What Can Conjoint Experiments Tell Us about Americans’ Abortion Attitudes?

Abstract

I report findings from a pair of conjoint experiments that presented respondents with a series of profiles of pregnant women and asked whether it should be possible for each to obtain a legal abortion. The profiles varied the reason for the abortion, gestational age, and demographic characteristics of the hypothetical woman. I find little evidence that women’s demographic characteristics—including their purported ethnoracial identities—affect these judgments. In contrast, the effects of gestational age and the reason for the abortion are substantial. Notably, the effects of gestational age appear to be linear and unresponsive to trimester and viability thresholds commonly cited in elite discourse. I also find that the reason for the abortion becomes more consequential as gestational age increases. Finally, I consider whether these effects vary with respondents’ party affiliation and gender. The findings offer new insights into the contours of abortion attitudes in the United States and illustrate the strengths and limitations of conjoint designs.

Keywords

conjoint abortion gender gap public opinion experiment

A substantial body of literature has examined the roles religiosity, gender, ethnoracial identity, and an array of other individual-level characteristics play in shaping broad support for permitting women to access abortion (e.g., Adamczyk & Valdimarsdóttir, 2018; Loll & Hall, 2019; Holman, Podrazik and Mohamed 2020, for reviews, see Jelen & Wilcox, 2003; Adamczyk et al., 2020). However, although Americans are often crudely characterized as either “pro-life” or “pro-choice,” attitudes about abortion are more complex than this distinction implies (Luker, 1984; Cook et al., 1992) and many Americans appear to have mixed feelings about this issue (Craig et al., 2002). For many, support for permitting abortion depends on the reason for the abortion, gestational age, and perhaps the characteristics of the pregnant woman (Munson, 2018).¹ Public attitudes about abortion can shape policy outcomes (Arceneaux, 2002), so it is important to understand the nature of these attitudes.

Abortion attitudes are most commonly measured using closed-ended survey questions. More often than not, these questions “focus on only one element of abortion—the reasons that a woman might have to seek an abortion” (Jelen & Wilcox, 2003, 490). However, a handful of studies have used survey experiments to pursue finer-grained assessments of the contours of the public’s abortion attitudes. For example, Hans and Kimberly (2014) report findings from a telephone survey that “built-out” randomly assigned aspects of a hypothetical pregnant woman’s situation sequentially.² Another study used the standard General Social Survey (GSS) battery of abortion questions as a starting point and randomly varied whether the question indicated that the woman was “less than 3 months pregnant” (Bumpass, 1997). In this research note I build on this work, reporting findings from a pair of conjoint experiments that presented respondents with detailed profiles of pregnant women and asked them whether the woman should be able to obtain a legal abortion. I make four contributions.

First, the conjoint design provides an avenue for disentangling the effects of considerations that may be correlated in people’s minds. For example, people may infer that abortions prompted by dangers to the health of the mother tend occur late in pregnancy or among older pregnant women. If so, differences in responses to a survey question that, say, asks whether abortion should be permitted if the woman’s health endangered and one that asks whether abortion should be permitted if the woman became pregnant as a result of rape may conflate the effects of the reason with considerations tied to gestational age or the age of the woman seeking the abortion. Similarly, in the second study, I account for the possibility that the apparent effects of other considerations mask effects of inferences respondents may make regarding the likely ethnoracial identity of the woman seeking the abortion (McClelland et al., 2020) by signaling whether the woman is white, Black, or Latina.

