This is a draft. Please contact the author to cite or for further details




старонка2/4
Дата канвертавання25.04.2016
Памер144.06 Kb.
1   2   3   4
Partisan Strength Treatment

Upon arriving in the survey center, group members each chose an individual computer terminal to privately complete an anonymous survey. The first question of the survey primed either a strong partisan affect (for those assigned to a strong partisan condition) or a weak partisan affect (for those assigned to a weak partisan condition). Modeled on the priming technique used by Lavine et al. (2012), the strong prime asks respondents to write some of the reasons for why they prefer their own political party and also to write some of the reasons why they dislike the opposing party. The weak prime asked respondents only to write some of the reasons they are dissatisfied with or critical of their own party (full text of primes are available in Appendix A). Immediately after this prime, the survey asked respondents to rank the importance of their party identification on a scale of 1 to 7, with 1 indicating ‘extremely unimportant’ and 7 indicating ‘extremely important.’ Indeed, upon analyzing these data, I could see that those assigned to the strong party prime ranked their party identification as significantly more important on a 7-point scale than did those assigned to the weak party prime (4.84 for strong partisans and 4.52 for weak partisans; p=0.06), demonstrating that this priming technique did have a statistically significant directional influence on party attachment, as was intended.3 Thus, the manipulation check confirms the success of the party strength manipulation.



Study Procedure

After completing their partisan strength prime, respondents – still seated at their individual computer terminals – read about the first policy issue: energy policy (specifically, how to decrease the price of gas and become more energy independent). Information regarding both the Democrats’ and Republicans’ perspectives on these issues was provided to all respondents and this information was identical across all conditions (full text of this issue is provided in Appendix B.) After reading about the issue at their individual computer terminals, respondents assigned to either a homogenous or heterogeneous group condition gathered with their fellow group members to discuss the issue and the alternative policy solutions about which they had all just read. Each group member stated his or her policy preference and then the group was given approximately five minutes to discuss the issue together. Group members were not explicitly told of each other’s partisan identities, but were only given as much information about the other members’ perspectives on politics as they themselves chose to reveal. Group members were allowed to refrain from commenting, if they so desired.

Following the group discussion, each member returned to his or her individual computer terminal, which was visually shielded from the others with an opaque glass barricade so as to ensure privacy. At the computers, each group member completed the duration of the survey, which included questions asking for their preferred policy solution and their perceived effectiveness of each of these policy solutions. Finally, they were asked to describe the level of ideological diversity they would want in a discussion group were they to discuss this issue with others.

Respondents assigned to non-social conditions endured the same procedure with one important exception -- they did not engage in any social interaction. Like in previous work (for e.g. Taber and Lodge 2006, Lavine et al. 2012), respondents merely read the information about the issues alone and then completed the duration of the survey in isolation.

After completing the energy policy survey, respondents then read about one more policy issue: health care (specifically, how to lower costs and widen coverage to more Americans; full text of this issue is provided in Appendix B). Once again, homogeneous and heterogeneous groups gathered to discuss this issue (non-social group conditions did not). Following the brief discussion, all respondents completed the duration of the survey privately, which included questions about health policy preference and the perceived effectiveness of each party’s policy, as well as a question asking the respondent to identify his preferred discussion group composition in the future.

Conditions

Respondents were thus treated to either a strong or weak partisanship prime, and to one of three group conditions: homogeneous, heterogeneous, or no group. This 2x3 design provides a 6-condition experimental design that allows me to distinguish the influence of the two types of social setting, in addition to a non-social setting, across both weak and strong partisans.



INSERT TABLE 1A

Table 1A displays the 2x2 factorial design and highlights the comparison groups for Hypothesis 1a and 1b – strong partisans versus weak partisans. Participants in Condition 1, Condition 2, and Condition 3 were given the strong partisanship prime, so as to induce a strong partisan attachment. Participants in Condition 4, Condition 5, and Condition 6 were given a weak partisanship prime, so as to induce a weak partisan attachment. Hypotheses 1a and 1b expect that, overall, respondents across all ‘strong partisanship’ conditions will express more partisan motivated reasoning than those in the ‘weak partisanship’ conditions. The outcome is operationalized by measuring (i) policy preference and (ii) perceived effectiveness of each party’s policy.



