1 Understanding Trust in Government Joseph Gershtenson, Eastern Kentucky University Dennis L. Plane, Juniata College Proposed Questions for the 2008 ANES Time Series Study 1. Standard NES Trust Question: How much of the time do you think you can trust the government in Washington to do what is right just about always, most of the time, or only some of the time? [Also record the volunteered Never response.] Point Trust Question: On a scale from 0 to 100, what percent of the time do you think you can trust the federal government in Washington? 3a. Open-ended Trust Question: When I asked about the how much of the time you trust the government in Washington, what sorts of things did you think about that made you say [ ]? Please be as specific as possible. 3b. Open-ended Trust Question: When I asked about the percentage of time you trust the federal government in Washington, what sorts of things did you think about that made you say [ %]? Please be as specific as possible. 4. State-level Trust Question, Standard NES Wording: How much of the time do you think you can trust the state government in [state capital] to do what is right just about always, most of the time, or only some of the time? [Also record the volunteered Never response.] 5. State-level 101-Point Trust Question: On a scale from 0 to 100, what percent of the time do you think you can trust the state government in [state capital]? 6a. Open-ended State-level Trust Question: When I asked about how much of the time you trust the state government in [state capital], what sorts of things did you think about that made you say [ ]? Please be as specific as possible. 6b. Open-ended State-level Trust Question: When I asked about the percentage of time you trust the state government in [state capital], what sorts of things did you think about that made you say [ %]? Please be as specific as possible. 7. During the PAST 12 MONTHS, have you worked with other people to deal with some issue facing your community? (NES Question 820) 8. If you were selected to serve on a jury, would you be happy to do it or would you rather not serve? (NES Question 831) Sample Design: We propose dividing the sample into four sub-groups for the purposes of these questions. All respondents would be asked the standard NES question on trust in the government
2 in Washington (#1 above). In addition, all respondents would be asked questions 2, 4, 5, 7, and 8. The sample would be split for the purposes of the open-ended items with half of the sample being asked open-ended items (3a, 6a) querying about responses to the standard NES format questions on both the national and state governments (1, 4) and the other half of the sample being asked open-ended items (3b, 6b) querying about responses to the 101-point items (2, 5). For each of these groups, we further propose that for half of the respondents the standard NES trust question should be administered close to the front of the survey, while the percent-of-the-time measures should be administered towards the end of the survey. For the other half of respondents, the administration of these questions should be reversed, with the percent-of-thetime measures at the front of the survey, and the standard NES trust question toward the end. Keywords: Political Trust, Citizenship, State Government, National Government Abstract We propose supplementing the current NES trust measure with a 101-point trust measure that is theoretically superior and has worked well on previous surveys (including the 2006 NES Pilot). Because citizens have different evaluations of different governments, we also propose including state-level NES trust measures. Since political trust means different things to different people, we propose adding open-ended prompts that will help determine what influences citizens trust reports. We have previously employed these open-ended prompts successfully in evaluating Kentucky citizens trust in the national and state governments. Finally, we propose questions to explore the link between trust and citizenship. Most trust research is based on the (often implicit) assumption that greater political trust is desirable and is the mark of a good citizen. Yet this assumption has not been examined empirically. Thus, we propose two previously employed questions related to citizenship (i.e., work on a community issue and
3 willingness to serve jury duty). When combined with other questions in the NES core (e.g., volunteerism, interest in politics, and knowledge) we can directly assess the relationship between trust and citizenship. The Merit and Theoretical Foundations of the Proposed Questions The questions we propose for the 2008 ANES Time Series Study will permit us to better understand the nature of political trust and its implications for beliefs and behaviors of the American public. To begin, our proposed measures of trust represent an improvement over the standard NES item. The proposed items are measured on a scale that allows for greater variation in response options. The standard trust in government question allows for only three or four response options (depending on how the volunteered never responses are treated) while the proposed questions yield a 101-point scale. In addition, with the proposed trust questions, the portion of the trust spectrum taken up by each response option is approximately equal. The standard trust question contains the unwieldy and expansive only some of the time category, and also the rather narrow just about always category. Third, the proposed trust question is balanced, explicitly giving respondents the opportunity to select responses at either end of the spectrum. In contrast, the standard NES question is lopsided towards the trusting end of the scale. Finally, the proposed questions use neutral language, avoiding the normative assumptions associated with the only some of the time verbiage of the standard question. While we find fault with the standard NES measure, we believe it warrants inclusion on the NES panel because it has been a mainstay of the NES and is widely used in academic research on elections. More importantly, however, including both the standard and the percentof-the-time questions on the same survey will allow for a close examination of what voters really mean when they report the extent to which they trust the government. For example, citizens
4 likely have different ideas in mind when they report that they trust government only some of the time. A parallel from the classroom may help illustrate this point: One student reports that he attends class only some of the time and shows up for about half of the classes. Another student also reports attending only some of the time, but rarely attends class. Thus, there is great variability in meaning within this one response option. Similarly, the difference in frequency between events that occur just about always and most of the time is unclear at best and likely varies from person to person. In sum, the vague wording of the response options on the standard NES question does a poor job at helping researchers understand how often citizens trust government. Including both the standard NES trust question and the 101-point percent-of-thetime question on the same survey will be invaluable for determining what respondents have in mind when answering the standard NES trust measure and will be helpful for analyzing political trust using existing surveys that do not include the improved measure. In addition to improving the NES trust measure, we hope to better understand the nature of trust. Evidence suggests that the standard NES trust measure captures temporal assessments of government. For example, Citrin and Luks (2001) have shown that political trust varies according to which political party controls Congress, with Democrats more likely to trust government when there is a Democratic-controlled Congress. Gershtenson, Ladewig, and Plane (2006) show that change in partisan control of just one chamber of Congress can affect trust evaluations. Other scholars have shown that citizens trust in government depends on whether their political party controls the presidency (Schaffner and Clark, 2004). The upsurge in political trust witnessed after the 9/11 terrorist attacks also points to the temporal nature of the standard trust measure (Hetherington and Globetti 2003).
