the selection of the network using the best methodological tools available, contextual effects do impact an individual’s vote choice.
may be an additional variable which not only determined whether or not they would become discussants but also which determines whether or not there is agreement or disagreement in the discussion network. However, the Huckfeldt et al. (2004) results are clear — there is correlation within discussion groups’ political preferences. The immediate consequence of a politically diverse discussion network on an individual voter’s ideological preferences is not, however, well understood.
A quick look at the breakdown of data available in the 2000 ANES demonstrates that indeed the respondents do live in politically diverse communities; that is, there are many individuals who voted for a Democratic presidential candidate but who live in a Republican congressional district, or who likewise voted for a Democratic presidential candidate but whose principle discussion network is composed of Repub- licans.2 Table 8.1 describes the percentages and raw numbers available in the data-set which separate the majority political party in either the respondent’s congressional district or discussion network and the vote into all four possible categories. The pres- ence of individuals whose vote is different from the majority party in their network ensures that there are enough of these cases to study.
Table 8.1 Goes Here
The puzzle begins when the data is tabulated both by vote, by respondent ide- ology, and by social context. Table 8.2 describes the number of instances where a self-identified liberal voter who lives in a majority Republican congressional district has chosen to vote for a Republican candidate or where a self-identified conservative voter who has in a majority Democratic discussion network has chosen to vote for a Democratic candidate. These percentages are significantly higher than if the individ- ual lived in, or had a network full of, individuals with similar partisanship. Looking at these percentages alone it seems possible that in fact voters are influenced by their context. Note that, for example, while only 1% of those voters who live in majority Democratic congressional districts and and who self-identified as liberals voted for
2Individuals are removed from the data-set if their discussion network is evenly split between Republicans and Democrats as it is then impossible to determine the majority party. Thus 126 individuals were dropped from the analyses.
the Republican presidential candidate, 6.55% of those voters who live in a major- ity Republican congressional district and who self-identified as liberals voted for a Republican presidential candidate. To what extent are voters influenced by others within their congressional district? Within their immediate social circle? Are their friends in fact imbedded within their congressional district?
Table 8.2 Goes Here
These results are presented in Figure 8.1 for closer inspection. The height of each of these bars indicates the percentage of individuals who fell into this category who voted for Bush. The categories are described at the bottom of each bar — the ideology of the respondent is listed first (either Liberal or Conservative) followed by the majority party category (Democratic or Republican). The first bar in each category denotes the congressional district effect, and the second bar denotes the discussion network effect. Note here that if there were no effect of congressional district or discussion network, we would anticipate the height of the bars on the far left, which indicate the percentage of individuals who voted for Bush but who are self-identified as liberal and live in a majority-Democratic congressional district (light blue) or have a majority-Democratic discussion network (purple) should be the same height as those immediately to their right, who are also self-identified as liberal but live in a majority-Republican congressional district or who have a majority-Republican discussion network. However, we observe a large shift in the percentage of Bush voters once the congressional district or discussion network changes majority party. This effect is identical for respondents who are self-identified conservative voters.
Figure 8.1 Goes Here
These results are next tabulated by Bush vote while examining only the majority party discussion network in Figure 8.2. Here again the height of each bar denotes the percentage of individuals who fall into that category who voted for Bush. Note that the greatest percentage of Bush votes come from individuals who report having an all-Republican discussion network. Individuals who reported having a mixed-party
discussion network fall into the middle category, and individuals who report having an all-Democratic discussion network fall into the lowest category.
Figure 8.2 Goes Here
All of this suggests that individuals are increasingly voting like those around them, and that the social context explanation for vote choice is potentially appropriate given the tabulations of the data seen above. One possibility is that voters are increasingly able to seek out like-minded individuals as it becomes increasingly easy to maintain connections across larger geographic distances. Additionally, if it is the case that a voter’s network is persuasive, then it is important to understand the role of diversification in networks for ideological stability.
In February 2004 the Pew Internet and American Life Project conducted a sur- vey on social ties in America. In the report on these survey results they find that
“traditional orientation to neighborhood- and village-based groups is moving towards communities that are oriented around geographically dispersed social networks” (Pew Internet and American Life Project 2006, 2). The Pew results also indicate that re- spondents were likely to get advice from people online. Many claim that Americans are increasingly “bowling alone” (Putnam 2000) and replacing neighborhood social interactions with online social interactions (Sunstein 2001). If the internet and other technological changes (such as nationwide long distance cellular phone plans) begin to change with whom voters interact, this is likely to affect their voting choices. In particular, if respondents are more likely to have discussion partners who are geo- graphically distant, then knowledge of the respondent’s geography is not sufficient to control for the information she would receive from other voters. It is then crucial to know something about the structure of her particular social network if we believe that voter-to-voter communication impacts voters choices. Most survey data fails to include questions regarding the respondent’s social ties. However, this seems like a likely place that a voter may actually be influenced.
Huckfeldt and Sprague (1987, 1988) and Huckfeldt, Sprague, and Levine (2000) did conduct surveys on political discussion partners. These surveys asked an initial
respondent a series of questions and also asked each respondent for the contact in- formation of their discussion partners. The researchers then contacted the discussion partners and administered the same survey. This type of survey is referred to as a “snowball” survey. The existing snowball surveys on political discussion partners are limited, both in terms of geographic scope and by being conducted for only a few specific election cycles. Furthermore, only non-relatives are considered in the scope of political discussion partners. The General Social Survey (GSS) did complete a social network battery in 1987 and 1988, but there is little to identify the network political ideology or the respondent’s political ideology. Finally, the 2004 ANES asked respon- dents if they ever discuss politics with their family or friends, and the frequency with which that discussion occurs. However, these questions fail to include the partisan nature of the discussants which plays a crucial role in determining the respondents’
preferences and information set.
Both the fields of political science and sociology have benefited from inclusion of questions attempting to analyze details of the respondent’s political network. The GSS social network battery produced innovations in the analysis of network survey data (Marsden 1990), an increased understanding of the structure of political discus- sion networks (Marsden 1987), and an increased understanding of the relationship between organizational affiliations and network density (Liedka 1991). It provided data from which to analyze with whom people discuss politics, the frequency of that discussion, and the impact of political discussion on political participation (Straits 1991; Knoke 1990). The ANES social network battery lead to an increased under- standing of the role between disagreement and social ties (Huckfeldt, Johnson, and Sprague 2004).