Menu
Home
Log in / Register
 
Home arrow Economics arrow Competing motives in the partisan mind : how loyalty and responsiveness shape party identification and democracy
Source

DUTY, STAKES, AND PARTY IDENTIFICATION CHANGE

An experiment was designed to manipulate the saliency of the norm of political autonomy versus the policy stakes of an upcoming election. The experiment complements the previous survey analysis by providing leverage over causality. Moreover, by measuring party identification before and after the experimental manipulation, it was possible to observe actual party identification change.

If party identification were found to be responsive, this would partially support revisionist claims. However, whereas most revisionist models portray party identification as serving an instrumental function in the quest for policy benefits, the dual motivations model suggests that the incentive to appear pragmatic and unbiased is likely to be the real driver of responsiveness. In other words, partisan updating is thought to serve an expressive function. When partisans change their identity to reflect their attitudes, they do so because they want to appear pragmatic and politically autonomous, not because they desire policy benefits.

The implications of this distinction are highly relevant to our understanding of democratic accountability. Scholars often draw analogies between democracy and markets. Voters are thought to demand policy from parties, just as consumers demand products from firms. To adapt Adam Smith's (1986) famous quote, it is not from the benevolence of the Republican or Democratic voter that we expect parties to be held accountable, but from their regard to their own interest. However, if partisan responsiveness is not driven by policy concerns but by civic duty, then benevolence is precisely what is needed to attain accountability.

Method

A sample composed of 1098 adult partisans and partisan "leaners" from across the United States was obtained over the Internet though YouGov/Polimetrix between July 29, 2008, and August 3, 2008. YouGov/Polimetrix matched subjects down to the known marginals of the general population of the United States on gender, age, race, education, and political interest. However, in this sample, partisan "leaners" were under-represented in the sample by a substantial margin relative to strong and weak partisans.[1] As in previous chapters,

“pure" Independents were excluded from analysis, because the purpose of the study was to determine what conditions lead partisans to defect from their party. Since “pure” Independents claim no partisan allegiance, it cannot be determined whether movement in their party identification constitutes movement toward or away from a favored party.

Procedure

This experiment employed a three-celled design in which subjects were primed to consider either the policy stakes of an upcoming election (i.e., instrumental concerns), the norms of civic duty and pragmatism (i.e., expressive concerns), or neither of the above. Both priming treatments were carried by a survey question administered near the beginning of the study. In all three conditions, subjects read the following introduction:

“Experts predict that nearly 200 million people will vote in the November election. In addition to the presidential race, they will be casting votes for representatives and senators who will represent them in Congress.

Most voters think of themselves as either a Republican or a Democrat, and most candidates are affiliated with one of those two parties.”

I n the control condition, respondents were not asked to comment. In the policy stakes condition, this introduction was followed by an open-ended question:

“Think about what's at stake in the upcoming election. Do you believe the country and you personally will be seriously affected by which party wins this election? Please explain your answer in a few sentences.”

In the duty priming condition, participants were asked:

“Think about what it means to be a good citizen. Do citizens have a duty to consider the issues, or is it okay to just vote based on party? Please explain your answer in a few sentences.”

Immediately after the prime, participants viewed a nonpartisan electronic voter guide outlining differences in the platforms of the Republican and Democratic parties. The voter guide discussed the economy, the war in Iraq/defense, education, healthcare, illegal immigration, abortion, and gun

“leaners.” For the purposes of analysis, the two subsamples were combined. Neither the sampling issue nor the additional wording affected random assignment. Therefore, any differences that arose between experimental conditions can be attributed only to the priming stimuli and to no other factor.

control. The parties' positions were presented side by side in succinct statements (see Appendix Figure 6.1).

Measures

A variable corresponding to past party identification strength was created by simply folding the standard seven-point party identification measure in half. The new scale ran from 0 to 3. Strong Democrats as well as strong Republicans were coded as 3, weak partisans were coded as 2, leaners were coded as 1, and pure Independents were coded as 0. Because YouGov/Polimetrix uses sample matching to obtain nationally representative samples, they have party identification on record for everyone in their respondent pool. Thus, it was possible to avoid problems that would obviously result if party identification were obtained immediately prior to the treatment.