5 Yet scholars often envision political trust as tapping into more deep-seated attitudes about the health of the political system. While presidential and congressional approval is expected to wax and wane as the political winds shift in Washington, political trust should remain more constant. That is, systemic trust is the belief that the American political system is essentially healthy and functional. While a decline in temporal trust should not cause much concern, a decline in systemic, diffuse trust should be disconcerting. In many ways, this debate over the nature of political trust traces its roots back to the early Citrin v. Miller debate (Citrin 1974; Miller 1974a; 1974b). By including open-ended items querying respondents what they were thinking when they indicated that their level of trust in government, we will be able to shed light on the systemic/diffuse debate and to more directly examine the origins of trust. That is, the open-ended items will help us move beyond the level of trust to examine why citizens may (or may not) trust government. Much of the research on political trust is built upon or motivated by the assumption that greater political trust is normatively desirable. In fact, this assumption can be seen in the wording of the standard NES question which includes the normatively-loaded only some of the time response options, suggesting that such a response is somehow suboptimal. The assumption that greater trust would improve democracy has not yet been thoroughly tested. Some research suggests that greater trust improves the quality of democracy by showing that trusting citizens are more likely to comply with laws (Tyler 1990, Tyler and Degoey 1995, Levi 1997, Scholz and Lubell 1998). However, there are also some theoretical reasons to expect that too much political trust threatens democracy. As Ruscio (2004, p. 4) puts it, A pathology of democratic life occurs when trust shades too far into unquestioned acceptance of a leader's dictates. In short, much remains unexplored. Thus, we propose questions to directly examine whether political trust is
6 associated with good citizenship. For example, does greater trust make it more likely that citizens will follow and participate in politics? Does greater trust increase citizens commitments to their communities and fellow citizens? The NES core already contains some relevant items (e.g., volunteer work, contributions to charity, following government and politics), but we feel that inclusion of the items on working with others in the community and on jury service are also central to evaluations of the relationships between trust and indicators of good citizenship. The final contribution of the proposed questions comes with regard to the level of government evaluated by respondents. To date, the trust literature focuses almost exclusively on the national government. While the national government is obviously important, in our federalist system, state governments retain their own sovereignty and have significant (and arguably, increasing) policy responsibilities. As a matter of daily life, decisions made by states may indeed have greater significance than those made by the federal government. State governments establish sales taxes, provide funding to public institutions, and create the civil and penal codes responsible for the vast majority of cases heard in judicial systems. There is some evidence that citizens are more approving of state governments than they are of the national government (Hibbing and Smith, 2001), with the perception of relative closeness to the people of state governments being suggested as the source of this pattern. Given this, one might anticipate citizens to also be more trusting of state governments than of the national government, but, we simply do not know if this is the case. Different perceptions of the levels of government may also be driven by the visibility of each level. Citizens tend to have little knowledge about politics in general, but the national government receives more media attention and is generally more visible.
7 Much as citizen awareness of government may influence the extent of trust, visibility of the levels of government may affect the basis on which citizens assess government. More generally, while some of the same forces likely influence state trust as influence national trust, there is no reason to assume that these determinants are identical. This is particularly true given the different responsibilities of each government and different citizen expectations of them. To date, the causes of trust in sub-national governments simply have not been explored. Empirical Performance of Proposed Questions and their Superiority over Alternatives As outlined above, we believe that the response options used for the standard NES trust measure ( just about always, most of the time, or only some of the time ) are problematic. Because the standard trust question has very few response options and the range of the trust spectrum taken up by each option is unequal (and likely overlapping), it can not be used to approximate a continuous, interval-level variable. We included two alternative sets of response options for trust measures on the 2006 NES Pilot. 1 Specifically, we included a five-point scale (always, most of the time, about half of the time, once in a while, and never) and a 101-point scale ( What percent of the time ). In short, the five-point scale did not perform much differently than the standard three-point scale because very few respondents claimed to trust the government always or never. In contrast, the 101- point scale yielded a nice dispersion in response options, with a mean of approximately 50% and a standard deviation of approximately 25%. Additionally, it was easy for respondents to understand, as only one respondent failed to provide a response. Finally, these 101-point trust measures were significantly correlated with the expected array of variables including approval of 1 The Pilot questions were slightly different, asking respondents how often they trusted the government to make decisions in a fair way and to do what is in the best interest of the country. While paired-difference t-tests show the items to yield distinctive responses, the two items demonstrated similar patterns with regard to their relationships (correlations) with other variables in the survey.