INSERT TABLE 1B

Table 1B illustrates this same experimental design, but now highlights the groups that are compared in order to test Hypothesis 2 and 3: respondents in non-social, homogeneous, and heterogeneous groups. Participants in Condition 1 and Condition 4 read about policy issues in isolation and then reported their preferences in an anonymous survey. Participants in Condition 2 and Condition 5 read about policy issues in isolation and then discussed the issues in an ideologically homogeneous group before reporting their preferences in an anonymous survey. Participants in Condition 3 and Condition 6 read about policy issues in isolation and then discussed the issues in an ideologically heterogeneous group before reporting their preferences in an anonymous survey. Hypotheses 2a and 2b expect that respondents in homogeneous groups (Conditions 2 and 5) will engage in more partisan motivated reasoning and thus will express a stronger preference for their own party’s policy and will deem the other party’s policy to be less effective, as compared to respondents in non-social settings (Conditions 1 and 4). Hypotheses 3a and 3b expect that respondents in heterogeneous groups (Condition 3 and Condition 6) will engage in less partisan motivated reasoning and will thus express more favorability toward the opponents and less bias regarding the effectiveness of the policies, as opposed to those in non-social settings.



INSERT TABLE 1C

Finally, Table 1C shows this same 2x3 factorial, but this time highlights the groups that are compared in order to test Hypothesis 4. Hypotheses 4a and 4b address the interaction of partisan strength and social setting. Whereas strong partisans are expected to engage in more partisan motivated reasoning, Hypothesis 4a expects that weak partisans in homogeneous social groups (Condition 4) will engage in more partisan motivated reasoning than will strong partisans in heterogeneous groups (Condition 3) and will thus express a greater preference for their own party’s policy. Hypothesis 4b predicts that weak partisans in homogeneous social groups (Condition 4) will be more biased in assessing the effectiveness of the two parties’ policies than will strong partisans in heterogeneous groups (Condition 3) .



Measurement of Effects

For both energy policy and health care, I used three sets of dependent variables to measure the extent to which partisan motivate reasoning is more evident among strong partisans versus weak partisans (Hypothesis 1), homogeneous groups versus non-social (Hypothesis 2) and heterogeneous groups versus non-social (Hypothesis 3), and weak partisans in homogeneous groups versus strong partisans in heterogeneous groups (Hypothesis 4). To measure policy preference, I first asked respondents to select their preferred policy choice. To measure partisan motivation in rating the effectiveness of each policy, I then asked respondents to rate their own party’s policy according to effectiveness on a 1-7 scale, and then to do the same for the opposing party’s policy. Finally, respondents selected their ideal discussion group according to ideological composition. This provides a test of downstream information acquisition. Those engaging in partisan motivated reasoning should prefer a group of all Democrats, whereas those with less partisan motivated reasoning should be more open to hearing information from the opposing side. All dependent variables are listed below in Table 2.



INSERT TABLE 2

Analyses & Results

Respondents first read about, discussed, and responded to questions regarding energy policy, and then they repeated this procedure for health care policy. Given the near identical results across both issues, I will present these results for both issues together. I will begin by presenting the results for policy preferences and perceived policy effectiveness, and then will turn to presenting results for the downstream effects regarding these two issues.


