Voter Attitudes towards Corruption and Government before a…

To view this page ensure that Adobe Flash Player version 9.0.124 or greater is installed.

Please login or register to make a comment!

Voter Attitudes towards Corruption and Government before and after Enron Results from Exit Polling in a Midwestern City David P. Redlawsk Department of Political Science University of Iowa Iowa City, IA 52242 david-redlawsk@uiowa.edu Abstract During the past two years issues surrounding corporate fraud and political corruption exploded into the news as one after another large corporations became the centers of scandal. In a number of cases, Enron, for example, these scandals had potential political implications as company executives and politicians were seen to be cozying up to each other.

This paper reports the results from two exit polls, one during the 2000 election and the other during 2002, carried out in Iowa City, IA. These polls were designed to understand voter attitudes towards political corruption, and included a battery of questions about voters' attitudes towards corruption. The paper describes the nature and effects of attitudes towards corruption from before the current wave of scandals and since, focusing specifically on how attitudes have evolved.

Paper prepared for presentation at the Annual Meeting of the Midwest Political Science Association, April 3-6, 2003, Chicago, IL 1 Voter Attitudes towards Corruption and Government before and after Enron Results from Exit Polling in a Midwestern City * ... more. less.

Enron. Tyco. Adelphia. Arthur Anderson. Captains of industry led away in handcuffs as the stock market collapses decimating paper fortunes and real pensions. Companies collapsing. Thousands thrown out of work. During 2001 and 2002 the headlines trumpeting corporate corruption and ties to political money were at times nearly overwhelming. A Harris Poll taken in late July 2002 reported that 68% of adults were at least familiar with reports of recent corporate scandals while 41% of adults believed that President Bush had done something dishonest or unethical as a businessman Even so, the public did not appear to directly blame Bush as only 30% believed he was at fault in the corporate scandals. At the same time, a 49% plurality felt the campaign finance system was in some way responsible at least in part. (Taylor, 2002b.) There is little doubt that at least for a time, the wave of corporate scandals not only shook the confidence of the public in business, but also may have contributed to a general sense that politicians themselves were corrupt. Much of that came from the Enron situation, where at least seven members of the Bush administration were connected to the company through stock ownership, previous employment, or, perhaps most importantly, the receipt of generous campaign contributions (O 9Harrow, 2003.) And, in a less visible but potentially politically relevant case, Haliburton, which Vice-president Dick Cheney led from 1995 to 2000, began to be investigated for potentially illegal financial dealings (Forest and McNamee, 2003.) * Thanks to Doug Rose for his work in the initial development and coordination of this project in 2000. Travis Lowe provided excellent research assistance for the 2000 election, as did Kimberly Briskey and Matthew Shultz for the 2002 poll in Iowa City. Karen Emmerson dug up data on the actual election results and US Census information, and spent a great deal of time documenting various ethical breeches, especially in corporations, that occurred between 2000 and 2002. And thanks, especially, to those students in my local politics class (2000) and political campaigning class (2002) who stood out at the polls trying to convince voters to participate in this study. 2 The corporate scandals came just as the political system was perhaps crecovering d from the Presidential election debacle of 2000. In a contested presidential election ultimately decided by the Supreme Court, George Bush squeaked out the narrowest of Electoral College victories while losing the popular vote by over 500,000 votes. From the public 9s point of view, the outcome of the election may have led to questions about the integrity of the process. The person responsible for certifying the contested Florida vote count, Secretary of State Katherine Harris, was Bush 9s campaign co-chair in the state. The Governor was Bush 9s brother. The decision of the Supreme Court hinged on the vote of Justices appointed by Bush 9s father. A Harris Poll found, as might be expected, a strong partisan divide about whether the votes had been counted correctly, with 58% of Democrats believing as well that the Supreme Court decision was political, while 70% of Republicans thought the ruling was impartial justice (Taylor, 2000.) In the same poll, 86% of respondents felt there was ca major problem in the way votes are cast and counted and this needs to be fixed. d It would not be surprising if the combination of the election of 2000 and the wave of corporate scandals left citizens less than enamored of both the political and business systems. Throw in a few more events that made headlines 3 Congressman Gary Condit 9s relationship with missing intern, Chandra Levy; Clinton 9s controversial pardons at the end of his term, which many felt were connected to campaign contributions; and the debate over campaign finance reform in Congress during which many of the cdirty secrets d of campaign finance were aired 3 and it may seem that the only rational thing for anyone to believe was that government and corporations were both the corrupt domains of crooks. If so, then voters in 2002 may have had a very different perspective on government and business than they had only two years before. 3 Yet countervailing trends may also have occurred, with the attacks of September 11, 2001. Prior to the attacks, there was a great deal of focus on both the inconclusive 2000 election and the developing corporate corruption scandals. Afterwards, of course, for quite a while little else seemed to matter except terrorism and the protection of the homeland. The percentage of Americans who had ca great deal of confidence d in the White House more than doubled from its January 2001 reading of 21% to 50% in January 2002 (Taylor, 2002a.) There was also a very large jump in presidential approval ratings (Taylor, 2001) with Bush 9s overall approval remaining quite high throughout 2002. And in a reflection of either a significant increase in trust in government or simply fear for one 9s own safety, a two to one majority of citizens polled by Harris in late September 2001 supported security measures (including phone tapping and the like) that might erode civil liberties, believing perhaps that the government could be trusted to use the measures properly (Taylor, 2001.) Events between the 2000 and 2002 elections clearly buffeted the public, but in which direction? Did voters become more cynical, more condemning of what they view as corrupt activities, more willing to believe that corrupt means are actually necessary to get things done in government? Or did the crally 8round the flag d effect engendered by September 11 override the scandals? Certainly, Republican Congressional campaigns of 2002 did everything they could to invoke a patriotic theme, while distancing from the corporate scandals and campaign finance debates. 1 To the extent that voters pay attention to campaigns it may be that the various corruption-oriented scandals became relatively unimportant as they receded into the past and as 1 In a broad study of competitive House and Senate election in 2002, Magleby and his associates documented the types of television ads and mailings that made up the campaign from both interest groups and candidate campaigns. They found a substantial reliance on September 11 related themes, particularly efforts to question the patriotism of some Democratic members of the House and the Senate. 