Opinions, Perceptions, and Networks: Opinion Types and Their Origins

Page 1

Opinions, Perceptions, and Networks: Opinion Types and Their Origins Peter William Hurford Project Advisor: Paul A. Djupe Department of Political Science Denison University Summer Scholars Project Date: 2011

Abstract: Many previous pieces in the political science literature have studied the causes that drive political opinions in respondents, but have not checked to see whether these cause differ when the content of the opinion differs. In this paper, we look to see if there are different kinds of political opinions with different underlying causes, and then look at what those causes are, with a special focus on the influence interpersonal networks of discussion partners.

Peter Hurford Opinions, Perceptions, and Networks: Opinion Types and Their Origins Political Science has long been interested in how voters acquire political information and use it to generate opinions and judgments about issues and people in politics. This study of opinionation represents the most basic level of political behavior, because it is impossible to do other political activities like vote or discuss politics unless you have judgments and opinions to vote by or discuss. Locating the sources of opinions would also define the field of politics much further, such as determing whether voters get opinions from outside groups, such from the media and campaign literature, or from discussing politics with acquaintances. All prior studies in opinionation have treated it as a single phenomena (but see Djupe 2011; Gilens 2001). Most have employed a composite index formed of opinions about policies and groups (Atkeson and Rapoport 2003; Berinsky 2002; Delli Carpini and Keeter 1993), but some have used the absence of party likes/dislikes (Atkeson and Rapoport 2003; Leighley 1991), some have evaluated opinions about political figures specifically (Gimpel and Wolpert 1996), and another has used non-neutral responses to issue questions (Jacoby 1995). Different studies have analyzed opinionation on multiple spectrums varying from whether the respondent will self-report certain characteristics of himself or herself (Kim, Wyatt, and Katz 1999), the respondent’s views on how the government should act on certain issues (Berinsky 2002; Kim, Wyatt, and Katz 1999), whether the respondent will rate a group (Djupe 2009), and whether a respondent will like or dislike a candidate (Atkeson and Rapoport 2003; Gimpel and Wolpert 1995) or a political party (Leighly 1991). Some studies, such as Francis and Busch (1975) and Rapoport (1985), looked at opinions on almost all questions in a survey that call for an opinion, regardless of what kind of information the question is specifically calling for. In each of these studies, the specific kinds of opinions used were stated to be representative of opinionation in 1

general. Generalizing opinionation carries the assumption that the factors causing people to state opinions on widely different questions are the same – that opinionation is a measure of confidence, willingness, and ability to state an opinion, regardless of what that opinion is. However, this may not be the case. At least one scholar has called on researchers to study different types of opinionation, stating that “different scholars employ various measures of information recall without considering the possibility (or the likelihood) that some kinds of information may be more desirable or consequential than others� (Druckman 2005: 517; also see Gilens 2001 and Djupe 2009). It could be instead that what factors lead a respondent to feel confident enough to rate a group may not be the same factors that lead a respondent to like a candidate or state a view on how the government should act. If there are different kinds of opinionation, then certain characteristics of the respondent might lead to holding opinions on some questions but not on others. Furthermore, if different studies into opinionation assume they are analyzing the same general concept of opinionation, but really are looking into different types of opinions, this could cause inconsistencies in the literature that would be resolved by recognizing different opinion types. An analysis of the existing literature may point to some of these differences. For instance, in some studies the race of the respondent had no statistically significant effect on opinionation (Gimpel and Wolpert 1995; Kim, Wyatt, and Katz 1999), but in others being nonwhite was found to negatively affect opinionation (Atkeson and Rapoport 2003; Francis and Busch 1975)1. Also, some studies found age to correlate positively with opinionation (Atkeson and Rapoport 2003; Gimpel and Wolpert 1995; Leighly 1991), one study found it to


Another difference that divides the two studies is that Atkeson and Rapoport (2003) and Francis and Busch (1975) include income, whereas Gimpel and Wolpert (1995) and Kim, Wyatt and Katz (1999) do not.