8 President Bush (both overall approval and approval of the President on the economy, foreign affairs, and the war on terrorism), political efficacy, party identification, and assessments of economic performance over the past year. We also included both the standard NES trust question and our 101-point alternative on a recent national survey conducted by the Center for Survey Research and Analysis (CSRA) at the University of Connecticut. The results provide further evidence that the response options for the standard question are wanting. For those who volunteered that they "never" trusted the government on the standard NES trust question, 43% of them said they trusted the government 0% of the time, but the mean was 8% and the responses ranged all the way up to 45% of the time. The "only some of the time" response option is much more ambiguous, with respondents providing a percentage range from 0 percent of the time up to 98% (with 41% as the mean and 50% as the mode). This range overlaps considerably with the "most of the time" respondents, who varied from 0% to 90% with a mean of 68% and a mode of 75%. Finally, the "just about always" respondents provided answers ranging from 0 to 95, with 71% as the mean and 90% as the mode. Clearly, within each response option of the standard NES question, citizens gave a wide range of responses to the percent of the time that they trusted government. We also have evidence regarding the performance of both the state-level government trust item and the open-ended items. Very similar questions were included on a survey administered to a random sample of Kentucky residents in the summer of The data both confirmed some expectations and yielded some surprises (Gershtenson and Plane 2006a, 2006b). As anticipated, Kentuckians expressed higher trust in their local (county) governments than in the state and national governments. On the other hand, Kentuckians were actually less trusting of their state government than they were of the national government in the survey from which our
9 data come. On its face, this might call into question previous assertions that citizens have greater confidence in state governments. However, we suspect that the political conditions in Kentucky during the administration of the survey are responsible for the results. The ongoing, high-profile inquiry into the hiring practices of Governor Fletcher s administration that resulted in indictments of several government officials almost certainly had a deleterious effect on citizen attitudes toward the state. (In the open-ended item asking respondents why they evaluated the state government the way they did, 16% cited political corruption or patronage.) The open-ended items on the survey provided information about the nature of trust and showed that citizens are able to distinguish between levels of government. For instance, 35% of respondents cited the war in Iraq and/or the war against terrorism when accounting for their trust in the national government, while 0% did so for the state government. Furthermore, the responses showed that there is some mix of systemic and diffuse considerations that enter citizens minds when asked about trust in government. Of course, getting additional data from the NES would extend this research beyond Kentucky. It would also allow for examining determinants of trust in the different levels of government. The Kentucky study suggested that variation exists across the levels, particularly with regard to the role of partisanship. Potential Statistical Analyses Using Proposed Questions The analyses envisioned on the basis of the proposed questions range from frequency distributions to multivariate causal models. First, we will be able to compare citizens responses on the standard NES trust question with their responses on the percent-of-the-time question to get a fuller understanding of what citizens mean when they report, for example, that they trust the government just about always. This analysis will be like that described above for the
10 CSRA survey and we anticipate it will reveal the problematic nature of the response options on the standard NES trust item. We also believe that the proposed 101-point trust measure offers other benefits over the standard NES item. In particular, the 101-point trust measure should perform better as both a dependent variable and as an independent variable. Consequently, we will conduct multivariate analyses replicating to the extent possible analyses from the literature. For example, we can model the vote for the incumbent party presidential candidate as a function of political trust, economic evaluations, political ideology, and other variables following Hetherington (1999). Similarly, we can examine the role of trust in individuals support for liberal governmental policies (Hetherington 2004), in affect toward candidates (Hetherington 1998), and in whether or not citizens vote in elections (Hibbing and Theiss-Morse 2002). Finally, we can model behavior characteristic of good citizenship as a function of political trust. With trust as a dependent variable, we can examine the roles of partisanship, economic evaluations, political news consumption, etc. following Hetherington (1998), Gershtenson, Ladewig, and Plane (2006), and others. Of course, we also believe that the determinants of state trust may differ from those of national trust. Consequently, the multivariate analyses would be done for both levels of government. We are also interested in examining differences in trust across levels of government and would therefore compare the magnitude of trust in the national and state governments. Perhaps the most onerous analyses will be those involving the open-ended items. Here, we will need to read responses in order to develop an appropriate coding scheme and perform content analysis. While time consuming, these analyses should provide considerable insights into the nature and determinants of political trust.