4 the Democrats failed to focus on them. 2 On the other hand, it might also be that voters carried with them a greater sense of distrust as they went to the polls in 2002 compared to 2000. One way to assess the extent to which voters attitudes towards corruption changed between the elections of 2000 and 2002 is to actually talk to voters. This paper reports on the results of exit polls carried out in a small Midwestern city during the 2000 and 2002 general elections. The exit polls were designed to tap the attitudes of voters towards political corruption, broadly defined, and to develop an understanding of how citizens view what is and is not corrupt and whether any real political consequences flow from those attitudes. Because a number of key questions about political corruption were repeated on the two polls, this study can look at changes that may have occurred in the tumultuous political environment between these two elections. Prior Research on Corruption and Politics It is certainly important to begin by understanding what is meant by ccorruption d. Corruption has certainly been a concern of political scientists (and many others) for a very long time. Brooks (1909) suggested that cin the whole vocabulary of politics it would be difficult to point out any single term that is more frequently employed than the word 8corruption 9 d Wilson (1966) believed that it is reasonable to expect public officials to engage in corrupt activities because (1) governing under the checks and balances system cannot happen without corruption, (2) members of groups once in office can be expected to give favor to other members of the same groups and (3) because cmen steal when there is a lot of money lying around loose and no one is watching. d Even so, a clear definition of what is meant by corruption has been elusive. Much 2 Perhaps because many were implicated in some ways as much as any Republican. Enron, as do many donors, spread its money around, with some going to Democrats as well as Republicans. And of course, campaign finance scandals revolved around the Clinton administration too. 5 like Justice Stewart Potter 9s description of obscenity, c& I know it when I see it, 3 d the definition of political corruption has escaped precision. This is not to say that scholars have avoided trying to define corruption. Prior work has viewed corruption from a number of perspectives, including that of illegality (Fackler and Linn, 1995) and the somewhat more expansive view of Peters and Welch (1978) who, while agreeing that clearly illegal activities are corrupt, suggest that public opinion may consider otherwise legal actions to also be corrupt. They go further in also suggesting that some activities which appear to be corrupt may in fact further the public interest. Johnston (1986) sees two categories of corruption, formal or social. Formal definitions rely on the idea that public officials abuse their roles and resources for their own benefit and are limited in their usefulness, especially for those who would reform corrupt systems. Instead, social definitions which are created by the perceptions of the public, while perhaps quite nuanced and difficult to measure, are the most interesting and useful to those who wish to understand how corruption is perceived and continues to exist. This latter view informs Redlawsk and McCann (2002) in which voters are given the opportunity to determine themselves what is and is not corrupt. While earlier work on corruption tended to focus on institutional perspectives (Heidenheimer, 1970; Meier and Holbrook, 1992) or to impose definitions of corrupt behavior, Redlawsk and McCann (2002) argue that if perceptions of corruption matter in how voters respond to the political environment, then it is necessary to develop a definition of corruption driven from the bottom up, rather than a definition imposed by the researcher. Other studies have shown that various groups of people often perceive corruption where no illegality exists and where public officials themselves may not see any clear wrongdoing (Johnson, 1986, 1998) 3 Jacobellis v. Ohio , 378 U.S. 184, 197 (1964) 6 This paper adopts the Redlawsk and McCann (2002) approach and uses some of the 2000 election data from that study, adding to it additional data from 2002. Results from the exit polls in 2000 show there is not a one-to-one mapping between plainly illegal behaviors and behaviors perceived by the public as corrupt. An eight question battery was administered on which voters could identify how corrupt they believed various actions by politicians and citizens to be on a 1-5 scale. Seven of the questions loaded cleanly onto two factors, with four of the questions clearly referencing illegal activities, while the other three all referenced activities that while not illegal, might be considered corrupt in their own way 4 (Redlawsk and McCann, Table 4, p. 23.) These two factors, lawbreaking and favoritism, had clear implications for the vote in 2000, with mobilization towards third party candidates not only predicated on a general sense of mistrust in government or lack of efficacy, but also on the moral judgments made about corruption. Ultimately, if the public perceives political figures and institutions to be corrupt, it may not matter whether they are corrupt in any legalistic definition. The mere perception of corruption is likely to be enough. If the public believes that all politicians are crooks, then it little matters whether they are or are not. Thus it is useful not only to understand how voters perceive corruption, but also how those perceptions may have changed over the course of the two years from 2000 to 2002, given the substantial potential impacts of the very public political and corporate corruption scandals occurring between the two elections. Study Design and Data 4 The eighth question, on raising campaign funds from a politician 9s office, loaded relatively evenly on both factors, probably reflecting the fact than many voters do not know that this is generally an illegal activity. 7 Data for this study were collected through exit polls carried out in Iowa City and Coralville, IA during the November 7, 2000 and November 5, 2002 general elections. 5 Exit polls cannot, of course, represent the general public, but they do have the advantage of capturing the attitudes of citizens directly engaged in the most basic of political processes. The polls were conducted under faculty direction by students enrolled in political science courses at the University of Iowa; the students received partial course credit for participating in questionnaire design and acting as interviewers. 6 Iowa City and its neighbor Coralville were chosen for convenience, not due to any assertion that they uniquely represent some particular area of the country. Iowa City is a college town of about 60,000 residents (including students.) Coralville, which shares a border with Iowa City, has a population of about 15,000, and is demographically somewhat older and more conservative than Iowa City, though both places (and indeed the entire county in which they are situated) are seen as liberal bastions within Iowa. Even so, the voters surveyed represented the full range of political attitudes, ideologies and party affiliations. For the 2000 poll, 13 voting precincts were randomly selected from the 29 precincts that cover the two cities. Each precinct was weighted according to its registered voter population so as to assure that every voter had an equal chance of being selected. In 2002, the same precincts were again surveyed, though because of redistricting that split precincts in Coralville, 14 precincts were included in the 2002 poll. Within precincts, interviewers systematically selected every third (in 2002) or fourth (in 2000) voter leaving the polling place and attempted to get her 5 The 2000 poll was part of a multi-city exit poll project that included surveys in Miami, New Orleans, New York City, Lafayette, IN, and Kenosha, WI in addition to the Iowa City area. For analyses of the complete 2000 dataset see Redlawsk and McCann (2002). The 2002 poll was only carried out in Iowa City and Coralville, IA. 6 See Cole (2002) for detailed discussion of the pedagogy involved in this project. 