that the causes endorsed on slate mail represent nothing more than the highest "bids" -- the candidates or ballot measure campaigns that were able to pay the most (Federighi 1992),

In a sense, both these claims represent the dark side of one of slate mail's greatest virtues, its fitness as a vehicle for coalitional campaigning. The primary message conveyed by a slate mailer is that all the causes advocated are consistent with the general thrust of the mailer, be it partisan, ideological, or civic. To the extent the mailer urges a vote for a candidate or for or against a proposition when that vote cannot reasonably be regarded as consistent with the general thrust, the mailer is misleading. Thus, about ten years ago critics were justified in complaining about a handful of slates that, in partisan races and without disclosure, endorsed Democratic candidates in Republican-oriented mailers and vice-versa (Simon 1989).

However, the claim that slate mail typically is deceitful in these ways is exaggerated. Over a period of nearly two decades of working with slate mail publishers, we have never heard any of them express a hope that voters would think their mailers originate from the "official" party organizations. This should not be surprising. In California, the "official" party organizations -- the state and county central committees -- have neither power nor visibility. Few voters could name a single member of any of these committees. Indeed, most voters are probably unaware of the committees’ existence. The critics of slate mail who make this charge make the common error of identifying American political parties with their "official" administrative structure. In fact, parties are amorphous entities whose political significance inheres far more in their candidates and in the loyalties of voters than in the official party organizations (Lowenstein 1993). Further undermining this criticism is the well-known fact that political parties are endowed with less than heroic properties, at least in the eyes of many voters (see Broder 1971; Wattenberg 1984; Dionne 1991). Slate mail publishers have neither need nor reason to try to mislead voters into believing mailers originate from these "official" organizations. The publishers of a partisan mailer may have reason to appeal to recipients as Republicans or Democrats, though the results we report later in this paper raise some doubts about the wisdom of this widely used strategy. Wisely or not, the publishers try hard to make it appear that the candidates and causes urged in the mailers would be likely to be favored by adherents of the party in question. To voters, "the Democrats" and "the Republicans" refer to leaders like Bill Clinton, Al Gore, Gray Davis, George W. Bush, and Pete Wilson, not to "official" party committees.

Therefore, the concern of many critics that slate mailers are designed to look like publications of official party organizations is misplaced. The more meaningful question is whether partisan-oriented slate mailers present to recipients an honestly partisan set of recommendations. To answer that question is difficult, because with the exception of races for partisan office in general elections, there is no ascertainably "correct" Republican or Democratic position. Most publishers of partisan slates have extensive ties within their parties. In addition, their clientele consists largely of partisan candidates and campaign consultants. Partisan political considerations necessarily have a major influence on which causes are endorsed on a slate mailer. But partisan considerations are by no means the only component of the endorsement decision. The question of whether the partisan orientation of slate mail is a legitimate facilitator of coalitional politics or a deceptive veneer thus leads us to the second claim made by critics, regarding the integrity of endorsement decisions.

That second claim--that slate mail endorsements are for sale to the highest bidder -- is also exaggerated. Slate mail publishers vary in their practices, but all need to take a variety of considerations into account when deciding which causes to endorse (Smith 1998). One of the considerations is likely to be which candidate or which side of a ballot measure campaign is willing and able to pay its share of the cost of the mailing. Even that consideration is not purely a matter of profit to the publisher. The more revenue is generated, the more extensive the mailing can be and therefore the more useful it proves to all the participants. However, the publisher must also be concerned about the consistency of the particular endorsement with the overall message of the mailing. When Proposition 209 -- the initiative to eliminate racial and gender preferences in public education, employment and contracting -- was on the ballot, Democratic slate mail publishers who supported it or Republicans who opposed it would have sacrificed the credibility of their mail in the eyes of many recipients. Thus, the popularity of candidates or ballot items is an important consideration in determining inclusion or exclusion. Other factors include the tendency to prefer repeat clients or candidates and causes that the slate mail publisher may represent in some other consulting medium (as mentioned previously, most slate mail publishers engage in other consulting activities). Furthermore, the campaign industry is small enough that the future prospects of a slate mail publisher can be affected by reputation. For all these reasons, publishers face a variety of political constraints on which causes they will endorse. True, after all political considerations are taken into account--including personal conviction, which, if strongly held, is likely to be decisive for most publishers-- there are likely to be some candidate races and propositions on which the publisher is more or less indifferent. In such cases, most publishers will make the choice on the basis of which side is willing to pay its share. Even here, the bidding war depicted by critics of slate mail rarely if ever occurs. According to slate mail publishers we have talked to, the usual practice is to make an offer to one side. The other side is approached only if the first side turns down the offer.