correlate negatively with opinionation (Kim, Wyatt, and Katz 1999), and other study found it to make no statistically significant change (Francis and Busch 1975). Lastly, consider gender – while being male has been found to promote opinionation (Atkeson and Rapoport 2003; Djupe 2009; Francis and Busch 1975), in one study it was found unimportant (Kim, Wyatt, and Katz 1999). Therefore, we pose two central questions regarding opinionation: (1) are there different types of opinionation and (2) if so, what factors drive opinionation in these different types? However, in order to get a meaningful answer to both these questions, they will have to be answered simultaneously because we cannot know for sure if there are different types of opinionation unless the different types have different causes2, and such an analysis of the causes of opinionation would answer the second question. I’d suggest that this would be a good place for something like, “We focus attention on one particularly useful determinant – personal social networks. Drawing on a long line of literature on the political effects of social networks, we test the specific, competing assumptions made about the provision of information that map onto the dimensions of opinionation we assess. This also constitutes the first tests of the effect of networks on opinionation, which also has led us to confront some of the specific challenges network research faces in studying public opinion. In what follows, we will look at survey questions from the 2000 ANES that call for opinions on different subjects, such as personal opinions on campaign issues, perceptions of where presidential candidates (Bush and Gore) stand on those campaign issues, personal opinions on retrospective issues, personal opinions on government spending, and personal opinions of different groups. We will assess if there are any differences both in amounts of opinion types held and in the underlying causes of opinionation. We will also specifically look not just at 2

We can, however, have initial evidence if it turns out that the amount of opinionation across different questions have different means.


the resources and characteristics of a respondent, but at their interpersonal networks to see if the socially supplied information from political discussion supports learning on all subjects, just some, or none at all.

Personal Resources as an Explanation for Opinionation A search for political information is demanding to the point that few people will undertake it (Downs 1957). This means that those who have more resources will face a less costly search and therefore will be more likely to possess political information and the political opinions that follow from that information. Studies in opinionation have repeatedly confirmed this. People with lower resources, such as less education and less income, have been found to have fewer opinions (Krosnick and Milburn 1990; Verba, Schlozman, and Brady 1995; Zaller 1992). Nonwhites have also been known as being disadvantaged with regard to resources and therefore also have been less likely to express opinions (Francis and Busch 1975). Sometimes resource differences leading to lower opinionation are sociological and psychological rather than material, and persist even when controlling for education and income. For instance, women also hold fewer opinions than men and this trend has continued despite women having much greater access to political resources and engaging in far more political activity than in previous decades (Atkeson and Rapoport 2003; Burns, Schlozman, and Verba 2001; Djupe 2011; Rapoport 1982, 1985).

Environmental Factors as Causes of Opinionation Additionally, opinionation is related to not just resources but characteristics of the environment in which the opinions are formed. For instance, opinionation rises when people pay more attention to campaigns – opinionation is not constant from election to election and instead varies based on the vibrancy of the campaign environment (Atkeson and Rapoport 2003; Caldeira and Smith 1996; Krosnick and Milburn 1990; Zaller 1992). Opinionation also increases when people pay


more attention to media (Kim, Wyatt, and Katz 1999). Interpersonal Networks as Causes of Opinionation While the existing literature has discussed how environmental factors can lead to opinionation and has touched on how political discussion correlates with opinionation and opinion quality (Djupe 2009; Kim, Wyatt, and Katz 1999), little has been written on how a respondent’s interpersonal network, the groups of people which the respondent discusses politics with, could promote opinion holding. This is not to say that there hasn’t been discussion of connections between opinions and interpersonal networks. Analysis of networks has also typically focused on how network factors such as amount of general disagreement (Djupe 2009; Huckfeldt et al. 2004; Huckfeldt and Mendez 2005; Kenny 1998; Mutz 2006; Sokhey and McClurg 2008), amount of political expertise (Huckfelt 2001; McClurg 2006), frequency of political discussion (McClurg 2003), and size (Kenny 1992; McClurg 2003; McClurg 2006) have affected opinion latency, opinion direction, and political participation. Results have shown those with larger and more politically knowledgeable networks strongly correlates with political participation (Kenny 1992; McClurg 2003; McClurg 2006) and that political disagreement strongly influences voting for the candidate who matches the voter’s preferences (Sokhey and McClurg 2008) but may also cause people to avoid political activity altogether (Huckfeldt and Mendez 2005; Mutz 2006). These measures, especially participation, are correlated with opinionation, but no study of social networks has investigated opinionation specifically. It turns out that studying the effects of networks on different types of opinionation is useful in helping to sort out how networks affect political behavior. There are competing assumptions made in the literature about how, for instance, networks affect political participation. Given the conceptual differences


between opinionation types, we can assess whether networks help provide information about transient figures in electoral politics, clarify stances on policy, help position the voter amidst a sea of reference groups, and/or help them understand the role of political institutions in shaping the current world. While the search for information is aided by more resources, it can also be aided by additional heuristics that an otherwise disadvantaged respondent could use to approximate this information, such as using endorsements from interest groups to emulate the behavior of well-informed voters (Lupia 1994) or socially supplied information from political discussion (Djupe 2009; Downs 1957; Kim, Wyatt, and Katz 1999; Sokhey and Djupe 2010). The key focus regarding networks is the suggestion that individuals involved in the political discussion will acquire political information (Berelson et al. 1954; Huckfeldt et al. 1995; Levine 2005) with the implication that this information can then be used to form opinions (Djupe 2009) – the idea that people learn from their networks. However, this brings up several questions, such as what exactly people are learning from these networks, and which, if any, kinds of opinions networks promote. Do networks help build opinions regardless of what kind of content the question calls for, or are networks only helpful in acquiring certain kinds of information sufficient to hold opinions?