8 or him to complete the survey instrument. 7 Those who agreed were handed the questionnaire and asked to complete it unaided by the interviewer. Upon completion, the exit poll was dropped in a box next to the interviewer. While in 2000 the weather was cooperative (though rather cold) in 2002 election day was generally wet and miserable, especially for those interviewers forced to stand outside the polls for four hours. 8 Exit polls are not, of course, a random sample of all citizens in a given locality. Obviously, exit polls can only survey those who are actually voting at the polls on election day. As such, then, the results can only be generalized to voters in a given election rather than the public as a whole. Yet for the purposes of this study voters are exactly the population of interest, since the study specifically focuses on how voters respond to corruption. Of course, one way of responding to perceived governmental corruption is to opt out, to fail to vote. And no doubt, the attitudes of those who do so are important and warrant study. But for this paper, the focus is on those who, no matter how they perceive the government and people within it, still show up to vote. That having been said, it is important to validate the exit poll data in some fashion. As reported in Table 1 there is a very close link between the actual vote for president and other offices and the vote reported by the exit poll samples. In 2000 only the Presidential race is available for comparison, and the exit poll data tracks quite closely to the actual results at the polls on election day with the two major party candidates 9 vote slightly under-reported, and 7 As always, some voters refused to participate in the survey. Poll takers were instructed to record the gender, race, and apparent age of all refusals to allow weighting of the data. Refusal rates in 2000 were 44% and in 2002 were much higher at 60%, probably due to the weather. The analyses in this paper were carried out on the unweighted data, since preliminary investigation showed no particular advantage to using the weighted data. 8 As in most states, Iowa law requires interviewers doing exit polls to remain outside of the doors to the main entrance of a polling place. In some cases, especially in 2002, sympathetic election judges allowed the students to stand inside to avoid the weather. But often the students found themselves trying to convince voters to take the survey while standing under an umbrella. 9 Ralph Nader 9s vote a bit over-represented. In 2002, voters were asked about the Governor, Senate, and House races, and in the first two the exit poll results are quite close to the actual numbers. In the House race, however, the poll under-reports the vote for the incumbent Republican, Jim Leach, while over-reporting for his Democratic challenger. Given that about 20% of Democrats in the poll admitted voting for Leach, the assumption here is that some number of Democrats who did vote for Leach simply lied about it on the poll, since they otherwise voted a Democratic ticket. Overall, there is strong evidence that the exit poll interviews fairly represent election day voters in Iowa City and Coralville in both 2000 and 2002 9 . [Table 1 about here] It is also important to recognize contextual differences in the datasets for the two elections. The 2000 election was a high-interest presidential election which turned out to be extremely close in Iowa, with Al Gore winning the state by about 4,000 votes. Locally there was also a Congressional seat on the ballot (the incumbent was not seriously challenged and won handily) as well as local County offices. Voter turnout in Johnson County, IA was about 72.4%. 10 As one might expect, turnout in 2002 was significantly lower, at 52.4%, even though there were several high profile races, including competitive elections for Governor, US Senate, and this time around, the local Congressional seat. More important than the turnout difference is the difference in those who turned out. The 2002 election sample includes more high income, married, and home-owning voters than in 2000. On the other hand, there are far fewer minorities in the 2002 sample (and fewer minorities than in the general public in both years.) The 2002 9 The exit poll results here are compared to the vote on election day, not including absentee or cearly d vote, which was substantial. The Johnson County auditor reports early vote separately from vote at the polls. In 2000, about 39% of votes were cast early; in 2002 the percentage was 38%. 10 Turnout figures are from the Johnson County, IA web site at http://www.jcauditor.com. 10 sample is also more Republican and conservative, while slightly less female. Where nearly 43% of the 2000 sample was under 30 years old, only 31.7% of the 2002 sample is in that age group. Finally, the 2002 sample is more highly educated, undoubtedly connected to the lower turnout of younger voters. Appendix A details key differences which must be taken into account in the analyses that follow. The exit poll questionnaire contained about 50 questions, include a demographic battery, questions about the votes just cast, and a series of questions designed to tap attitudes towards political corruption. This latter included the eight question battery designed to tap attitudes towards a range of activities that might be considered to be corrupt. This battery included items that are clearly illegal as well as items that are less clear cut. Previous analyses of these items from the 2000 multi-city poll show the emergence of two clear factors, lawbreaking and favoritism (Redlawsk and McCann, 2002.) In addition, several specific questions about perceived corruption in government were asked. Most of these questions, including the corruption battery, were asked in both 2000 and 2002, allowing a direct comparison to be made between the two years. Unfortunately, however, the several questions on attitudes towards corporations were not asked in 2000, meaning that while the effects of the corporate corruption scandals noted earlier cannot be tested over time, they can be tested at one point in time and compared to attitudes towards government at that time. The full questionnaires with marginals (eliminating those with no answer from each question) are included as Appendix B. Results How Much Corruption? 11 Several questions were designed to tap general attitudes towards government (local and national) and corporations (only in 2002). As a first cut, the univariate results of those questions are presented in Table 2. Keeping in mind the variation in the samples, the most notable finding is that there are few differences in the responses between the 2000 and 2002 samples. Voters in the Iowa City area did not change their basic opinions about local government corruption, about any increase in corruption in the federal government (just under half feel it has increased cin the past 20 years d) or about identifying whether the national government is more corrupt than either the state or local governments. Basic efficacy questions 3 trust in the federal government and whether public officials care 3 show virtually no change as well. Two indicators of how voters feel about the federal government do show significant change, however. Respondents were asked whether cIn government corrupt means are needed to achieve important goals d and given the choice of cusually d or crarely. d In 2000, nearly one-third of voters answered cusually d but the number dropped considerably, to just under one-quarter in 2002 a significant and substantive change (X 2 1 =9.218, p <.01.) Likewise, fewer voters thought that cquite a few d of the people crunning the government d were ccrooked d in 2002, compared to 2000 (42.6% vs. 50.6%, X 2 2 =14.243, p <.001.) These two measures appear to indicate a more positive perception of government in 2002. This can be contrasted with the views of voters on corporations. Since these questions were not asked in 2000, no changes over time can be considered, but in 2002, voters are much less supportive of corporations than government. While only a quarter thought corrupt means were needed in government, nearly 40% felt corrupt means were needed to achieve goals in corporations. Further, more than half (52.6%) believed that cquite a few d of those crunning corporations d were crooked. These differences are statistically significant (corrupt means: X 2 1 =121.608 p <.001; number crooked: X 2 4 =233.187, p <.001.) 