Whether or not the criticisms directed at slate mail axeare well-founded, the California state legislature has responded by enacting a variety of disclosure requirements. Proposition 208, the campaign finance initiative approved by the voters in 1996, contains additional requirements, and these that are currently the subject of constitutional litigation.

The legislature's first and major regulation of slate mail was enacted in 19971987. The statute imposed two types of disclosure obligations on slate mail publishers, who are referred to in the statute as "slate mail organizations." First, they must register with the Secretary of State and file disclosure reports that are essentially equivalent to the reports filed by campaign committees. Second, they must include certain disclosures or notices, commonly referred to as "disclaimers," on the face of the slate mailers. Aside from the name and address of the publisher, the mailer must contain a box with the following statements, printed in bold-face 8-point type.

The disclaimers required by the 1987 legislation thus address the critics' concerns in several ways. Their concern that slate mailers will be mistaken for "official" party publications is addressed by the first sentence, which is emphasized by appearing in all capital letters. The concern that endorsements are bought and paid for is addressed by the requirement that the candidates and ballot measure campaigns who have paid the publisher be identified by an asterisk. The second sentence of the required notice cautions the voter against the assumption that all causes supported by the slate mailing are necessarily mutually supporting.

In the 1990s the statute was amended to require a slate mailer with. a partisan orientation to point out specifically any candidate endorsed for a partisan office, if the candidate is either the nominee of a different party or a candidate in a primary whose opponent has been endorsed by the party state central committee. Perhaps because, as we have seen, the regulations did nothing to diminish the proliferation of slate mailers, critics remained dissatisfied. The proponents of Proposition 208, a campaign finance initiative approved by the voters in 1996, inserted three provisions revising the slate mail regulations. The first requires that the box containing the notice to voters appear two or more times depending on the format of the mailer, rather than once under the 1987 legislation. The second requires that candidates and propositions who have paid a share of the cost of the mailer be identified with three dollar signs ($$$) rather than with an asterisk (*). The third requires that if a ballot measure campaign has paid to participate in the mailer, the two largest contributors of $50,000 or more to that campaign be identified on the face of the mailing. The third requirement applies to all paid advertising in the ballot measure campaign, not just slate mail. However, it falls especially onerously on state mailers, which may include advertising on numerous ballot propositions and therefore could be required to identify a long list of contributors in a limited space.

Slate mail publishers, although not necessarily enthusiastic about the 1987 legislation, generally accepted it. Before the law was enacted many of them had adopted the practice of using an asterisk to identify identifying causes that had paid to participate with an asterisk before the law was enacted and most were publishing disclaimers comparable to those required by the statute. However, they perceived the changes in Proposition 208 as gratuitous, hostile, and damaging. Accordingly, in 1997 they filed a constitutional challenge. Their case was consolidated with four other cases challenging various parts of Proposition 208 and, after a trial, U.S. District Judge Lawrence Karlton declared certain provisions unconstitutional and preliminarily enjoined enforcement of the entire proposition without reaching the merits of many of the challenges, including the challenges to the slate mail provisions. The Ninth Circuit Court of Appeals affirmed the preliminary injunction but returned the case to Judge Karlton for him to rule on the remaining issues. On August 13, 1999, Judge Karlton ordered the parties to identify any additional evidence they wish to introduce and called for a status conference on November 10. No ruling on the merits is likely until well into 2000, meaning that Proposition 208 will remain enjoined at least through next year's primary and possibly through the general election. In the meantime, a new campaign finance initiative with additional slate mail provisions, sponsored by maverick Republican Ron Unz, is likely to appear on the ballot in 2000.