The primary dependent variable of interest in this study was current party identification strength. This variable was coded identically to past party identification strength except that those partisans who crossed over from one party to the other from pre-treatment to post-treatment were coded as having zero strength.

The measure of i ssue distance was based on subjects' assessments of the parties' positions relative to their own positions across a number of issues preceding party identification in the post-treatment instrument. Participants were asked to take positions on each of the issues discussed in the voter guide. They were then asked to place the Republican Party and Democratic Party on identical seven-point scales. These items were either taken directly from the ANES or modeled after ANES questions. (The actual wording of the measures appears in Appendix Figure 6.1.) These items were used to construct measures of each respondent's distance from his or her own (pre-test) party on each issue dimension (relative to the opposition party). Later in the survey, subjects were asked to rate the importance of each of these issues on a seven-point scale ranging from “not that important” to “extremely important.” Each issue position was then weighted by the importance rating (rescaled to run from 0 to 1) assigned to it by the respondent.[2] These weighted values were then added together to form a single summary measure of issue distance and rescaled to run from -1 to 1.[3] Individuals whose weighted issue positions were equidistant from the two parties received a value of zero. Those who favored their own party's positions relative to the opposition party's positions received positive values. Those who favored the opposition party's positions received negative values.

Dummy variables were created to represent exposure to the duty prime condition or policy stakes prime condition. Subjects assigned to a treatment condition received a value of “1”; all others received a value of “0.”

Results

I began by examining subjects' responses to the two priming questions. The experiment was built on the assumption that subjects assigned to the stakes condition would say that there was, indeed, something important at stake in the 2008 election. In the norms condition, they were expected to endorse the norm of voting on the issues as opposed to simply voting for one's party. As expected, 82.6% of subjects assigned to the stakes priming condition and 85.0% of respondents assigned to the norms priming condition responded in the expected manner.

Next, I checked random assignment to determine whether any chance associations existed between the treatment and any other variables related to party identification strength. Although no significant differences in previous party identification strength or political sophistication emerged between experimental conditions, differences in age and Democratic identification appear to have arisen between conditions by chance.1 1 More specifically, before treatment, those assigned to the policy stakes condition showed a higher propensity to report party identity on the Democratic side of the scale (mean (M) = .585, standard deviation (SD) = .026) than those assigned to the control condition (M = .484, SD = 0.251; p < .01).12 In addition, subjects assigned to the duty prime were older on average (M = 49.21, SD = 14.62) than subjects assigned to either the control group (M = 47.15, SD = 15.64; p < .05) or the policy stakes prime (M = 46.71, SD = 14.71; p < .05). These variables were therefore included as [4] [5] [6] [7] [8] [9] [10]

controls in the analyses that follows. Predicted values were estimated with age set to its mean, party identification strength set to “strong,” and Democrat identification set to zero.

Past party identification and current party identification are cross-tabulated in Table 6.2. Although there was a clear tendency toward stability, there does appear to be variance in party identification to explain. However, the question is, of course, whether this variation constitutes experimentally induced systematic variation or just random measurement error (Green & Palmquist 1990, 1994; Green, Palmquist, & Schickler, 2002).

The first step was to examine the effects of both primes on current party identification while controlling for past party identification. I had hypothesized that the duty prime would weaken party identification whereas the stakes prime would have little effect. The probability of maintaining a strong party identity across experimental conditions is displayed by condition in Figure 6.2. As in previous chapters, these estimates were obtained with the use of ordered probit regression to account for potentially inconsistent intervals in party identification strength. Consistent with expectations, the results in Table 6.3 show that those exposed to the duty prime were less likely to maintain a strong party identity than those assigned to the control group (p < .05). On the other hand, the stakes prime had no effect on party identification whatsoever.