12 Finally, voters in 2002 were much more trusting of either government (46.4%) or unions (35.7%) than corporations (17.9%) when asked which of the three they trusted most. [Insert Table 2 about Here] The eight-item corruption battery also provides some interesting results as shown in Table 3. It is clear looking at the battery that voters perceive the obvious differences between the types of activities listed. Those items which are illegal are all condemned at a high level, with means of well over four on a 1-5 scale, where five is cextremely corrupt d. The only exception is for fundraising in a government office, which people apparently do not see as necessarily illegal, though it generally is. Given the prominence that this issue had during both the 1996 and 2000 presidential elections, when Vice-president Al Gore was accused of such activities, it is somewhat surprising that voters are not as intolerant of it as other illegal activities. The remaining items generally are not illegal, and may, in fact, be seen by many as cpolitics as usual. d Two of the three are not viewed as particularly corrupt 3 supporting a candidate promising to fix potholes, and recommending an out of work friend for a job 3 though in both cases significant minorities of voters still consider these activities to be very corrupt. However, interestingly, voters much more strongly condemn tax cuts for rich supporters of an elected official. The question simply asked how corrupt it would be if can elected official with many wealthy backers supported a tax cut that largely benefited the rich. d This would seem to many political observers to be the essence of interest group pluralism 3 interests work to get politicians elected who will support what they want! Voters in the Iowa City area, though, appear to be very wary of such a process. While factor analyses (not shown, see Redlawsk and McCann, 2002 for details) indicate that in both 2000 and 2002 the items separate into the same two clear factors (with one item, 13 fundraising in the office not loading clearly on either), the individual items show some real change over the two years. Turning first to the claw breaking d items, voters became more condemning of these actions, showing significant differences in three of the four items on this scale and on the overall index (created as the mean of the four items; 4.38 in 2002 vs. 4.29 in 2000, p <.01.) In general, the voters in 2002 were harder on these clearly illegal activities. On the other hand, the items making up the favoritism scale show the opposite result. Voters in 2002 are easier on favoritism, with the strongest difference coming on the item about recommending an out-of-work friend for a job. The overall favoritism scale is significantly lower in 2002 than in 2000 (2.88 vs. 3.00, p <.05) indicating a higher tolerance for this set of activities. Most of the difference is attributable to the out-of-work friend item; perhaps the perceived decline in the economy over the two years makes voters somewhat more sympathetic to anyone out of work. [Insert Table 3 about here] Individual Differences in Attitudes towards Corruption What underlies the differences between 2000 and 2002 displayed in the exit poll data? One obvious source of difference is the variations in the samples, which display significant socioeconomic and political differences, as would be expected for different kinds of elections. Because socioeconomic factors and partisanship predict attitudes towards corruption (Redlawsk and McCann, 2002) it is important to control for these differences. Table 4 examines some socioeconomic and political differences in attitudes towards the lawbreaking and favoritism indices, along with other measures of corruption pooling the 2000 and 2002 data. No clear differences arise by partisanship in attitudes towards lawbreaking, but Democrats (2.94 on a five point scale) and Independents (2.87) are somewhat less tolerant on the favoritism index than are 14 Republicans (2.76), though the difference is not large. Women are generally less tolerant on both indices, with a fairly large difference on favoritism (3.02 vs. 2.72 for men.) College graduates are more condemning of illegal activities than others (4.46 vs. 4.23 for some college and 4.21 for no college), but those with at least some college (2.84) are more tolerant of the cpolitics as usual d implicit in the favoritism index than are those with no college at all (3.05). Wealthier voters are more condemning of illegality (4.53 vs. 4.24), but there is no difference between them and those less well off on favoritism. And as voters get older, they are clearly less tolerant of either illegal behavior or favoritism, with a rather monotonic increase in the scale by age group. Looking at the questions directed towards government, political independents are more likely to choose the answers that indicate belief that government is crooked on all of these questions than are partisans of either stripe, who show little difference between them. Political independents are clearly simply less positive about government. Women are also generally less positive, as are those with less education. Interestingly, in this Midwestern town (which most would agree is not particularly corrupt) those without any college education are more than twice as likely to indicate that there is more corruption locally than in other places they have lived (23.9% vs. 10.9% for some college and 8.1% for college graduates.) Older voters are generally less likely to be negative towards either the local or federal government, while the youngest voters are especially more likely than all other age groups to see a need for corrupt means in government to reach important goals (34.2% vs. 17.8% to 22.4% for other age groups) as are those with lower incomes (who are probably often the same voters). Finally, Table 4 shows the 2002 results for questions about corporations. Here there are some strong partisan differences, as might be expected, with Republicans less negative about corporations than either Democrats or independents. And, not surprisingly, Republicans are 15 much more likely to trust corporations (37.0% trust them more) than are Democrats (7.9%) and independents (15.8%.) On the flip side, Democrats and independents are much more trusting of unions (46.6% and 46.7%) than are Republicans (7.8%). Oddly, given typical Republican anti- government rhetoric, more Republicans trust government (55.2%) than either of the other two, while for Democrats about the same share chooses government (45.5%) as unions while independents give unions the edge, 46.7% to 37.5%. Again women are less positive about corporations as they are with government. On the other hand, the age and education differences are mixed here, though younger voters are