Hypothesis 1: Strong versus Weak Partisans

I begin with Hypotheses 1a and 1b that suggest, regardless of social interactions, we should see greater partisan motivated reasoning among strong partisans – that is preferences more consistent with one’s party when party strength is strong. Recall that all of the respondents I analyze identify with the Democratic Party and thus in every case the Republican Party is thus the “other”. Hypothesis 1a states that strong (i.e., less ambivalent) partisans will express a stronger preference for attitudinally congruent information – i.e. the Democratic Party’s policy. Figure 1a displays the results across all conditions for strong and weak partisans (without distinguishing by social setting)



INSERT FIGURE 1A

Figure 1a displays strong and weak partisans’ preferences for energy policy on the left side and their preferences for health care policy on the right side. The darker blue point indicates strong partisans’ preferences and the lighter blue point indicates weak partisans’ preferences. Surrounding each point are two horizontal lines indicating the 95% confidence interval. Along the y-axis, the response scale ranges from the Republican policy at the lowest point (“Only invest in drilling for oil” as the energy policy solution and “Only increase competition” as the health care solution) and the Democratic policy at the highest point (“Only invest in alternative fuels” as the energy policy solution and “Only expand subsidies” as the health care solution). The mid-point on the scale represents equally prioritizing both parties’ policies.

On the left side, Figure 1a indicates that, when it comes to energy policy, strong partisans are significantly (p<0.01) more in favor of investing in alternative fuels (5.41) as opposed to drilling for oil, as compared to weak partisans (4.89). On the right side of the graph, we see that, with respect to health care policy, strong partisans are again more in favor of their own party’s policy. Strong partisans report a 4.92 on the 7-point scale, which is significantly (p<0.01) greater than weak partisans (4.54).

Hypothesis 1b expects that strong (i.e., less ambivalent) partisans will perceive the Democratic party’s policy to be more effective, and the Republican party’s policy to be less effective, than will weak partisans. Figure 1b displays these accompanying results for both energy policy and health care.



INSERT FIGURE 1B

On the left side of Figure 1b, we see how strong and weak partisans rate the effectiveness of the Democratic policy of alternative fuels and the Republican policy of drilling for oil. Here we can see that strong partisans rate alternative fuels as more effective (5.82) than do weak partisans (5.66).This difference is directionally in line with what Hypothesis 1b predicts, though it does not reach conventional levels of statistical significance (p=0.17). This may be due to the widespread popularity of this particular policy among both strong and weak partisans. When it comes to drilling for oil, strong partisans rate this Republican policy at a low 2.86, which is significantly (p<0.01) less effective than the perceived effectiveness of this Republican policy among weak partisans, who rate it at 3.32.

On the right side of Figure 1b, we see this same pattern when it comes to health care policy. Strong partisans rate the Democratic policy of expanding government subsidies at 5.45, which is significantly (p<0.05) more effective than the weak partisans’ rating of 5.15. Strong partisans perceive the Republicans’ policy of increasing competition as significantly (p<0.05) less effective (3.99) as compared to weak partisans (4.34). Overall, I find support for Hypotheses 1a and 1b.
Hypotheses 2 and 3: Homogeneous, Non-Social, and Heterogeneous Settings

I next turn to Hypotheses 2 and 3, which expects differences in partisan motivated reasoning according to social setting. Here I analyze average responses only by social setting, without taking partisan strength into account. Hypothesis 2 states that respondents in homogeneous groups (i.e., groups that include only fellow partisans) will engage in more partisan motivated reasoning as opposed to those in non-social settings. This will result in more biased preferences for the Democratic party’s policy (Hypothesis 2a) and a perception that the Democratic party’s policy is more effective, and the Republican party’s policy is less effective (Hypothesis 2b).

Hypothesis 3 states that respondents in heterogeneous groups (i.e., groups that include other partisans) will engage in less partisan motivated reasoning as opposed to those in non-social setting. This will result in a less biased preference for the Democratic party’s policy (Hypothesis 3a) and a perception that the Republican party’s policy is more effective, and the Democratic party’s policy is less effective (Hypothesis 3b).

INSERT FIGURE 2-3A

Figure 2-3A displays responses for the energy policy options on the left side and the health care policy options on the right side. Here, the dark blue dot indicates respondents in homogeneous (all Democrat) groups. The white dot indicates respondents without groups (non-social settings). The purple dot indicates respondents in heterogeneous groups (ie. a group of 4 Democrats and 4 Republicans). Again the horizontal bars surrounding each dot indicate the 95% confidence interval.