Although these results provide compelling evidence in favor of the duty and stakes hypotheses, it is possible that the duty prime simply encouraged closet partisanship. To be more specific, participants may have simply reported a weaker party identity in order to avoid the appearance of partisan

TABLE 6.2. CROSS-TABULATION OF PAST PARTY IDENTITY AND CURRENT PARTY IDENTITY

Current Party Identification

Past Party Identification

Strong

Dem

Weak

Dem

Leaning

Dem

Indep

Leaning

Rep

Weak

Rep

Strong

Rep

Strong Dem

81.2%

14.6%

2.1%

0.9%

0.3%

0.0%

0.9%

Weak Dem

7.6%

74.1%

10.6%

5.3%

1.2%

1.2%

0.0%

Lean Dem

5.6%

16.7%

66.7%

11.1%

0.0%

0.0%

0.0%

Indep

1.4%

2.8%

12.7%

67.6%

12.7%

2.8%

0.0%

Lean Rep

0.0%

0.0%

1.7%

15.5%

65.5%

6.9%

10.3%

Weak Rep

0.0%

2.7%

2.7%

5.9%

13.9%

68.4%

6.4%

Strong Rep

0.0%

0.0%

0.0%

1.1%

4.5%

10.4%

84.0%

Note: Percentages are calculated so that rows sum to 100%.

The figure illustrates the predicted probability of maintaining a strong party identity from pre-treatment to post-treatment

Figure 6.2 The figure illustrates the predicted probability of maintaining a strong party identity from pre-treatment to post-treatment. Predicted probabilities were based on ordered probit regression. For the purpose of prediction, past party identification strength was set at 3 (strong), Democrat identification at 0, and age at 47.54.

bias. They may not have actually brought their identity into alignment with their issue positions.

To determine whether the duty prime truly increased responsiveness motivation rather than merely eliciting closet partisanship, I examined whether issue positions moderated the effect of the duty prime. Although issue positions were measured after the stimulus, the results show that issue positions were entirely unaffected by the experimental manipulation. Therefore, they can safely be treated as exogenous for the purpose of testing the moderation prediction. To do this, the issue distance variable was interacted with each of the two treatment variables. The results presented in Table 6.4 indicate that the duty prime, but not the stakes prime, moderated the influence of issue positions on party identification. The results displayed in Figure 6.3 suggest that exposure to the duty prime led partisans to bring their identity into line with their issue positions. Those partisans whose issue positions were close to the positions of their own party, relative to the other party, tended to maintain their party identity regardless of condition. However, substantial differences emerged between conditions among those people who tended to disagree with their own party on the issues. Subjects in the stakes condition and in the control group who expressed more agreement with the opposition party than with their own party across a number of issues still managed to maintain a stable party identity approximately one-third of the time. This probability dropped to approximately one twentieth in the duty priming condition. In sum, responsiveness motivation was increased by priming considerations of

TABLE 6.3. EFFECTS OF DUTY PRIME AND STAKES PRIME ON PARTY IDENTIFICATION

Party Identification Strength B

  • (SE)
  • (N =1083)

Duty Prime

  • -.189**
  • (.089)

Stakes Prime

.010

(.091)

Party Identification Strength (f-1)

  • 1.38***
  • (.059)

Democrat (f-1)

.112

(.075)

Age

.007***

(.003)

Cut 1

  • 1.43
  • (.186)

Cut 2

  • 2.40
  • (.188)

Cut 3

  • 3.79
  • (.206)

Note: Results are based on ordered probit regression. The coefficients (B) and standard errors (SE) reported in the table were obtained via ordered probit models.

***p < .01, **p < .05, *p <.10

duty and the virtue of pragmatism over partisanship. This led partisans to bring their identity into alignment with their issue positions.