more likely to trust unions compared to corporations, while for the oldest group government is by far the most trusted of the three. Finally, lower income voters are less positive about corporations, just as they were about government. [Insert Table 4 about here] If nothing else, Table 4 clearly demonstrates that it is important to take the demographic and political differences of the two samples into account when trying to determine what changed, if anything, between the two elections. The remaining analyses in this paper all control for these differences. Changes in Attitudes between 2000 and 2002 One interesting place to start is to look at the exit poll question giving voters the opportunity to check which one ccandidate quality d matters most. In both years a list of nine options was provided, including one about trusting the candidate. A quick examination shows that across the full sample, 14.3% of those who chose an option marked ctrust d. Given nine options, this is not much more than the 11.1% that would be expected by chance. The difference between 2000 (15.1%) and 2002 (13.0%) is not statistically significant, and trends in the wrong 16 direction if trust, per se , mattered more in 2002 than in 2000. So while many things happened in the political environment between the two years, nothing appears to have made trusting the candidate more important over that time. 11 However, even without a change in the salience of candidate trust it is certainly possible that attitudes towards corruption and government changed over this period. Table 5 reports significant results from a multivariate model simultaneously examining changes in attitudes towards both of the corruption indices, lawbreaking and favoritism. The model includes main effects for the various socioeconomic and political characteristics of the samples (partisanship, ideology, gender, income, age, race, education) and the year (coded 1 for 2002), as well as an interaction term between each characteristic and the year, to examine changes within particular groups over time. The model also controls for efficacy, measured as a response to a question about whether government cares about cpeople like me d and for trust in Washington, DC. 12 While there are a number of main effects, the analysis here is focused on the changes over time. 13 Making sense of the results of this analysis (and ones to come) may be a bit tricky so while the parameter estimates for the full model are presented in Table 5, graphical analyses 11 It is worth noting that there are strong partisan differences here. Democrats show no significant change over the two elections (9.6% in 2000, 11.6% in 2002) while Republicans were substantially more likely to choose ctrust d during the presidential election (24.5% vs. 13.6% in 2002) no doubt in response to a major Bush campaign theme (X 2 1 =6.977, p <.01.). 12 It is, of course, possible that efficacy and trust have changed over time. OLS regression analyses pooling the 2000 and 2002 data with efficacy or trust as the dependent variable and the various individual differences as controls show no significant difference from 2000 to 2002 and no interaction effects between the year and individual differences. Given this, use of the efficacy and trust measures as controls in the remaining analyses will be as a main effect only. Efficacy is jus the opposite, showing no effects on lawbreaking, but significant effects on favoritism; as respondents become more efficacious, they become less condemning of favoritism. 13 While the individual demographic correlates of lawbreaking and favoritism are certainly interesting, they are not the focus of this paper and are not addressed here. Some highlights are worth pointing out, however. Women are substantially more likely to condemn favoritism. No differences are seen in these data for partisanship, but liberals are less condemning of lawbreaking than either moderates or conservatives 3 though there is no significant difference for favoritism. In general there is a monotonic increase in condemnation of both illegal activities and favoritism by age, with the youngest cohort less likely to condemn either type of activity than their elders. Finally, those voters who have never attended college are less condemning of illegality than college graduates. More detailed analysis of the socioeconomic bases of corruption attitudes in the 2000 election is available in Redlawsk and McCann, 2002. 17 of the significant effects are shown in Figures 1 and 2. Turning first to the lawbreaking index, three groups show significant changes from 2000 to 2002. Liberals appear to become more accepting of illegal actions compared to both conservatives and moderates, as do non-whites compared to whites. Respondents with some college education become more condemning of these activities, compared to college graduates. Figure 1 clarifies these results. Controlling for all other factors, and assuming the mean values for efficacy (3.43) and trust (2.07), all ideological groups show some decline in how harshly they rate illegal activities, though liberals clearly become more accepting than any other group. Turning to race, there is virtually no change in how whites view lawbreaking, but non-whites are substantially less likely to condemn such activities in 2002 compared to 2000. And, again for education, all groups show some decline, but both college graduates and those with no college at all become much more accepting, while those with some college but no degree change very little. It is worth pointing out, however, that respondents in general still remaining quite convinced that these activities are very corrupt, since the means remain about 4.0 on a 1-5 scale. There are also few differences among groups over time on the favoritism index. Again there seems to be somewhat of an overall decrease in how corrupt voters see these activities, though the coefficient on the year does not reach statistical significance. Only two groups of voters show any significant change. As Figure 2 reflects, both Democrats and Independents were less condemning in 2002 than in 2000, while Republicans remained in virtually the same place. On the other hand, older voters rated favoritism as substantially more corrupt in 2002, while other age groups rated these activities as less corrupt. In summary, despite the apparent increase in how corrupt voters viewed lawbreaking seen on Table 3, after controlling for the sample differences voters are slightly more accepting of 18 these activities, if anything. The changes seem to be limited to a few groups, however. The various events that occurred during the two years had either little effect, or cross-cutting effects, on this measure. Some specific groups do show change, but not in any fashion that indicates clear direction, and in fact, the results are counter to what would be expected if voters became more wary of politicians as a result of the corruption scandals. The same is mostly true of favoritism though there is some evidence that there is an overall increase in tolerance for these activities. While Republicans and older respondents are relatively more condemning of favoritism in 2002, the overall means otherwise do show some declines. Voters in 2002 may have been more tolerant overall of cpolitics as usual d as captured in the kinds of activities that make up the favoritism measure. This may well reflect the predominance of security and terrorism concerns as election 2002 arrived which could have resulted in a generally more positive view of what politicians do (that is not clearly illegal) than otherwise might have been the case without September 11. [Insert Table 5 and Figures 1 and 2 about here] In addition to the eight question battery on corruption, voters had the opportunity to respond directly to questions about the prevalence of corruption in government