Data and Measurement To find out whether or not there are multiple kinds of opinionation, and to find out what the underlying causes of opinionation are, data was collected from the 2000 National Election Survey 3.

If there truly are multiple, distinct kinds

of opinionation, these factors will drive each kind of opinionation differently. For instance, perhaps some factor will correlate significantly with personal opinionation, but won’t matter at all when it comes to determining candidate 3

We chose the 2000 NES specifically because it contains questions that allow us to analyze network effects.



Personal Opinionation In this survey respondents were asked, among other things, to provide numerous opinions on a wide variety of questions. One set of questions asked respondents for their personal stances on many campaign issues, such as whether or not abortion should be legal, whether it should be more or less difficult to get a gun, whether environmental regulation should be increased or decreased, whether defense spending should be increased or decreased, whether or not the government should guarantee jobs, whether the government should provide aid to blacks, and the ideology of the respondent. In addition to the standard “don’t know” and “refuse” options, on the last five issues respondents were specifically given the choice of saying they “haven’t thought much” on the issue instead of giving a stance, which we counted as no opinion. When “haven’t thought much” was an option for the issue, the amount of respondents failing to report an opinion increased, except on ideology, consistent with other studies that show an increase in nonresponse when such an option is added (Bishop et al. 1980; Shuman and Presser 1981). Candidate Opinionation Additionally, all of the respondents were asked of their perception of the candidates – George W. Bush and Al Gore – on these same seven issues, for a total of fourteen possible perceptions. On these fourteen questions, there was no “haven’t thought much” option, so all fourteen questions were identical in form. Table 1 shows the levels of personal and candidate opinionation on each issue. Overall, there are no statistically significant differences between perceptions of the two candidates on the same issue, but there is a statistically significant difference between the amount of respondents holding a candidate perception on an issue and the amount of respondents holding a personal opinion on the issue.


Retrospective Opinionation Another source of questions that call for opinions involve issues where the voter is asked about retrospective issues; questions about how America has changed, for better or for worse. The questions in the 2000 NES focused specifically on four areas – the economy, national security, the amount of crime, and “moral decay” – asking the respondent to report whether America got better or worse in this area, and then asking whether the respondent thinks Clinton made America better or worse in this area. Voting behavior literature has long suggested that elections are referendums on the current administration, where citizens cast their vote based on how the most recent administration preformed (Key 1965). However, if voters do not have opinions on these retrospective issues, for example failing to connect the Clinton administration to how America has changed one way or another, they cannot vote retrospectively. Table 2 shows the amount of retrospective opinionation viewed with and without considering Clinton. Respondents appear highly opinionated on these issues and there is no statistically significant difference between opinionation levels on the initial retrospective issues and those that ask the respondent to assign blame or praise to Clinton. Spending Opinionation A fourth set of questions that call for opinions is the battery of questions asking for the respondent’s opinion on government spending – whether spending should be raised or lowered on highway repair, welfare, AIDS research, foreign aid, food stamps, aid to the poor, social security, environmental regulation, public schools, crime prevention, child care, illegal immigration prevention, and black aid. On these issues, respondents also appear very highly opinionated, with opinion levels on individual issues seen in Table 3. Group Opinionation A fifth set of questions call for opinions on groups using thermometers that


measure the respondent’s “warmness” or “coolness” to 24 different groups of people, whether that group is an institution of the government, an organization, or defined by a common characteristic, such as liberals, people on welfare, or blacks. Table 4 shows a complete list of these groups and the opinionation on each of them, which run above 90% on every group except fundamentalists (%) and The Christian Coalition (%). Measuring Means An initial way to determine whether these four types of opinionation really are different is to see if the mean amount of opinions held in each type is different to a statistically significant degree. Because all four types had a different amount of questions asked, the only way to do this is to convert each group into a percentage, where respondents are measured by a percent of all possible opinions on each of the four types. The mean percentage of opinionation held across all five types is shown in Figure 1, which reveals that there are statistically significant differences among personal opinionation, candidate opinionation, and group opinionation; with each those three different from retrospective opinionation and spending opinionation, which do not themselves have statistically significant differences in the mean amount of opinions held.4 However, a difference in means is not enough to indicate that each type of opinionation is distinct, because each type could have the same causes, just with different strengths. Looking at Networks The 2000 NES data allows us to evaluate the networks of the respondents, by asking the respondents to report up to four people they discuss politics with. 74.27% of respondents have a network, which means they indicated political discussions with at least one person. Additional questions ask the respondent how much they discuss politics with each person they identified, giving us a measure of discussion frequency; whether the respondent knew other people in the