It was the litigation over Proposition 208 that gave rise to the research project we describe in the next section of this paper. One of the co-authors of this paper (Lowenstein) is serving as co-counsel for the slate mail plaintiffs. One of the other co-authors (Iyengar) was retained by the plaintiffs to perform an experimental study of the effects of the additional disclosure boxes and the identification of paying candidates/causes with dollar signs rather than asterisks. During the course of conducting that study and presenting the results to the court, we became convinced that the experimental study of slate mail could serve several useful purposes. Knowledge of the specific consequences of various disclosure requirements would be relevant to the immediate issues in the litigation and, more broadly, to issues concerning the regulation of political speech. By manipulating specific attributes of slate mail, we could assess not only the effectiveness of slate mail generally, but also the influence of these attributes on voter support for different categories of candidates and ballot questions.

Following the study conducted for the lawsuit over Proposition 208, we conducted two follow-up experiments in connection with the 1998 primary and general elections. The first of these focused on the effects of identifying "paid" candidates and proposition campaigns with dollar signs instead of asterisks. The second of these studies was designed to assess the general effectiveness of slate mail and the differential effectiveness, if any, of mailers that feature more prominent partisan appeals. It is this study that is reported in detail in the following sections.

IV. RESEARCH DESIGN

We designed a field experiment to assess the impact of slate mail in the 1998 general elections. The experiment was administered in the city of Santa Monica, an affluent coastal suburb of Los Angeles, during the last week of the campaign.

Santa Monica is a Democratic stronghold (Tom Hayden served as Mayor in the 1970s). Our intervention was based on an actual slate mailer, a "voter information guide" mailed to all registered Democrats and independents in Santa Monica. The mailers were prepared by Voter Information Guide of Sherman Oaks, a major Democratic consulting firm. The guide endorsed the full slate of Democratic candidates running for statewide and local office and indicated a preferred position on several statewide initiatives. The versions mailed to Democratic and independent voters were exactly the same except for a banner which stated either "Attention Democrats" or "Attention Independent Voters."

We modified the voter guides to create "partisan" and "nonpartisan" versions. (A copy of each version is provided in the Appendix.) The partisan version was based on, though not identical to, the actual version that had been mailed to registered Democrats. It was not identical because we added an explicit reference to the impeachment issue in the "closing statement." The nonpartisan version was identical to the version that had been mailed to registered independent voters (and did not contain any reference to impeachment).

We had two purposes in varying the partisan appearance of the guide. First, we wished to test for differences in source credibility (and hence persuasiveness) under partisan and nonpartisan labels. Slate mail represents a classic persuasion situation: the more credible the organization endorsing the candidate or initiative, the greater the probability of acceptance, given exposure. There were two possible credibility effects: of credibility. On the one hand, given the generally low repute of political parties in the current milieu, an ostensibly nonpartisan mailer may have greater potential, ceteris paribus, to sway voters than one that appears to be overtly partisan. If this is true, slate mailers would have an incentive to disassociate themselves from party organizations. On the other hand, since voters are known to favor in-group cues, we might expect acceptance or rejection of mail endorsements to vary with recipients’ party affiliation, Democrats and independents preferring to follow the partisan and nonpartisan versions, respectively. Accordingly, the use of two formats for the voters guide allowed us to assess these competing possibilities.

Our secondary objective in varying the partisan appearance of the guide was to examine the ability of mail campaigns to link state and local races with salient national events. The 1998 elections were the first ever to occur in the midst of an ongoing presidential impeachment process. The fate of President Clinton had been the paramount news story for months, easily overshadowing the issues and candidacies in California. This context may have "nationalized" the 1998 elections. As suggested by any number of pundits, voters may have punished Republicans for their zeal in pursuing impeachment of a popular President. As Washington Post columnist E. J. Dionne Jr. put it, "Citizens finally got to say at the ballot box what they've been trying to tell Washington through the polls: However much they disapprove of Bill Clinton's transgressions, they do not think the Monica Lewinsky affair is the driving issue in American politics and they don't want Clinton impeached" (Dionne 1998; also see Apple 1998; Online News Hour, November 4, 1998; for evidence suggesting a backlash against Republican candidates among moderate swing voters, see Abramowitz 1999).