with two questions aimed at tapping how they felt government was run. One asked chow many people running the government do you think are crooked d while the other asked how often corrupt means are needed in government to cachieve important goals. d These two measures are reasonably correlated (r=.246, p <.001) while tapping different aspects of how government is viewed. A government corruption index was created averaging the two measures, scaled so that a high value on the index indicates a strong belief that government is corrupt. As with the lawbreaking and favoritism indices, a model controlling for socioeconomic, partisan, and 19 ideological differences was built to estimate the effects of time on this index. The model is shown in Table 6, while Figure 3 displays graphically the significant differences between groups. As with the previous analysis, little overall change is seen over the two years. However, once again several groups show significant change, after controlling for efficacy and trust in Washington. In particular, Republicans and Democrats both become relatively less likely to believe government is run by crooks and requires corrupt means, than independents, who while at roughly the same place with partisans in 2000, are the group more likely to believe government is corrupt in 2002. On the other hand, age differences which appear in 2000, when younger voters were substantially higher on the scale than others, disappear in 2002, as younger voters become less convinced that government is crooked, while other age groups show little change. The other change is among the least educated voters, those who have not attended college. This group was well above other education levels in its belief that government was corrupt in 2000, but by 2002 those with no college look no different than those who attended or graduated from college. As with attitudes towards illegal activities and favoritism, beliefs about how corrupt government is appear to have moderated a bit between 2000 and 2002. The changes here are quite clear for several groups 3 ratings on this scale are lower if they have changed at all. In particular, though, ideology and partisanship show some interesting patterns. Ideologues (liberals and conservatives) show a greater increase in tolerance (shown by a decline on the scale) for lawbreaking than do moderates, while partisans (Republican and Democrat) show a greater decline in their assessment of how corrupt government is in 2002 than do Independents, who show relatively little change. [Insert Table 6 and Figure 3 about Here] 20 The overall picture seems to be that the corruption scandals played little role in the evolution of voter attitudes towards corruption and government between 2000 and 2002. Had the scandals really influenced voters, it is likely that attitudes towards corruption would have become more severe, not generally less, as appears to be the case for this group of voters. Attitudes towards Government and Corporate Corruption in 2002 Unfortunately no questions about corporate corruption were asked on the 2000 exit poll so there is no opportunity to examine change over time. However, it may be useful to directly compare attitudes towards government with those towards corporations in 2002. There are two opportunities to do this in these data. First, a corporate corruption index, analogous to the government corruption index above can be constructed using two questions about the number of crooks running corporations and the need to use corrupt means to obtain important ends in corporations. These measures are fairly highly correlated (r=.457, p <.001) but each brings a different perspective on the question. The corporate corruption index is highly correlated with the government corruption index for 2002, (r=.538, p <.001) indicating that voters who see crooks in one place are relatively likely to see crooks in the other. Yet, having said this, there may well be differences in who views which entity as corrupt. In order to make a direct comparison between government and corporations, a multivariate GLM model was constructed with the two indices as the dependent variables, and the various socioeconomic and political factors as predictors. The results are presented in Table 7. As it turns out there are some interesting differences in attitudes towards government and corporations. Democrats, as would be expected, are more positive towards government than they are towards corporations, while conservatives (controlling for party ID of course) are much more 21 positive towards corporations (both represented by negative coefficients, since the dependent variables are coded so that a high score indicates belief that crooks run government.) Neither of these results is particularly surprising. Turning to socioeconomic traits, it is clear that women, in general, are more likely than men to believe that crooks run both corporations and government . On the other hand, those without any college education are no more likely to see government as corrupt, but they are significantly more likely to view corporations as corrupt. The same effect is seen with age, as younger voters (18-29) are much more negative about corporations but no more so about government than other age groups. On the other hand, voters over 65 are substantially less likely to see corruption in corporations than other age groups. [Insert Table 7 about Here] Clearly, despite the high correlation between the two, beliefs in the number of crooks and need to use corrupt means in government are not the same as beliefs about the same things in corporations. Different groups of voters hold different attitudes towards the two. In order to examine the effects of these beliefs and of attitudes towards corruption on trust in government and corporations, voters participating in the 2002 exit poll were given the chance to choose which of three entities 3 government, corporations, or unions 3 they most trusted. Only one of the three could be chosen, forcing a decision. It seems likely that in addition to the socioeconomic and political factors that have been used in previous analyses, attitudes towards governmental and corporate corruption, along with attitudes toward corrupt activities (lawbreaking and favoritism) should predict which of these three voters trust most. It seems fairly certain that those who see a great deal of corruption in government would be unlikely to pick it as the most trusted, while those who see corruption in corporations would be less likely to 22 trust them. Likewise, attitudes towards the corrupt activities, since they mostly reference political officials, may predict the likelihood of trusting government over the others. Table 8 presents the results of a multinomial logistic regression analysis with ccorporations d as the reference group. The analysis predicts the likelihood of choosing either government or unions as the most trusted, compared to choosing corporations. As expected, a number of socioeconomic and political variables predict differential preferences. Republicans are substantially less likely than independents or democrats to choose unions as the most trusted and somewhat less likely to choose government. Liberals, on the other hand are much more likely to choose unions, compared to moderates, but ideology has no effect on choosing government as the most trusted. Those with no college tend to trust unions over the other two, while older voters are far more likely to trust government than other age cohorts. As would be expected, a high sense of efficacy or trust in Washington leads to choosing government as the most trusted compared to either corporations or unions. But the real story of interest is in the measures of attitudes towards corruption. Neither the lawbreaking nor favoritism index predicts any differential preference. It may be simply that