network, giving us a measure of network insularity; how much they thought each partner knew about politics, giving us a measure of expertise within the network; and who their partner voted for president, which indicates disagreement if the vote is different from the respondent. However, in creating these variables we have a choice – do we include in our measurements of network characteristics all of those who have no network, and therefore would experience no discussion frequency, no network insularity, no network expertise, and no network disagreement? This choice causes a dilemma, for if we do include these people without networks, all of our measurements become collinear to the point that they cannot all be included in the same model without invalidating each other. However, if we do not include these people without networks, we won’t have any idea whether the possession of a network causes a respondent to have access to political information or not, and therefore indicate whether networks can stimulate opinion holding. Networks and Resource Effects In addition to the dilemma surrounding those without networks, we have another problem – while ideally network analysis would provide a way to look at how respondents learn information independent from resource effects, it doesn’t work this way as people have a tendency to associate with people like themselves (Huckfeldt and Sprague 1995; Lazarsfeld and Merton 1954), thus forming social networks with similar demographics (Berelson et al. 1954). This means that any resource effects present within the respondent will be present within the network as well. For instance, those with low education will have less access to educated respondents, and thus miss out on network expertise. Also, those who are uninvolved in politics as a whole are less likely to have sought out and created a political network. This problem can be seen in the data. Figure 2 shows that those respondents who have a network also consume significantly more media, are more partisan, and are significantly more likely to be politically interested. This is currently an unsolved problem in political science for us to note, and work is underway to


explore more advanced models to take this into account, such as models that consider “treatment effects” in non-experimental data. Creating Models To look at the actual causes among opinionation, we will create and compare five different models, with each model containing one type of opinionation as the dependent variable and the same potential causes as the independent variables. Because of the skewed distribution of opinionation among all five, with many more holding all possible opinions than no opinions, we will represent personal opinionation5, candidate opinionation, and group opinionation in quartiles and create ordered logistic models for these. Retrospective opinionation and spending opinionation, where more than 80% of respondents held all possible opinions, will be dichotomized into a variable that represents whether the respondent held all possible opinions or not. The independent variables will include many of the resources discussed earlier and thought to affect opinionation, such as the respondent’s age (eg., Atkeson and Rapoport 2003; Huckfeldt and Mendez 2004; Gimpel and Wolpert 1995), gender (e.g., Djupe 2009; Gimpel and Wolpert 1995; Leighly 1991), education (eg., Atkeson et al. 2003; Djupe 2009; Huckfeldt and Mendez 2004), whether the respondent is employed, whether the respondent is a homemaker (Atkeson and Rapoport 2003), the respondent’s income (Atkeson and Rapoport 2003), the respondent’s race (Atkeson and Rapoport 2003; Kim, Wyatt, and Katz 1999; Leighly 1991), the respondent’s sense of political interest (eg. Atkeson and Rapoport 2003; Djupe 2009; Leighley 1991), the extremity of the respondent’s partisanship (eg., Atkeson and Rapoport 2003; Djupe 2009; Huckfeldt and Mendez 2004), and political activity levels (Atkeson and Rapoport 2003). 5

Only the “haven’t thought much” questions (environmental regulation, defense spending, ideology, government jobs, and black aid) were considered because they are methodologically distinct from the two questions where “haven’t thought much” was not an option (abortion and gun control), which could skew the results. Since the questions involving “haven’t thought much” were more numerous, they made for a better model.