Our design allowed us to shed some light on the nationalization of the 19981988 California elections, at least from the perspective of the behavior of Democratic and independent voters, by observing whether supporters of President Clinton were especially likely to vote Democratic when given the partisan version. More specifically, we varied the closing statement of the voter guide to increase the salience of impeachment as a relevant issue for the 1998 elections. In the partisan version, the voter guide ended by exhorting readers to "END THE WASHINGTON WITCH HUNT. TELL THE REPUBLICANS TO GET BACK TO THE BUSINESS OF RUNNING THE COUNTRY." In place of this charged partisan message, the nonpartisan guide merely asked readers to "USE THIS GUIDE WHEN YOU VOTE FOR GOVERNOR AND OTHER IMPORTANT OFFICES AND ISSUES." The explicit reference to impeachment politics in the partisan version would, we anticipated, "prime" voters with their feelings about President Clinton and the behavior of his critics. Voting choices among participants exposed to the partisan version would, we expected, be more closely connected to their impeachment-related attitudes.

In addition to varying the apparent partisanship of the voter guide, we embedded a "time of response" manipulation into both versions. Typically, slates are mailed during the last week of the campaign. As indicated previously, some voters (the more loyal partisans, perhaps) take the guides with them to the polling place so that they may vote "correctly." A larger number, no doubt, place them in the recycling bin immediately on receipt. We attempted to capture some of the naturally occurring variability in attentiveness to campaign mail by asking one-half of the participants in the treatment conditions to take the guide home with them and to mail back the "ballot" indicating their voting choices immediately following the election. Of the 80 participants in this "take home" condition, 53 sent in their ballots, for a response rate of 66 percent.

The remaining "treatment" participants (N=79) were handed the guide immediately after completing a brief pretest questionnaire dealing with their personal background. They were asked to look it over on their own before completing the ballot. The fact that this condition guaranteed some level of exposure to the guide coupled with the short time lag between exposure and elicitation of the posttest responses led us to expect that any potential effects of the manipulation would be enhanced. Conversely, we suspected that the effects of the manipulation would be weakened among participants whose actual level of exposure was unknown and who completed the posttest following the election.

The design also included a control condition in which participants were asked to complete the ballot without having previously seen the guide. These participants were told that the researchers were conducting a survey of voter preferences. All told, therefore, the design consisted of five conditions: the baseline (control) group, the partisan and nonpartisan versions of the voter guide, and the immediate and delayed response conditions. The experiment was conducted on two groups of subjects, Democrats and independents, under each condition. A total of 209 subjects participated in the study.

A pair of trained research assistants recruited subjects from a popular shopping area in Santa Monica. Potential subjects were screened for their place of residence, citizenship, and political affiliation (only Democrats and independents were eligible). They were informed that the study concerned the effects of "campaign mail" on different types of voters’ attitudes and that they would be paid $10 in cash. Those who agreed to participate completed an informed consent form and a brief pretest questionnaire dealing with their socio-economic background, level of media exposure, and political interest. Subjects were then handed the guide and posttest ballot and informed that they were free to consult the guide as they saw fit before returning the completed ballot to the research assistant. We used a condensed form of the official state ballot used in Santa Monica. On returning the ballot, subjects were paid and debriefed.

The study participants, who were drawn from a weekend farmers’ market on the Third Street Promenade area of Santa Monica, proved reasonably representative of the population of Santa Monica. In comparison with the population of Santa Monica, our sample proved somewhat older (with a median age of 44 as compared with 41 for the population), and relatively educated. Respondents without college degrees were underrepresented by some five percent, while the percent of respondents with graduate training exceeded the population norm by 15 percent. Despite the overrepresentation of the well educated, our respondents were no more affluent than the typical city resident. The 1990 median annual family income for Santa Monica residents was $51,085; the median in our sample fell in the group reporting annual family earnings of between $31,000 and $50,000. The gender composition of our sample also mirrored the city population.

Finally, our participant pool showed a small ethnic bias resulting from an undersampling of Hispanics in favor of whites. Hispanics represent 14 percent of the city but only 4 percent of our sample, while whites comprise 75 percent of the city, but 83 percent of our sample.

Our dependent measures consisted of participants’ intended votes (or reported votes in the case of those in the "take home" condition) in the races for Governor, U.S. Senator, U.S. Representative, Lieutenant. Governor, Secretary of State, Attorney General, Controller, State Assembly, Insurance Commissioner, and Board of Equalization. In each case, the endorsed candidate in these partisan races was the Democrat. Participants also indicated their votes on Propositions 1, 1A, 2, 3, 4, and 5. With the exception of Proposition 4, the guide endorsed a "yes" vote.