Second, I consider whether the demographic characteristics of the woman—her age, marital status, number of children, income, and race/ethnicity—affect willingness to permit abortion. These factors may be consequential, but have received scant attention in existing survey research. For example, people may be more inclined to permit abortion for women who are young, unmarried, or poor because they view these women as facing more substantial hurdles to bearing the costs associated with carrying a pregnancy to term or raising a child. Another possibility is that people are more inclined to permit abortion if a woman already has children, perhaps because they view experience as a mother as improving a woman’s ability to make a well-reasoned judgment about whether to seek an abortion. A particularly troubling possibility is that willingness to permit abortion depends on the race or ethnicity of the woman. Existing work suggests that people may be less inclined to protect the rights of racial or ethnic minorities (e.g., Doherty & Stancliffe, 2017). This said, the nexus between abortion policy and race in the United States is fraught and entangled with historical connections to eugenics movements (Luna, 2018). Thus, an alternative possibility is that public opinion—perhaps enhanced by a sense of racial threat (Craig et al., 2018)—echoes this history: people may be more inclined to permit women who are racial or ethnic minorities to access abortion.

Third, the design offers avenues to explore finer-grained and conditional characteristics of the public’s abortion attitudes. For example, legal arguments often treat trimester and fetal viability thresholds as critical considerations. However, I find that the effects of gestational age on support for permitting abortion are almost perfectly linear—a pattern that suggests a disjuncture between public attitudes and elite discourse. There is also theoretical reason to expect that the mitigating effects of offering a justification for the abortion vary with gestational age and some existing work finds evidence of this type of conditionality (Bumpass, 1997; Zigerell & Barker, 2011). Specifically, many people may find abortion to be broadly acceptable early in a pregnancy, but feel that a strong justification must be offered for later-term abortions. The evidence from my experiments is consistent with this expectation.

Finally, I assess whether the effects of the treatments vary with respondents’ party identification and gender. Republicans are unsurprisingly less inclined to support permitting abortion than Democrats in the experimental tasks (Adams, 1997; Carsey & Layman, 2006). However, this partisan gap is smaller when reason for the abortion is “hard” (Granberg & Granberg, 1980, e.g., when the life of the mother is endangered) and when the pregnancy is further along. Although existing work has yielded mixed findings regarding differences in summary abortion attitudes by gender (Cook et al., 1992; Hertel & Russell, 1999; Barkan, 2014; Loll & Hall, 2019), I find that women were broadly, and substantially, less supportive of permitting abortion when presented with profiles of a woman considering abortion. Previous work is also finds some evidence that the factors that shape abortion attitudes vary with gender (Walzer, 1994). Men and women responded similarly to most of the treatments in my experiments, with one clear exception: among women, support for permitting abortion drops more sharply with gestational age than it does for men—a pattern that is consistent with the results that emerged in observational analysis of the 2006 Cooperative Congressional Election Study (Zigerell & Barker, 2011).

Design and Data

I fielded two experiments online in fall 2019 (Study 1: September 23–25, 2019; Study 2: November 13–17, 2019). Participants were recruited via Lucid using quota sampling to be descriptively representative of the national adult population. One fourth of respondents in each study were randomly assigned to complete a set of closed-ended questions regarding when abortion should be permitted that mirror those asked on the General Social Survey (GSS), rather than the experimental tasks. Respondents’ demographic characteristics and reported abortion attitudes closely track those from the 2016 and 2018 GSS surveys (see Supplemental Table A1).³ In concert with emerging evidence that experiments conducted using convenience samples typically yield conclusions that track those found in probability samples (e.g., Krupnikov et al., 2021), this offers some reassurance regarding the validity and generalizability of the findings reported below. The experimental analysis includes all respondents who provided responses to demographic characteristics (Study 1, N = 1133; Study 2, N = 817).

Respondents in the first study were presented with a series of five profiles of hypothetical women and were asked whether they thought it should be possible for the woman “to obtain a legal abortion in these circumstances.” The woman’s age, family income, marital status, and number of children, as well as gestational age (in weeks) and the reason for the abortion were presented in tabular form (see Figure 1 for an example). The second study was identical, but for the fact that a name that signaled whether the woman was white, Black, or Latina was presented in the top row (Butler & Homola, 2017, see Supplemental Appendix A.2.1). The distribution of characteristics was selected with an eye toward balancing the need for sufficient statistical power to estimate effects along each treatment dimension with an effort to present respondents with examples that are reasonably representative of the characteristics of women who seek abortions in the United States (Jerman et al., 2016, see de la Cuesta et al.,. Forthcoming for a discussion of the value of being attentive to real world distributions when constructing conjoint tasks). A more detailed discussion of the distribution of treatments I used and rationale for those choices is presented in Supplemental Appendix A.2.