Additionally, the independent variables will include measures of the respondent’s political environment, such as how often the respondent was contacted by campaigners (Djupe 2011), how much media the respondent consumes (eg. Djupe 2009; Huckfeldt and Mendez 2004; Leighley 1991; Wyatt and whoever), and whether or not the respondent lives in a battleground state – a state where the election was decided by less than 5% – which would typically be a more vibrant and politicized environment (Zaller 1992). Analyzing Model Results When it comes to candidate opinionation, shown in Table 5, the amount of perceptions held about candidate opinions is increased by age, being male, having a higher income, being more partisan, consuming media, self-reporting political interest, and participating in politics. Being more educated, living in a battleground state, being employed, being a homemaker, being contacted by outside groups, and having a network do nothing statistically significant to affect candidate opinionation. Personal opinionation (Table 6), however, is increased by being male, being educated, having a higher income, being white, consuming media, self-reporting political interest, and having a network; with age, being partisan, participating politically, living in a battleground state, being employed, being a homemaker, and being contacted by outside groups having no statistically significant effect. Retrospective opinionation (Table 7) is increased by being male, consuming media, and having a network; and is not affected by age, income, race, political interest, partisanship, participation, battleground state status, employment, homemaker status, or being contacted. The model of spending opinionation (Table 8) was much worse and no variables could be shown to have a statistically significant effect, with the exception of age, which makes spending opinionation go up. Lastly, group opinionation (Table 9) is increased by being male, being


educated, consuming media, participating in politics, and having a network; decreased by age; and not effected by income, race, political interest, partisanship, battleground state status, employment, homemaker status, or outside contact.

Conclusions Overall, our results answer both our questions. First, we demonstrate that since all five types of opinionation have distinct underlying causes, there are multiple types of opinionation and future analysis must take this into account. However, when we look at which factors correlate with which types of opinionation, we can reveal an explanatory narrative of how information is used to form different types of opinionation, answering our second question. The three big players when it comes to explaining opinionation seems to be resource effects, media consumption, and having a network.

Resource Effects Resource effects seem to increase opinionation regardless of the kind, demonstrating that information truly is less difficult to obtain for the advantaged members of society, regardless of what that information is. Democracy is profoundly affected by inequalities in opinionation brought on by resource effects, since those who lack opinions will not be able to deliberate over policy and make informed choices (Atkeson and Rapoport 2003; Huckfeldt and Sprague 1995), which will prevent them from participating in politics (Atkeson and Rappoport 2003; Leighley 1991). Additionally, these deep roots in resource inequalities strongly indicate that people who are resource disadvantaged will also be politically disadvantaged, unable to politically participate. This creates a vicious cycle, as those who are disadvantaged are the ones who most need to use the political process to get help, yet are also most likely to be unable to make use of the political process. Our data confirms that disadvantaged people are also less likely to be able to form 13

social networks, and when they do, they are less likely to have a network that is as effective, lacking in discussion frequency and political expertise.

Media Effects With the exception of spending opinionation which remains inexplicable for reasons unknown, media consumption also increases opinions across the board, regardless of the content of opinion called for by the question. This may either indicate that the media is effective in disseminating the wide variety of information necessary to form all sorts of opinions, that the people who consume media are just more likely to be the type of people interested in forming opinions (Kim, Wyatt, and Katz 1, or some combination of the two.

Network Effects Kim, Wyatt, and Katz (1999) wrote that “conversation is the soul of democracy�, and this seems true here as well, though with some reservation. Having a network increases group, retrospective, and personal opinionation, but has no effect on spending or candidate opinionation. This helps confirm the theory that political discussion does not help people gain general information in the same way that media consumption does, but that personal deliberation helps those involved connect their values to a rating of key campaign issues, retrospective issues, or various groups. Specifically, a discussant may know some facts about abortion, but may not turn these facts into a personally held opinion until he or she is asked to declare his or her opinion to others, or adopts opinions from others that are consistent with known facts (see McPhee 1963).


References Atkeson, Lonna Rae, and Ronald B. Rapoport. 2003. “The more things change the more they remain the same: Examining gender differences in political opinionation, 1952-2000”. Public Opinion Quarterly 67:495-521. Berelson, Bernard, Paul Lazarsfeld, and William McPhee. 1954. Voting. Chicago: University of Chicago Press. Berinsky, Adam J. 2002. “Political Context and the Survey Response: The Dynamics of Racial Policy Opinion.” The Journal of Politics 64 (2): 567584. Bishop, George, Robert Oldendick, and Alfred Tuchfarber. 1980. “Experiments in filtering political opinions.” Political Behavior 2 (4): 339-370. Burns, Nancy, Kay Lehman Schlozman, and Sidney Verba. 2001. The Private Roots of Public Action. Cambridge, MA: Harvard University Press. Caldeira, Gregory A. and Charles E. Smith Jr. 1996. “Campaigning for the Supreme Court: The Dynamics of Public Opinion on the Thomas Nomination.” The Journal of Politics. 58 (3): 655-681. Delli Caprini and Keeter, Scott. 1993. “Measuring Political Knowledge: Putting First Things First.” American Journal of Political Science 37 (4): 11791206 Djupe, Paul A. 2011. “Political Pluralism and the Information Search: Determinants of Group Opinionation.” Political Research Quarterly 64(1): 68-81. Downs, Anthony. 1957. An economic theory of democracy. New York: Harper and Row. Druckman, James N. 2003. “The power of television images: The first KennedyNixon debate revisited.” Journal of Politics 65: 559–571. Druckman, James N. 2005. “Does Political Information Matter?” Political Communication 22: 515-519. Francis, Joe, and Lawrence Busch. 1975. “What we don't know about 'I don't knows'.” Public Opinion Quarterly 39: 207-218. Gilens, Martin. 2001. “Political ignorance and collective policy preferences.” 15