The relatively close demographic match between our ‘sample’ and the target population provided some assurance that the observed voting pattern would be representative. In fact, as shown in Figure 1, the distribution of our participants’ votes for the various elections reliably approximated the actual results. Study participants were slightly more Democratic in their votes for candidates, and slightly less "Democratic" in their votes on the propositions. Averaging across all 16 races, the difference between the actual and sampled results was only one percent.

(Figure 1 here)

V. ANALYSIS AND RESULTS

The hypothesis underlying this study is that the importance of campaign mail varies inversely with voter information. The more voters "know" about any particular election, the lower the likelihood that their choice can be influenced by exposure to voter guides. In effect, we predict increasing returns from the experimental treatment as voters make their way through the ballot – negligible effects on voting for U.S. Senator or Governor, more noticeable effects on voting for less prominent offices, and substantial effects on proposition voting.

Campaigns generate differing amounts of information, depending upon the electoral context. Major statewide contests (such as the races for U.S. Senator and Governor), attract regular coverage from the news media (although in recent cycles, the amount of news accorded statewide races has dwindled to a trickle (see Kaplan 1999)) and an abundance of paid advertising. In these relatively visible races, voters can acquire considerable information about the candidates – who they are, their past record, general ideological posture, and so on -- at relatively little cost. The same cannot be said of races for lower-level offices. News coverage is nonexistent and the candidates cannot afford to mount significant advertising campaigns. As a result, voter information is typically limited to the candidate’s party affiliation, since this information can be acquired merely by examining the ballot.(The exception, of course, would be candidates running in nonpartisan races.) An even greater level of voter uncertainty characterizes most aballot propositions. As Lupia has pointed out, the subject matter of these elections tends to be complex and esoteric, and voters cannot readily associate the choices with partisan cues or other substitutes for factual information (Lupia 1994). In short, for purposes of assessing the effects of campaign mail, we differentiate between three levels of campaign visibility: races generally characterized by minimal information and the absence of partisan cues (Propositions 1-5), down-ballot, partisan races contested by generally unfamiliar candidates (the elections for State Assembly, Controller, Board of Equalization, Attorney General, Treasurer, U.S. Representative, and Controller), and partisan, high information races (the elections for U.S. Senator and Governor). We expect the effects of campaign slate mail to decrease as voter information increases.

We begin by presenting (see Figure 2) the overall effect of exposure to the voter guide (the effect pooled across all treatment conditions) on voter choice in each of the three categories of elections. We also decompose the overall effect into the effects of the source credibility and time-of-response manipulations. These results are presented separately for Democrats and independents.

(Figure 2 here)

The pattern of results underscores the importance of partisanship to electoral choice. In both high and low visibility candidates races, Democrats supported their party with near unanimity. Among independents, the probability of voting for a Democrat was no more than 0.5. Given the virtual consensus among Democrats, our manipulation had little leverage over their candidate preferences; if anything, Democrats with access to the guide were slightly less disciplined in their voting. The persuasive effects of the guide were more pronounced among nonpartisans in both classes of candidate elections, averaging an increase of seven percentage points.

As expected, voting on the propositions proved most susceptible to the influence of slate mail. The overall effects of exposure on vote choice were robust (p < .05), as were the distinctive effects among nonpartisans (as measured by the interaction effect, p < 05). The percentage of independents who voted in the prescribed direction rose by 20 points in the treatment conditions, as compared with a five percent increase among the ranks of the Democrats. As a result of the disproportionate effect among independents, proposition voting under conditions of exposure to the guide provided the only instance of nonpartisan vote choice: independents and Democrats who read the guide were equally likely to adopt the endorsed positions. These results suggest that under conditions of low information and the absence of partisan cues, the effects of campaign slate mail on vote choice can be pivotal.

The middle and bottom panels of Figure 2 correspond to the effects of the source credibility and time of response manipulations. The generally parallel Democratic and independent slopes for both versions of the guide suggest the absence of partisan selectivity or bias; both groups deferred to the guide more frequently when it appeared to be nonpartisan. This pattern held across all three levels of voter information although the relative advantage of the nonpartisan version was most apparent in the case of proposition voting. It appears that to be perceived as nonpartisan may be an asset for direct mail organizations.