On one hand, these results strongly support revisionist models of party identification by showing that, under at least some conditions, partisans will change their identity to reflect their issue positions. Thus, the “unmoved mover” portrayal of party identification appears to be a mischaracterization. On the other hand, these results provide a striking contrast with the general assumption that party identification is instrumental to the attainment of policy benefits. From these results, it appears that consideration of policy stakes has little or no influence on party identification. Instead, party identification change is driven by the incentive to express one's civic duty and pragmatism. Although the instrumentality of party identification cannot be

TABLE 6.4. EFFECT OF ISSUE POSITIONS AND PARTY IDENTIFICATION MODERATED BY DUTY AND STAKES

Party Identification Strength B

  • (SE)
  • (N = 1082)

Duty Prime

  • -.420
  • (.302)

Stakes Prime

  • -.238
  • (.297)

Issue Distance

Duty Prime*Issue Distance

Stakes Prime*Issue Distance

- .153 (.421)

Party Identification Strength (f-1)

Duty*Party Identification Strength (f-1)

- .017 (.126)

Stakes*Party Identification Strength (f-1)

.124

(.126)

Democrat (f-1)

.147**

(.077)

Age

.004*

(.003)

Cut 1

1.31 (.247)

Cut 2

2.34 (.248)

Cut 3

3.81 (.263)

Note: Results are based on ordered probit regression. The coefficients (B) and standard errors (SE) reported in the table were obtained via ordered probit models.

The figure illustrates the probability of maintaining a strong party identification from pre-treatment to post-treatment given one's issue positions

Figure 6.3 The figure illustrates the probability of maintaining a strong party identification from pre-treatment to post-treatment given one's issue positions. Predicted probabilities were based on ordered probit regression. For the purpose of prediction, past party identification strength was set at 3 (strong), Democrat identification at 0, age at 47.54, and word count at 14.35. The maximum negative issue distance was -.518 and the maximum positive issue distance was .857. These were the most extreme observed values in the data. Neutral issue distance was set at 0.

substantial effect of the duty prime suggests that partisans had plenty of room to move. Clearly, party identification was not in perfect alignment with issue positions coming into the study, and this relationship only increased in the duty priming condition—leading to weaker party identification.

Although individuals are motivated to see themselves as loyal partisans (and to avoid the cost of partisan disloyalty), they are also motivated to believe that their identity is issue-based and not merely rooted in partisan bias. The results reported here paint the image of a voter who is concerned with avoiding the appearance of partisan bias but not so concerned with maintaining the functionality of her party identity to facilitate the attainment of policy benefits.[14]

Discussion

This chapter asked the question, “To whatever degree partisans update their identity, what is their motivation for doing so?” Although a great number of studies have examined the degree to which party identification changes, few have given attention to the motivations underlying party identification change. It is generally assumed that, because citizens have a stake in political outcomes, they will want to identify with the party that offers them the most policy benefits. However, this chapter demonstrates that, if there is any psychological cost associated with party identification change, expected policy benefits are unlikely to be great enough to outweigh this cost.

Because the American voter's motivation to attain policy benefits may not be sufficient to overcome her motivation to remain loyal to her party, we should, by implication, be wary of assuming that democratic accountability is guided by the invisible hand. Ironically, it appears that we may not be able to count on voters to act in their own policy interests (see also Caplan, 2007). Partisans appear motivated to act responsively—like good citizens— only when norms of civic duty are salient to them and not simply when their personal interests are at stake.

This also means that party identification may not function as an efficient shortcut for voters. For instance, consider a purely instrumental model in which party identification develops out of the need to make accurate political evaluations while minimizing information costs (Shively, 1979). For party identification to function efficiently, allowing relatively uninformed individuals to vote as if they had objectively weighed the available information, partisans would need to update their identity to reflect their issue positions. These results suggest that policy incentives are insufficient to offset the cost of partisan updating. Instead, it appears that the heuristic efficiency of party identification depends on instilling norms of civic duty through cultural socialization and civic education. Without such norms, partisans would have little incentive to hold their party accountable for its policies. To the degree that voters rely on their party identification to make decisions, these decisions would not accurately reflect their true preferences, and short-term democratic accountability would be threatened.