Looking at the left side of the figure, we can see that those in homogeneous groups are most biased in favor of the Democratic policy (5.67). Those in non-social settings are significantly (p<0.01) lower on the 7-point scale (5.09). Respondents in heterogeneous groups are even closer to the midpoint of the scale. Compared to those in non-social settings, respondents in heterogeneous groups are significantly (p<0.01) lower on the scale (4.13).

Regarding health care policy, we see this same strong pattern. Those in homogeneous groups are closest to the Democratic policy (5.06), which is significantly (p<0.01) higher than those in non-social groups (4.74). There is a significant (p<0.01) difference as well between those in non-social groups and those in heterogeneous groups who rate their preferred policy at 4.03. Across both issues, Hypotheses 2a and 3a are thus both strongly supported. Partisan motivated reasoning is thus clearly contingent upon social setting. In the company of heterogeneous others, motivated reasoning is drastically tempered, as compared to settings in which respondents are learning information in isolation. When respondents are among like-minded co-partisans, on the other hand, we can see that partisan motivated reasoning is considerably enhanced.



INSERT FIGURE 2-3B

Turning to Hypotheses 2b and 3b, I report Figure 2-3B, which displays perceived effectiveness of the Democratic policy and the Republican policy among those in homogeneous groups (the blue dot on the figure), non-social groups (the white dot), and heterogeneous groups (the purple dot). On the left side of the figure, we can see the perceived effectiveness of the Democratic solution to energy policy (alternative fuels) at the top and the Republican solution to energy policy (drilling for oil) at the bottom. From the figure, we can see that respondents in homogenous groups perceived alternative fuels to be significantly (p<0.01) more effective (6.35) than do those in nonsocial groups (5.59). These perceive drilling for oil to be significantly (p<0.05) less effective (2.75) than do those in non-social groups (3.12). Hypothesis 2b is thus supported.

Hypothesis 3b expects that respondents in heterogeneous groups will engage in less partisan motivated reasoning and will therefore perceive their own party’s policy to be less effective and the opponent’s to be more effective, as opposed to those in non-social settings. With respect to energy policy, the left side of Figure 2-3B displays that this is indeed the case. Respondents in heterogeneous groups rate alternative fuels to be significantly (p<0.01) less effective (4.74) than those in non-social settings, and they rate drilling for oil to be significantly (p<0.01) more effective (3.75) that do those in non-social settings.

Hypothesis 2b and 3b are also strong supported when it comes to health care policy. The right side of Figure 2-3B displays the perceived effectiveness of the Democratic solution to health care policy (expanding government subsidies) and of the Republican solution (increasing competition among insurers). Again we see that respondents in homogeneous groups rate the Democratic policy to be significantly (p<0.01) more effective (5.57) than do those in non-social groups (5.24). Respondents in heterogeneous groups rate the Democratic policy to be significantly (p<0.05) less effective (4.82) than do those in non-social settings. Regarding the Republican policy of competition, respondents in homogeneous groups perceive it as significantly (p<0.10) less effective (3.79) than do those in non-social settings (4.14) and those in heterogeneous settings rate it as significantly (p<0.01) more effective (4.98).


Hypothesis 4: Weak Partisans in Homogeneous Settings versus Strong Partisans in Heterogeneous Settings
Finally, I turn to Hypothesis 4 to test the interaction between partisan strength and social settings. Hypothesis 4a predicts that weak partisans in homogeneous networks will express a strong preference for attitudinally congruent information – i.e. their own party’s policy – than will strong partisans in heterogeneous networks (Hypothesis 4a). Figure 4a displays these results. We see the weak partisans in homogeneous groups indicated by the light blue dot. The strong partisans in heterogeneous groups are indicated by the dark blue dot. Again the horizontal bars sandwiching the dots indicate the 95% confidence interval.