American Political Science Review 95 (2): 379-96. Gimpel, James G. and Robin M. Wolpert. 1996. “Opinion-holding and public attitudes toward controversial Supreme Court nominees.” Political Research Quarterly 49 (1): 163-76. Graber, Doris A. 2001. Processing politics: Learning from television in the Internet age. Chicago: University of Chicago Press. Huckfeldt, Robert, Paul A. Beck, Russell J. Dalton, and Jeffrey Levine. 1995. “Political environments, cohesive social groups, and the communication of public opinion.” American Journal of Political Science 39: 1025-54. Huckfeldt, Robert, and John Sprague. 1995. Citizens, Politics, and Social Communications: Information and Influence in an Election Campaign. Cambridge: Cambridge University Press. Huckfeldt, Robert. 2001. “The Social Communication of Political Expertise.” American Journal of Political Science. 45 (2): 425-438. Huckfeldt, Robert, Jeanette Morehouse Mendez, and Tracy Osborn. 2004. “Disagreement, Ambivalence, and Engagement: The Political Consequences of Heterogeneous Networks.” Political Psychology 25 (1): 65-95 Huckfeldt, Robert, and Jeanette Morehouse Mendez. 2005. “Managing Disagreement Within Communication Networks: Moths, Flames, and Political Engagement.” Jacoby, William G. “The Structure of Ideological Thinking in the American Electorate.” American Journal of Political Science. 39 (2): 314-335. Kenny, Christopher B. 1992. “Political Participation and Effects from the Social Environment.” American Journal of Political Science 36(1): 259-67. Kenny, Christopher B. 1998. “The Behavioral Consequences of Political Discussion: Another Look at Discussant Effects on Vote Choice.” Journal of Politics 60: 231- 244. Key, V. O. Jr. 1965. The Responsible Electorate: Rationality in Presidential Voting, 1936-1960. Harvard: Belknap Press. Kim, Joohan, Robert O. Wyatt, and Elihu Katz. 1999. “News, Talk, Opinion,


Participation: The Part Played by Conversation in Deliberative Democracy.” Political Communication 16: 361-385. Krosnick, Jon A. and Michael Milburn. 1990. “Psychological determinants of political opinionation.” Social Cognition 8 (1): 49-72. Lazarsfeld, Paul and R. Merton. 1954. “Friendship as a Social Process: A Substantive and Methodological Analysis.” Freedom and Control in Modern Society, ed. Morroe Berger, Theodore Abel, and Charles H. Page. New York, NY: Van Nostrand. Leighley, Jan. “Participation as a Stimulus of Political Conceptualization.” The Journal of Politics 53 (1): 198-211 Levine, Jeffrey. 2005. “Choosing Alone? The Social Network Basis of Modern Political Choice.” The Social Logic of Politics: Personal Networks as Contexts for Political Behavior, ed. Alan S. Zuckerman. Philadelphia, PA: Temple University Press. Lupia, Arthur. 1994. “Shortcuts versus encyclopedias: Information and Voting Behavior in California Insurance Reform Elections”. American Political Science Review 88 (1): 63–76. McPhee, William, with Robert B. Smith and Jack Ferguson. 1963. “A Theory of Informal Social Influence.” In William McPhee, ed. Formal Theories of Mass Behavior. London: Collier-Macmillan, Free Press. (pp. 74-103). Mutz, Diana C. 2006. Hearing the Other Side: Deliberative Versus Participatory Democracy. Cambridge: Cambridge University Press. McClurg, Scott D. 2003. “Social Networks and Political Participation: The Role of Social Interaction in Explaining Political Participation.” Political Research Quarterly 56 (4): 449–64. McClurg, Scott D. 2006. “The Electoral Relevance of Political Talk: Examining Disagreement and Expertise Effects in Social Networks on Political Participation” American Journal of Political Science 50 (3): 737-754. Rapoport, Ronald B. 1982. “Sex Differences in Opinionation: A Generational Explanation” The Public Opinion Quarterly 46 (1): 86-96. Rapoport, Ronald B. 1985. “Like mother, like daughter: Inter- generational