The time of response manipulation also proved ineffective. Contrary to expectations, there was no substantial erosion of experimental effects when participants were allowed to control their own exposure to the guide. Among participants who took the guide home and mailed back their ballots, the level of voting Democratic was reduced by 8 and 4 percent respectively in the high and low visibility candidate elections. However, the result was reversed for the propositions; participants who responded after the election were more likely to have voted (by a margin of 6 percent) consistently with the guide. The fact that the effects of the guide were not weakened in the more unobtrusive and realistic form of the manipulation bolsters our confidence in the validity of the pooled effect. The observed effects of slate mail on proposition voting are not an artifact of our design: no matter how salient the manipulation, exposure to the guide has the intended effect.

In sum, we find that slate mail can be a decisive force in the typical ballot measure election. For candidates in partisan races whose party is indicated on the ballot, at least those who are running for statewide office, the benefits of mail are on a much smaller scale.

We re-estimated the pooled effects of exposure to the voter guide while controlling for participants’ party affiliation, their vote in the 1996 presidential election, and a series of other background factors known to influence voting choice. As shown in Table 1, partisanship and 1996 presidential vote together accounted for a substantial share of the variance in 1998 candidate choice in both upper and lower ballot races, but played a much smaller part in shaping initiative voting. In fact, participants' 1996 presidential vote proved completely irrelevant to their votes on the propositions. The significant impact of campaign slate mail in initiative races is thus clearly attributable to the nonpartisan nature of these contests; slate mail it fills the void created by the paucity of information and the absence of party labels.

(Table 1 here)

Finally, we consider the role of partisan rhetoric in activating voters’ feelings towards President Clinton as a criterion for their 1998 voting choices. Evaluations of the incumbent president are known to influence voting in midterm elections; the less popular the incumbent, the greater the number of seats lost by his party (see Abramowitz and Segal 1996; Atkeson and Parten 1995). We are especially interested in the consistency between participants’ 1996 and 1998 vote choices across the two versions of the guide. Given the prominent reference to impeachment politics in the partisan version, we anticipated that supporters of President Clinton would be especially likely to rally to the cause of fellow Democrats in this condition. Stated more technically, we expected a significant interaction between 1996 presidential vote and the partisan appearance of the voter guide.

Our preliminary examination of the results revealed that the effects of 1996 vote on 1998 vote were indeed strongest in the conditions featuring the partisan version of the guide. There were no clear differences in 1996-1998 vote consistency between the control and nonpartisan conditions. These conditions were therefore merged for purposes of this analysis; the relevant comparison is that between participants exposed to the partisan form of the guide and all others.

We proceeded to estimate the coefficient representing the interaction between experimental condition (operationalized as the partisan version conditions vs. all other conditions) and 1996 presidential vote. The baseline effects of 1996 vote choice and the increased magnitude of this effect (captured by the interaction effect) on 1998 voting are shown in Table 2.

(Table 2 here)

Voting for President Clinton was strongly related to voting Democratic in both upper and lower ballot candidate races even among participants who were given the nonpartisan version of the guide. In both categories of candidate races, however, exposure to the partisan version of mail doubled this already substantial effect. This difference was robust for the low visibility elections, and only marginally significant for the high visibility contests. Among participants receiving the "end the Washington witchhunt" message, support for President Clinton was a more powerful force in shaping 1988 1998 vote choice than party affiliation.

The priming effects of exposure to partisan rhetoric were limited to the partisan elections. As noted earlier, the 1996 presidential election and the 1998 propositions were essentially independent choices. The level of partisan rhetoric provided by the guide did little to alter this state; if anything, presidential vote choice was inversely related to proposition voting among participants exposed to the partisan version.

All told, the evidence suggests that campaign slate mail can heighten voter sensitivity to ongoing national issues and events. References to impeachment made Democrats and independent supporters of the president vote more staunchly Democratic in 1998. The effect was limited to partisan choices; in nonpartisan elections, the impeachment-related message was uninformative.

VI. CONCLUSION

Slate mail can be persuasive, at least in elections where voters lack alternative sources of information. In those campaigns where geographical or economic factors preclude the use of radio and television advertising, slate mail provides an effective substitute. It is true that our results show significant persuasion effects only on proposition voting. In future experiments we hope to test whether the same result will extend to nonpartisan elections, which in California include judicial elections and all local elections. Such results seem likely, as there is no reason to suppose voters are more informed about races for city councils, county committees or superior court judgeships than they are about the propositions.