because the references in the questions comprising the index are focused on specific actions, rather than more general references to government or corporations, that expecting to find any effects on a trust preference is wrong. On the other hand, the government and corporate corruption indices, which combine beliefs about the number of crooks and the need for corrupt means in government or corporations, are significant predictors in the model. It appears that generalized beliefs about these entities, rather than about specific actions taken by citizens and politicians, lead to trust in the institutions. A belief that government is particularly corrupt significantly lowers the probability that government will be chosen as the most trusted, while 23 having no effect on choosing unions. But a belief that corporations are especially corrupt leads to a significantly higher likelihood that either government or unions will be seen as most trusted. These latter two are then differentiated by the government index; those believing both government and corporations are corrupt are more likely to choose unions as the most trusted of the three. [Insert Table 8 about here] The overall story of Table 8 indicates that for trust in institutions it is the general beliefs about the institutions that count, not attitudes towards how corrupt certain activities are or are not. Essentially, beliefs about how government and corporations work translate directly into beliefs about how trustworthy they are. Political Consequences of Voter Attitudes Finally, it is worth considering whether the attitudes towards corruption translate in any way into the vote choice itself. It is certainly possible that given that all respondents are voters, corruption attitudes have little to do with the vote choice, assuming there are no obvious scandals involving corruption and candidates actually on the ballot. After all, perhaps those for whom corruption really matters are not even voting, and thus not captured in these exit polls. Previous work, however, provides reason to believe that corruption attitudes do influence the vote with the Redlawsk and McCann (2002) finding that voters more condemning of favoritism were more likely to cast a third party vote, at least in the 2000 presidential election. In 2002, though, there was no presidential election; in Iowa the US Senate race between incumbent Democrat Tom Harkin and Republican Greg Ganske was at the top of the ticket. A multinomial logistic regression model can attempt to differentiate between the probability of a vote for one of these 24 major party candidates or a vote for the Green party alternative. 14 Such a model is reported in Table 9 with a vote for the Republican candidate as the reference category. The analysis finds no significant effects for beliefs about corruption in either government or corporations, though the coefficients for a third party vote indicate some tendency for those believing these entities are corrupt vote third party. The only statistically significant effect is found in the lawbreaking index and third party voting. In a reversal of Redlawsk and McCann (2002) the favoritism index has no ability to differentiate between candidates, but the lawbreaking index has very strong positive effects. Voters who feel more strongly that the activities in this index are in fact corrupt are much more likely to vote third party. It is not at all clear why the findings in this study should reverse the 2000 findings, though that study included a wide range of cities and a much larger number of third party votes. Perhaps this is simply idiosyncratic of the Iowa City area, or perhaps it is related to the Senate race which pitted an incumbent against a challenger, versus a presidential race with no actual incumbent. In any case there are no data here to tease out this difference. But even so, it is clear that attitudes towards corrupt activities have potential to affect the vote choice. To make a bit clearer the effects of these attitudes towards lawbreaking, the probabilities of a third party vote were calculated at each of its levels, keeping all other predictors at their means or reference categories (see Figure 4.) Voters at the lowest (most tolerant of lawbreaking) level of the scale (1) are less than 1% likely to vote third party, while those in the middle (3) are just about 2% likely, while those at the most condemning end of the scale (5) have more than a 12% chance of voting third party, substantially greater than the share that actually cast a third party vote (6.1%.) 14 Given the relatively small number of third party voters in this race (6.1%, or 34 voters cast third party votes) the following analysis is tentative at best. Still the results should be suggestive. 25 [Insert Table 9 and Figure 4 about here] Discussion Exit polls are not useful for studying the general population, of course, but they do provide an understanding of how voters feel at about the time they cast their vote. Thus, for anyone interested in how political attitudes affect the only political activity in which most citizens engage, exit polls can be quite useful, as long as their limitations are understood. And, of course, an exit poll in only one community, as presented here, does not allow any strong conclusion to be drawn about voters elsewhere. Even so, the findings in these exit polls, providing as they do the opportunity to examine change over time on an important issue of the day, are both useful and suggestive. First, even though corporate corruption was all over the front pages in the days before the terrorist attacks of 2001 and even somewhat afterwards, by the time the 2002 elections rolled around, there was little evidence that voters became any more concerned about corruption in government. If anything, there is some evidence that attitudes towards corrupt activities actually became slightly less negative, as voters appeared to moderate their views on how corrupt government really is. This moderation may have come about because much of the talk about corruption was muted by election 2002, replaced by talk about security, patriotism, and the role of government in protecting American citizens. Second, there are differences in beliefs about how corrupt corporations and government actually are. When asked about the need to use corrupt means and about how many working in government and corporations are corrupt, voters of different political and socioeconomic groups had different responses. Democrats like government more, as would be expected, while women 26 in general were less trusting of both institutions. And while age has no effect on beliefs about governmental corruption, age and education were both significant predictors of belief that corporations are corrupt, with young people and those without college degrees were more likely to see corruption in corporations, while elderly voters were less. These differences make clear that not all voters see corruption the same way, nor do they respond to the information environment in the same ways. Meanwhile the beliefs in the levels of corporate or governmental corruption have direct effects on whether these institutions are trusted. Third, as Redlawsk and McCann (2002) show in the 2000 presidential election across six cities of varying sizes and cultures, attitudes towards corrupt activities have potential to influence the vote choice. While these attitudes do not differentiate between Republican and Democratic votes, they do influence the likelihood of a third party vote. Third party voters are likely disillusioned with the political system as it is; one wellspring of that disillusionment may be a sense that government is corrupt and a tendency to condemn the kinds of activities they believe politicians routinely engage in. To be fair, the findings in the US Senate election in Iowa City in 2002 are somewhat different from those found across six cities in the presidential election of 2000. Where in 2002 it was attitudes towards illegal activities that predicted greater third party voting, in 2000 it was voters who condemned favoritism who were more likely to vote third party. Except to suggest that there is something idiosyncratic about either Iowa City (reasonably likely) or that Senate elections with an incumbent tap different considerations than do copen seat d presidential elections (also likely) there is no clear explanation for this difference. Even so, this study, limited as it is to voters in only one small geographic area, provides some sense of both the nature of attitudes towards corruption and beliefs about government and corporations, and how both may influence politics more generally. 