transmission of DK response rates.” Public Opinion Quarterly 49 (2): 198208. Schuman, Howard, and Stanley Presser. 1981. Questions and Answers: Experiments in Questionnaire Wording. New York: Academic Press. Sokhey, Anand E. and Scott D. McClurg. 2008. “Social Networks and Correct Voting.” Journal of Politics. Verba, Sidney, Kay Lehman Schlozman, and Henry E. Brady. 1995. Voice and equality: Civic voluntarism in American politics. Cambridge, MA: Harvard University Press. Zaller, John R. 1992. The nature and origins of mass opinion. Cambridge, UK: Cambridge University Press.


Table 1: Percent of Respondents Holding a Personal/Candidate Opinion on Seven Issues, with T-Tests Issue Personal Gore Bush t-test t-test Opinion? Perception? Perception? (Per/Gore) (Gore/Bush) Gun Control 99.45 83.45 82.84 <0.0001 0.6251 Abortion 99.00 74.76 72.61 <0.0001 0.1408 Defense Spending 78.64 77.64 79.03 0.4689 Environmental Reg. 74.65 74.99 72.61 0.8192 Ideology 98.95 86.83 86.55 <0.0001 Providing Jobs 88.27 80.13 79.58 <0.0001 Providing Black Aid 86.39 83.45 82.84 <0.0001 N=1807, Source: 2000 NES; The bottom five issues have “haven’t thought much” as a response option whereas the top two do not.

Table 2: Percent of Respondents Holding an Opinion on Retrospective Issues Perception? Issue Personal Clinton t-test State of Economy 98.51 97.40 0.0188 State of Security 97.07 97.01 0.9219 Amount of Crime 97.40 96.29 0.0570 Moral Decay 98.28 97.73 0.2340 N=1678, Source: 2000 NES

Table 3: Percent of Respondents Holding an Opinion on Spending Issues Issue % of Resp. Highway Spending 98.9 Welfare 98.23 AIDS Research 97.95 Foreign Aid 97.57 Food Stamps 97.18 Aid to the Poor 98.07 Social Security 97.90 Environmental Reg. 98.11 Public Schools 99.50 Crime Prevention 99.06 Child Care 97.78 Illegal Immigration 97.18 Black Aid 95.08


0.3129 0.1039 0.8066 0.6784 0.3401

N=1807, Source: 2000 NES

Table 4: Percent of Respondents Holding an Opinion on Groups Issue % of Resp. Supreme Court 95.43 Congress 95.82 Military 97.56 Federal Government 96.21 Blacks 93.38 Whites 93.83 Conservatives 92.54 Liberals 92.35 Unions 94.41 Big Business 96.14 Poor 92.99 People on Welfare 92.99 Hispanics 92.48 Fundamentalists 82.77 Women’s Movement 94.92 Old People 97.17 Environmentalists 95.43 Gays 93.12 The Christian Coalition 81.35 Catholics 92.93 Jews 91.25 Protestants 91.25 Feminists 91.77 Asians 91.13 N=1555, Source: 2000 NES

Table 5: Ordered Logistic Model of Total Amount of Perceptions Stated of Candidates on the Seven Issues (Split into quartiles) Variable Coeff. Std. Err. p-value R.’s Age -0.014 0.004 <0.001 Is R. Male? +0.861 0.115 <0.001 R.’s Education +0.039 0.025 0.118 R.’s Income +0.060 0.022 0.007 Is R. White? -0.245 0.135 0.069 R.’s Political Interest +0.165 0.043 <0.001 R.’s Partisanship +0.197 0.054 <0.001 20

R.’s Media Consumption +0.212 0.058 <0.001 R.’s # of Votes +0.251 0.084 0.003 Is R. In Battleground? -0.139 0.113 0.221 Is R. Employed? -0.216 0.139 0.115 Is R. a Homemaker? -0.017 0.200 0.931 R.’s Contact -0.011 0.047 0.822 Does R. Have Network? +0.177 0.132 0.181 /cut1 -0.771 0.458 /cut2 +0.310 0.457 /cut3 +1.348 0.458 N=1357, LogLikelihood: -1615.6593, LRX2: 274.31, Sig-X2: <0.0001, PsuedoR2: 0.0782, Source: 2000 NES