Even though exposure to slate mail failed to alter voting in partisan elections, the factors underlying vote choice were influenced by the presence of explicit partisan cues. By linking the Democratic candidates with opposition to impeachment, voting Democratic became more closely associated with support for President Clinton. However, since Clinton voters were just as likely to vote Democratic on other grounds (i.e. their partisanship), the increased salience of impeachment did little to increase the level of support for Democratic candidates. Thus the presence of significant priming effects did not result in higher levels of Democratic voting.

Because slate mail is a form of campaign advertising, it is at first blush susceptible to the same family of criticisms -- that the message is inherently superficial, biased, and/or designed to mislead rather than inform voters. However, notwithstanding the similar objectives of senders of slate mail and other political advertisers, there are significant differences in their products. Unlike television advertising, which focuses on individual candidates, slate mail is a team-oriented, "public good." Lesser-known candidates are associated with better-known, popular leaders so as to raise the stock of the former. Slate mail is thus much less of a decentralized and "solo" enterprise than other forms of advertising. Since the candidates on the slate share partisanship, one goal of slate mail is to strengthen party voting and, therefore, the vitality of the political parties.

The fact that slate mail aims to boost support for multiple candidates accounts for a further difference between mail and broadcast advertising. In recent years, the tone of political advertising has turned increasingly negative, to the point that negative advertising actually represents a dominant campaign strategy (see Ansolabehere and Iyengar 1995). The nature of slate mail, however, creates a different set of incentives. Because the psychology of slate mail advertising consists of "affection through association," there is a strong positivity bias to the content of slate mail. Because many of the candidates or causes are virtually unknown, the function of slate mail is to provide positive rather than negative information. Although we have not conducted a systematic content analysis of the medium, our findings based on a review of the 1998 mailers used in our experiments are suggestive. Combining the Republican and Democratic slates we studied (our primary election study focused on Republican mailers), there were 18 "boxed" endorsements for individual candidates. In only three cases did the endorsement make reference to a negative attribute of the endorsed candidate's opponent. Thus, the positive far outweighs the negative in slate mail. This effect is actually enhanced since slate mail, by its very nature of supporting multiple candidates, also seeks to demonstrate the similarity of positions taken by the best-known, most popular figures and lesser-known candidates who are running for more minor positions. For this reason, slate mail is more substantive than televised advertising and more focused on platform or program than on individualistic strategies. Paradoxically, this is true even if there is little or no ostensive issue content in the mailer. Accordingly, we would characterize this form of campaign advertising as more positive and mobilizing than the typical television-based campaign (for discussion of the demobilizing effects of negative campaigns, see Ansolabehere and Iyengar 1995; Ansolabehere, Iyengar, and Simon 1999; for critiques of the "demobilization" thesis, see Finkel and Geer 1999; Wattenberg and Brians 1999).

The predominance of positive cues in slate mail is, we believe, one of the major factors behind slate mail publishers’ opposition to the labeling provisions of Proposition 208. The measure would require all candidates or propositions paying to be included in a mailer to be identified by three dollar signs rather than an asterisk. It also would require listing of $50,000 contributors to ballot measure campaigns endorsed on the slate. In the contemporary political arena, most observers would agree that money is not a positive symbol; to a voter lacking additional information, the sight of three dollar signs next to the candidate's face or name and references to large contributions are unlikely to elicit positive associations and might well undermine the suggestion that the voter should identify with the causes being urged. In fact, we have conducted two experiments on the effects of the labeling of paid candidates in slate mailers. The results (which we will report in a later paper) were unequivocal: voters were significantly more likely to vote for the endorsed candidate when the label was an asterisk instead of the dollar signs.

In conclusion, slate mail represents an important yet hitherto unstudied form of campaigning for large numbers of California candidates and political organizations. . . . (Dan: can you complete this last para.?) Except in the courts, the normative debate has been almost entirely one-sided, consisting of rhetorically-charged criticism of slate mail by reformers and journalists. We have suggested that their claims are exaggerated and that they have overlooked public benefits that derive from slate mail. The suggested benefits and drawbacks of slate mail, as well as slate mail's electoral effects more generally, deserve empirical testing. In addition, the empirical study of slate mail may provide leverage over many important but elusive aspects of voter behavior. The research we have conducted to date, including the experiment reported in this paper, is only a small first step.

VII. REFERENCES