Ideology, Not Affect: What Americans Want from Political Representation

Article Summary

While a large body of research shows Americans hold increasingly negative views of non-copartisans, it is unclear how those views influence evaluations of representatives relative to policy congruence or constituency service. More simply, while partisans may deeply dislike the opposing party, they may look for qualities other than partisan disdain in their elected officials. Costa (2021) adds clarity to this topic, answering the question of whether evaluations of elected officials are driven by affect or ideology.

Costa utilizes 3 conjoint studies to answer this question, randomizing multiple attributes of elected officials, a full description of which is given in Table 1. By randomizing across all attributes, Costa can identify the causal effect of the inclusion of one attribute level in a profile relative to another. If affect influences behavior, respondents will be more or less likely to select or positively evaluate profiles of representatives giving affective signals. The effects of affective and ideological signals can then be compared to each other to determine the relative weight with which respondents rely on such signals when making evaluations of their representatives.

Table 1: Randomized Attributes of Study 1

Study	Attribute	Levels
Study 1	Sex	Male
		Female
	Party	Democrat
		Republican
	Race/ethnicity	White
		African American
		Hispanic
	Latest Tweet	“[Democrats/Republicans] are CORRUPT and IMMORAL. I will not sit silent while they lie to Americans and steer us in the wrong direction.”
		“Everyone in this country deserves basic health care. This is why we must support Medicare for all Americans.”
		“Providing Medicare for all Americans is NOT the right solution to fix health care in this country.”
		“One way to start fixing the immigration problem in this country is to increase spending for border security between the U.S. and Mexico.”
		“Increasing spending on border security between the U.S. and Mexico would NOT fix immigration policy in this country.”
	Constituent Relations	Answers more than 90 percent of constituent mail
		Answers less than half of constituent mail
Study 2	Sex	Male
		Female
	Party	Democrat
		Republican
	Race/ethnicity	White
		Black
		Latino
		Asian
	Latest Tweet	“[Democrats/Republicans] are not good for this country. Shameful.”
		“I am disgusted at todays [Democrats/Republicans]. They are dishonest, dangerous, and bad for America.”
		“Listening to [Democrats/Republicans] makes me so angry and fearful for this country. What a disgrace.”
		“We [Democrats/Republicans] are working hard to solve the nation’s problems.”
		“Show your support for [Democrats/Republicans] in Congress and follow us on Twitter!”
		“Proud of all the good work being done by my fellow [Democrats/Republicans].”
	Constituent Relations	Answers more than 90 percent of constituent mail
		Answers less than half of constituent mail
	Ban Assault Rifles	Voted for
		Voted against
	Increasing tax rate for high income people by 3%	Voted for
		Voted against
Study 3	Member quote	“We should do everything it takes to make sure [Democrats/Republicans] lose the next election” (Out-party loss)
		“We should do everything it takes to make sure [Democrats/Republicans] win the next election.” (In-party win)
		Not shown
	Issue priority	Respondent’s most important issue
		Respondent’s third most important issue
		Not shown

Several structural features change between the three studies. First, the dependent variable is either the choice of one profile over another as a respondent’s desired representative (study 1) or approval for a single representative’s profile using a 5-point scale (studies 2 and 3). Second, the randomization of policy versus affective statements is done concurrently in study 1, but separately in studies 2 and 3. Finally, study 3 has the addition of a “true control”, where no information is given about affect or ideology.

Across the three studies, Costa finds respondents tend to penalize representatives who display negative affect toward the opposite party. Respondents select such representatives less frequently relative to representatives who take stances on policy positions instead (study 1) and rate such representatives worse compared to those who emphasize positive affect toward in-party members (studies 2 and 3). Respondents reward representatives who display greater levels of service toward constituents and share similar policy positions.

These findings suggest limits to the reaches of affective polarization; while partisans may feel negatively toward each other, they are more motivated by shared policy attitudes and constituency service when making evaluations of representative quality. This opens an unanswered question; if voters punish politicians who engage in negative affective displays, when do politicians continue to denigrate members of the opposite party? Regardless, Costa paints a more hopeful picture of American democracy acting in a healthier manner than is often portrayed.

Methods and Analysis

Was the study and its analyses pre-registered?: No

Did the study rely on proxy variables to measure polarization?: No

Were standard p-value thresholds used (p<.05 or 95% Confidence Intervals that don’t overlap zero)?: Yes

• Largest p-value presented as significant: 0.05

Were correlational results interpreted with causal language?: No

Limitations / Weaknesses

Costa’s 3 conjoint designs account for the many of the potential effects of affective polarization on behavior, and such thoroughness is a strength of the paper. However, the effect of out-party animus independent of policy congruence, while estimated for the dependent variable of single-candidate evaluations, is not estimated in the context of preference for one candidate over another. This means the treatment condition of partisan negativity is comparing negativity to policy congruence, not the lack of negativity. This candidate versus candidate context is where we may expect such appeals to negative partisanship to be most effective: when one politician is pitted against another. Costa’s results seem to indicate the severity of punishments exerted on representatives when evaluated in isolation may not exist in a more competitive context; the effect of negative partisanship is much smaller when such negative sentiments are stoked by members of a respondent’s own party in study 1. Additionally, the use of the term “conjoint” to describe studies 2 and 3 is misleading. While conjoints can use continuous scales as dependent variables, they typically do so while comparing two or more profiles, often with repeated choice tasks. The power of conjoints comes from the ability to average effects against the joint distribution of all other attributes without having to show respondents all possible combinations, which is typically why respondents see multiple choice tasks and researcher cluster standard errors at the respondent level. When a single profile is given without another to compare to and with only a single trial, the nature of this effect as it is typically estimated in a conjoint is ambiguous. This design can more accurately be called a “vignette” experiment. Furthermore, separating partisan affect from policy may require more than the independent randomization of such features, as policy stances can be signals for strength of partisanship (Dias and Lelkes, 2022). The positive effect of policy congruence, then, could be as much an indication of policy-minded respondents exercising deeply-held preferences as an indication of deeply tribal partisans attempting to determine who the “true Democrat/Republican” is.

Open Data & Analyses

Does the article make the replication data publicly available?: Yes

Does the article make the replication analysis scripts publicly available?: Yes

Link to replication data.

Article Citation

Costa, M. (2021), Ideology, Not Affect: What Americans Want from Political Representation. American Journal of Political Science, 65: 342-358. https://doi.org/10.1111/ajps.12571

Bibtex

@article{https://doi.org/10.1111/ajps.12571,
author = {Costa, Mia},
title = {Ideology, Not Affect: What Americans Want from Political Representation},
journal = {American Journal of Political Science},
volume = {65},
number = {2},
pages = {342-358},
doi = {https://doi.org/10.1111/ajps.12571},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/ajps.12571},
eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1111/ajps.12571},
year = {2021}
}