Table 6: Ordered Logisitic Model of Total Opinions Stated on the Five Issues Where “Haven’t Thought Much” Was an Option (Split into quartiles) Variable Coeff. Std. Err. p-value R.’s Age +0.002 0.004 0.702 Is R. Male? +0.738 0.125 <0.001 R.’s Education +0.099 0.026 <0.001 R.’s Income +0.066 0.027 0.012 Is R. White? +0.629 0.137 <0.001 R.’s Political Interest +0.132 0.048 0.005 R.’s Partisanship -0.069 0.059 0.242 R.’s Media Consumption +0.204 0.063 0.001 R.’s # of Votes +0.082 0.087 0.342 Is R. In Battleground? -0.192 0.124 0.122 Is R. Employed? -0.078 0.148 0.600 Is R. a Homemaker? +0.268 0.213 0.208 R.’s Contact +0.015 0.051 0.758 Does R. Have Network? +0.447 0.137 0.001 /cut1 -0.116 0.485 /cut2 +1.223 0.481 /cut3 +2.552 0.486 N=1357, LogLikelihood: -1322.6592, LRX2: 239.47, Sig-X2: <0.0001, PsuedoR2: 0.0830, Source: 2000 NES

Table 7: Logistic Model of Total Perceptions Stated on Retrospective Issues (Dichotomized) Variable Coeff. Std. Err. p-value R.’s Age -0.013 0.007 0.052 21

Is R. Male? +0.826 0.234 <0.001 R.’s Education +0.040 0.042 0.345 R.’s Income +0.014 0.046 0.758 Is R. White? -0.156 0.250 0.533 R.’s Political Interest +0.096 0.081 0.233 R.’s Partisanship -0.068 0.102 0.506 R.’s Media Consumption +0.247 0.109 0.023 R.’s # of Votes +0.278 0.143 0.053 Is R. In Battleground? +0.263 0.223 0.237 Is R. Employed? -0.142 0.264 0.589 Is R. a Homemaker? -0.326 0.323 0.313 R.’s Contact -0.097 0.090 0.280 Does R. Have Network? +0.551 0.222 0.013 _cons +1.469 0.792 0.064 N=1357, LogLikelihood = -384.38809, LRX2: 74.64, Sig-X2: <0.0001, PseudoR2: 0.0885

Table 8: Logistic Model of Total Perceptions Stated on Spending Issues (Dichotomized) Variable Coeff. Std. Err. p-value R.’s Age -0.025 0.006 <0.001 Is R. Male? +0.242 0.172 0.160 R.’s Education -0.043 0.036 0.232 R.’s Income -0.051 0.028 0.072 Is R. White? -0.027 0.212 0.898 R.’s Political Interest -0.009 0.066 0.894 R.’s Partisanship +0.014 0.083 0.869 R.’s Media Consumption +0.052 0.088 0.550 R.’s # of Votes +0.173 0.126 0.169 Is R. In Battleground? -0.402 0.166 0.016 Is R. Employed? +0.059 0.209 0.780 Is R. a Homemaker? -0.026 0.309 0.932 R.’s Contact +0.094 0.072 0.191 Does R. Have Network? -0.003 0.200 0.987 _cons +3.349 0.710 <0.001 2 N=1357, LogLikelihood = -545.00613, LRX : 37.79, Sig-X2: 0.0006, Pseudo-R2: 0.0335

Table 9: Ordered Logistic Model of Total Opinions Stated on Groups Variable Coeff. Std. Err. p-value 22

R.’s Age -0.023 0.004 <0.001 Is R. Male? +0.571 0.126 <0.001 R.’s Education +0.121 0.026 <0.001 R.’s Income +0.008 0.025 0.754 Is R. White? +0.154 0.147 0.295 R.’s Political Interest -0.014 0.048 0.764 R.’s Partisanship +0.023 0.059 0.693 R.’s Media Consumption +0.127 0.064 0.047 R.’s # of Votes +0.348 0.089 <0.001 Is R. In Battleground? +0.006 0.126 0.962 Is R. Employed? -0.039 0.150 0.796 Is R. a Homemaker? -0.021 0.212 0.920 R.’s Contact -0.067 0.052 0.200 Does R. Have Network? +0.314 0.138 0.023 /cut1 -0.353 0.494 /cut2 +0.687 0.493 /cut3 +1.375 0.494 N=1357, LogLikelihood: -1338.2374, LRX2: 182.07, Sig-X2: <0.0001, PsuedoR2: 0.0637, Source: 2000 NES

Figure 1: Graph of Mean Percentage of Opinions Held Among All Four Types of Opinionation

Source: 2000 NES


Figure 2: Resources of Respondent by Whether or Not the Respondent Has a Network

Source: 2000 NES, All variables are displayed as a percent of maximum