INSERT FIGURE 4A

On the left side are policy preferences for energy policy. As indicated in the figure, here we see a flip in the relative influence of partisan motivated reasoning between strong and weak partisans, Whereas strong partisans generally showed themselves to have the greater bias toward their own party’s policy, here was can see that weak partisans in homogeneous groups actually report a stronger bias toward the Democratic policy of investing in alternative fuels than do strong partisans in heterogeneous groups. Here, where we take social setting into account, we can see that the weak partisans in homogeneous settings are significantly (p<0.01) more likely to favor an exclusively Democratic approach to energy policy (5.44), and the strong partisans in heterogeneous groups report a much more bipartisan preference (4.5).

This novel finding holds across the health care issue as well. Whereas weak partisans show less of a bias towards the Democratic policy when social setting is not taken into account, when we examine weak partisans in homogeneous settings, we see a stark bias (4.89) that is significantly (p<0.05) stronger than the preference shown among strong partisans in heterogeneous settings (4.35). Hypothesis 4a is thus strongly supported.

Whereas strong partisans tend to engage in the most partisan motivated reasoning, which leads to perceiving one’s own party’s policy as the most effective and the opponent’s policy as least effective, Hypothesis 4 expects that weak partisans in homogeneous networks will engage in strong partisan motivated reasoning as a result of their social surrounding. Strong partisans in heterogeneous networks, on the other hand, will engage in less partisan motivated reasoning, thanks to their ideologically diverse group. Figure 4b displays the accompanying results.



INSERT FIGURE 4B

On the left side, we see that weak partisans in homogeneous groups (light blue), in fact, rate the Democratic energy policy solution as significantly (p<0.01) more effective (6.31), than do strong partisans in heterogeneous networks (dark blue dot; 4.93). Conversely, strong partisans in heterogeneous networks rate drilling for oil as significantly (p<0.10) more effective (3.42) than do weak partisans in homogenous networks (2.95).

The right side of the figure shows that these findings hold regarding health care policy. Here we see that weak partisans in homogeneous groups rate the Democratic health care solution of subsidies as significantly (p<0.05) more effective (5.47) than do strong partisans in ideologically heterogeneous networks (4.94). When it comes to the Republican policy of increasing competition, weak partisans in homogeneous networks rate it as significantly (p<0.05) less effective (3.95) as compared to strong partisans in heterogeneous setting (4.84). Hypothesis 4a and 4b are thus supported, indicating that the influence of social settings appears to exert a more powerful effect over partisan motivated reasoning than does partisan strength alone. Although responses were collected in isolation and with an assurance of anonymity, it is indeed still possible that the mechanism at play is one of social pressure, a force that has been shown to boost turnout among voters (Green, Gerber, and Larimer 2008). Alternatively, it could be that the discussion within the group is, in fact, has a weakening effect on the partisan identity. The precise mechanism at work is surely a question that deserves further scholarly attention.
Downstream Effects

My final dependent variable measures the potential downstream effects of the treatments. How do partisanship strength, social setting, and the interaction of the two affect respondents’ preferences for ideological composition of future discussion groups? Hypothesis 1 expects that strong partisans will be more likely to prefer a group with more like-minded respondents (Democrats). Figure 5 displays the results, which indicate that indeed strong partisans prefer a group with significantly (p<0.01) more like-minded respondents (i.e. fellow Democrats) to discuss energy policy (on the left side of the figure) and that strong partisans also prefer a group with significantly (p<0.01) more like-minded respondents to discuss health care policy (on the right side of the graph).



INSERT FIGURE 5

Hypothesis 2 expects that those in homogeneous groups (across both partisan groups) will prefer more of a like-minded discussion group in the future, as compared to those in non-social settings and Hypothesis 3 states that respondents in heterogeneous groups will prefer a more diverse discussion group. Figure 6 displays these results, which indicate that this is indeed the case. The left side of Figure 6 illustrates the fact that, regarding energy policy discussions, respondents in homogeneous groups prefer a significantly (p<0.01) more like-minded group (4.57) than do those in non-social settings (4.23). Respondents in heterogeneous groups prefer a significantly (p<0.01) more diverse group in the future (3.63).


1   2   3   4


База данных защищена авторским правом ©shkola.of.by 2016
звярнуцца да адміністрацыі

    Галоўная старонка