27 Table 1 Exit Poll Vote Compared to Actual Vote Reported Vote Actual Vote Variance 2000 President Bush (R) 31.0 31.9 -0.9 Gore (D) 59.1 59.2 -0.1 Nader (G) 9.6 7.4 2.2 2002 Governor Gross (R) 32.9 32.8 -0.1 Vilsack (D) 61.0 60.1 -0.9 Others 6.1 7.1 1.0 US Senate Ganske (R) 32.9 32.1 -0.8 Harkin (D) 63.2 62.7 -0.5 Others 3.9 5.2 1.3 Congress Leach (R) 52.9 49.3 -3.6 Thomas (D) 44.9 49.3 4.4 Others 2.2 1.4 0.8 Actual votes are from the Johnson County, IA Auditor 9s Office, http://www.jcauditor.com. Only votes cast on election day itself are included; absentee votes are removed since absentees could not have been interviewed on election day. Totals include only the vote in Iowa City and Coralville and do not include the rest of Johnson County. 28 Table 2 Marginals for Corruption Questions More Corruption More Corruption Trust Govt Public Officials Locally in Fed Govt Most Corrupt Level In Washington Don 9t Care % Yes % Yes Local State National % Only Sometimes % Agree 2000 12.2 45.9 12.0 7.2 80.8 50.2 40.4 2002 10.0 48.0 8.0 8.9 83.1 48.4 39.0 In Government&&&.. In Corporations&&& Corrupt Means Number of Corrupt Means Number of Which do you Are Required Crooks Are Required Crooks Trust More % Usually % Quite a few % Usually % Quite a Few Govt Corp Unions 2000 32.2 49.8 n/a n/a n/a n/a n/a 2002 24.7 42.6 39.8 52.5 46.4 17.9 35.7 Table entries are percentages of those answering the question as indicated. 29 Table 3 8-Item Corruption Battery 2000 Poll 2002 Poll Difference Mean % Extremely Mean % Extremely (t-test) Would it be corrupt if: (sd) Corrupt (sd) Corrupt The following Items load on a cLaw Breaking d Factor 4.29 4.38 t=2.623, p<.01 (.70) (.62) ------------------------------------------------------------------------------------------------------------------ ---------------------- Someone on the government payroll did no 4.39 59.8 4.40 60.7 t=.210, n.s. work for the pay? (.89) (.90) A police officer accepted money not to write 4.33 59.5 4.48 65.1 t=2.965, p <.01 a traffic ticket on a speeding driver? (.95) (.84) A citizen claimed government benefits to which he or she 4.29 53.1 4.38 56.1 t=2.115, p <.05 was not entitled? (0.88) (.81) A government official gave a contract to a campaign 4.12 49.8 4.26 52.8 t=2.328, p <.05 contributor without considering other contractors? (1.06) (.91) The following items load on a cFavoritism d Factor 3.00 2.88 t=-2.291, p<.05 (1.04) (1.00) ------------------------------------------------------------------------------------------------------------------- ---------------------- An elected official with many wealthy backers supported 3.66 38.1 3.57 35.8 t=-1.412, n.s. tax cut that largely benefited the rich? (1.35) (1.39) Voters supported a candidate for office in return for a 2.69 14.3 2.57 14.2 t=1.285, n.s. Promise to fix potholes on their street? (1.39) (1.38) An official recommended an out-of-work friend for a 2.66 11.5 2.45 7.8 t=-2.839, p <.01 government job? (1.27) (1.20) The following item does not load cleanly on either factor An elected official raised campaign funds while inside his 3.51 24.0 3.44 23.5 t=-.932, n.s. or her government office? (1.16) (1.20) Note: Factor loadings were derived via principal component analysis, with VARIMAX rotation. All items scored on a five-point s cale (1 = cnot at all corrupt, d 5 = cextremely corrupt d). The two factors explain 53.4% (2000) and 52.9% (2002) of the variance; a third factor would have an ei genvalue well below 1.0. 30 Table 4 Socioeconomic and Political Differences in Attitudes towards Corruption 2000 and 2002 Pooled Education Partisanship Gender No Some College Age Income >$50K Rep. Dem. Ind. Male Female College College Grad <30 30-44 45-64 >64 Yes No Law-Breaking 4.32 4.34 4.31 4.30 4.36 4.19 4.15 4.43 4.03 4.45 4.58 4.71 4.47 4.22 (.67) (.66) (.68) (.67) (.67) (.84) (.74) (.58) (.72) (.58) (.52) (.41) (.71) (.71) Favoritism 2.77 3.02 2.98 2.76 3.12 3.16 2.95 2.92 2.89 2.93 3.02 3.25 2.91 2.98 (1.11) (.97) (.99) (1.02) (.99) (1.06) (.98) (1.04) (.95) (1.05) (1.08) (1.07) (1.02) (1.02) More Local 12.6% 10.1% 12.1% 11.6% 11.2% 17.8% 15.4% 8.7% 15.4% 8.9% 9.4% 5.6% 7.9% 13.6% Corruption (% Yes) More Corruption in 29.1% 39.2% 31.7% 42.9% 50.8% 66.3% 49.7% 43.4% 50.3% 42.1% 47.6% 47.8% 43.1% 49.4% Fed. Govt. (% Yes) Govt Needs Corrupt 23.9% 29.0% 33.4% 26.8% 31.0% 39.6% 35.5% 24.6% 38.4% 23.7% 21.3% 24.6% 22.1% 34.3% Means (% Usually) # of Crooks in 43.0% 44.7% 52.9% 42.1% 51.2% 53.3% 50.8% 44.0% 54.9% 43.2% 40.1% 37.5% 41.4% 50.7% Govt (% quite a few) Only asked in 2002: Corps Need Corrupt 32.9% 40.8% 44.4% 38.8% 41.3% 32.6% 52.2% 36.1% 53.1% 34.5% 35.2% 23.7% 33.9% 45.6% Means (% Usually) # of Crooks in Corps 37.1% 57.9% 59.3% 45.1% 59.7% 56.5% 58.8% 49.5% 64.9% 49.4% 44.9% 43.6% 47.3% 57.6% (% quite a few) Trust Government 55.2% 45.5% 37.5% 47.3% 44.6% 52.3% 43.6% 46.9% 44.0% 44.8% 45.7% 71.4% 52.0% 41.1% More: Corporations 37.0% 7.9% 15.8% 22.3% 13.4% 11.4% 16.8% 19.1% 12.6% 17.8% 23.9% 14.3% 21.0% 15.0% Unions 7.8% 46.6% 46.7% 30.4% 42.0% 36.4% 39.6% 34.0% 43.4% 37.4% 30.4% 14.3% 27.0% 43.9% 31 Table 5 Changes in Attitudes towards Types of Corruption, 2000-2002 Lawbreaking Favoritism Year (1=2002) -.330 (.206) -.496 (.334) Year x Partisanship Democrat .128 (.084) .050 (.136) Republican .042 (.108) .326* (.175) Ideology Liberal -.209*** (.081) .076 (.132) Conservative -.096 (.109) -.179 (.177) Socioeconomic Traits Female .014 (.069) -.145 (.111) Income > $50K .054 (.074) .080 (.120) Non-White -.383** (.188) -.210 (.304) Education No College -.079 (.136) -.202 (.221) Some College .171** (.081) .017 (.132) Age 18-29 -.060 (.094) .052 (.153) 30-44 .036 (.091) .047 (.147) 65 and Over .018 (.168) .569** (.270) Partisanship Democrat .074 (.065) -.009 (.106) Republican -.065 (.082) .084 (.133) Ideology Liberal -.163*** (.063) .056 (.102) Conservative .079 (.082) -.177 (.134) Socioeconomic Traits Female .067 (.052) .262*** (.084) Income > $50K .125** (.056) .004 (.091) Non-White -.333** (.162) .071 (.263) Education No College -.282*** (.101) .053 (.163) Some College .017 (.064) .059 (.104) Age 18-29 -.478*** (.071) -.165 (.114) 30-44 -.081 (.065) -.071 (.106) 65 and Over .194* (.112) .410** (.182) Sense of Efficacy .012 (.009) -.052*** (.015) Trust in Washington -.057*** (.016) -.039 (.026) Constant 4.444*** (.175) 3.353*** (.284) Adjusted r 2 .160 .056 * p <.1 ** p <.05 *** p <.01 Parameters are estimated by multivariate General Linear Modeling with both Lawbreaking and Favoritism as dependent variables, standard errors in parentheses. n=1373. Reference category for Party is Independent, for Ideology is Moderate, for Education is College Graduate, for Age is 45-64 Years Old. 32 Table 6 Change in Beliefs about Corruption in Government 2000-2002 Year (1=2002) .349 (.370) Year x Partisanship Democrat -.246* (.153) Republican -.248 (.200) Ideology Liberal -.125 (.148) Conservative -.247 (.199) Socioeconomic Traits Female .006 (.125) Income > $50K -.109 (.135) Non-White .101 (.335) Education No College -.409* (.248) Some College -.024 (.148) Age 18-29 Years Old -.396** (.172)

less