With this said, my intention has not been to argue that party identification is devoid of heuristic value. These results merely suggest that we should carefully consider how efficiently party identification actually operates as an information shortcut (also see Bartels, 1996). If there were no cost to updating one's party identity, we might safely assume that identification with responsible parties promotes democratic accountability by enabling relatively uninformed citizens to make sense out of a complicated political landscape. However, given the psychological cost of partisan updating, norms of civic duty may be the only thing preventing party identification from hopelessly biasing political assessments and undermining citizens' incentive to hold parties accountable. In short, good citizenship—at least as it pertains to party identification—is rooted in norms of civic duty and not in mere self-interest. Therefore, the promotion of these norms is vital if we want partisans to act like good citizens who hold parties accountable rather fans who support their team unconditionally.

This page intentionally left blank

  • [1] The full sample was composed of two subsamples of 600 people surveyedfrom July 20 through July 31, 2008 and from August 1 through August 3, 2008,respectively. After the first sample was collected, it was realized that partisans"leaners" had mistakenly been left out of the sample. There was also concernregarding the length of subjects' responses to the stimulus. Therefore, a sentencewas added to both treatment conditions asking participants to "Please explainyour answer in a few sentences." The study was put back into the field thefollowing day, and a new sample was collected—this time including partisan
  • [2] 9 . Carsey and Layman (2006) argued that partisans update their identity to reflecttheir issue positions when they are aware of differences between the two partiesand when they consider the issue to be important. However, on issues they considerless important, they update their issue positions to reflect their party identity.
  • [3] T-tests showed no significant differences in issue positions between conditions.Principal axis factor analysis of the seven issues resulted in a unidimensionalsolution with an Eigenvalue of 4.40.
  • [4] 11. Political sophistication was measured with the use of a seven-item battery.Therefore, the political sophistication variable ran from 0 (none correct) to
  • [5] (all correct). This battery contained a variety of multiple choice questions
  • [6] about political figures and institutions. Respondents were asked to identify
  • [7] the jobs filled by Nancy Pelosi, John Roberts, Dana Perino, and Gordon Brown.Participants are also asked how many votes it takes to override a veto, which
  • [8] branch of government has the power to determine whether a law is constitu
  • [9] tional, and which branch has the power of the purse.
  • [10] . Democratic identification was a dummy variable.
  • [11] p < .01, **p < .05, *p < .10 ruled out on the basis of a single null finding, the correspondence between thedual motivations theory and this result is noteworthy. A skeptic might arguethat individuals' party identities already reflected their concerns over policystakes coming into the experiment, so there was no room for partisans to bringtheir identities into closer alignment with their issue positions. However, the
  • [12] p < .01, **p < .05, *p < .10 ruled out on the basis of a single null finding, the correspondence between thedual motivations theory and this result is noteworthy. A skeptic might arguethat individuals' party identities already reflected their concerns over policystakes coming into the experiment, so there was no room for partisans to bringtheir identities into closer alignment with their issue positions. However, the
  • [13] p < .01, **p < .05, *p < .10 ruled out on the basis of a single null finding, the correspondence between thedual motivations theory and this result is noteworthy. A skeptic might arguethat individuals' party identities already reflected their concerns over policystakes coming into the experiment, so there was no room for partisans to bringtheir identities into closer alignment with their issue positions. However, the
  • [14] . The parallel between these findings and those in the literature on racial prejudice is worth mentioning. Whereas racial attitudes can have a substantial influence on individuals' policy positions (Devine, 1989; Kinder & Sanders, 1996;Mendelberg, 2001; Valentino, Hutchings, & White, 2002), these influences arediminished when racial implications are made explicit, because they conflictwith norms of egalitarianism (Devine; Mendelberg). In the cases of both partyidentification and racial prejudice, the influence of group bias appears to bereduced when conflict between biases and norms is made salient.
 
Source
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
 
Subjects
Accounting
Business & Finance
Communication
Computer Science
Economics
Education
Engineering
Environment
Geography
Health
History
Language & Literature
Law
Management
Marketing
Mathematics
Political science
Philosophy
Psychology
Religion
Sociology
Travel