Figure 1.

Screen Shot of Conjoint Task. Example task from Study 2. I did not include the Name row in Study 1. Order of characteristics was randomized for each respondent, then fixed across tasks within respondent.

Findings

I stack the data so each observation corresponds to one profile and estimate OLS regression models predicting responses (1 = “Yes, should be possible”) with indicators for each of the treatments (binning continuous treatments as illustrated in Figure 2; see Bansak et al., 2019) and a vector of pre-treatment controls to improve the precision of my estimates (respondent gender, ethnoracial identity, age, ideological self-placement, party affiliation, and frequency of religious attendance). The analysis includes 5640 observations for Study 1 and 4063 for Study 2.⁴ In a model interacting each treatment with an indicator for Study 2, none of the interactions are statistically significant (p > .05 in all cases). Thus, to maximize statistical power, in the analysis that follows I pool data across studies.⁵ Treatment effects are reported in Figure 2. The average probability of indicating abortion should be permitted (marginal mean) was approximately 46% among all cases where a profile did not present a specific reason for the abortion. Neither the “did not use birth control,” nor the “used birth control, but it failed” treatments affected support for permitting abortion (relative to this “no reason given” reference group). However, the other reasons significantly mitigated opposition to permitting abortion. The “strong chance of a serious defect in the baby” justification increased support by 14 percentage points, “became pregnant as a result of rape” increased support by 22 percentage points, and “woman’s own health is seriously endangered” increased support by just over 30 percentage points.⁶

Figure 2.

Treatment Effects from Conjoint Experiments. Whiskers are 95% confidence intervals. N = 9703; 1950 respondents. Study 2 participants presented with a profile with a purportedly “white” name are the reference group for the other race/ethnicity indicators and for the indicator for Study 1 participants (where woman’s race/ethnicity was not signaled).

These effects are comparable to, but somewhat smaller than the differences that emerge in responses to the closed-ended questions completed by a random subset of my respondents and by 2018 GSS respondents. Compared to levels of support in response to the “for any reason” question (51% support among respondents in each group), when the question specified a “strong chance of serious defect” support was 23 and 25 percentage points higher among my “closed-ended respondents” and among 2018 GSS respondents, respectively. The corresponding differences were 34 and 28 percentage points when the pregnancy resulted from rape, and 36 and 39 percentage points if the woman’s health was endangered (see Supplemental Table A1).

The effects of the trimester of pregnancy are also substantial. Support for permitting abortion dropped in a seemingly linear fashion with gestational age: by approximately 11 percentage points with each trimester. I consider this apparent linearity in further detail by estimating a LOWESS line using the un-binned gestational age treatments in Figure 3. In other words, rather than collapsing the gestational age treatments into trimesters, I use the specific number of weeks presented in the profiles. The LOWESS line is almost perfectly linear.⁷ This suggests that public attitudes are not sensitive to the trimester or viability demarcations commonly cited in elite discourse.

Figure 3.

Linear Relationship between Gestational Age and Support. Solid line is a smoothed LOWESS line illustrating relationship between gestational age (in weeks) and the probability of a respondent supporting abortion. The (nearly identical) dashed line illustrates the linear fit. The three markers are marginal means, binning weeks by trimester; whiskers show 95% confidence intervals. N = 9703; 1950 respondents.

In contrast, the effects of the demographic characteristics of the woman are trivial. The largest estimated difference is the estimated 3.2 percentage point difference between the first and third quintile of family income (p < .05). Tests of the joint significance of the age, income, marital status, and number of children indicators were each statistically insignificant. Similarly, in Study 2, the indicators for the race/ethnicity treatments are statistically indistinguishable from one another (p > .10 for all tests).

Variation in Effects of Reason by Trimester

Next, I assess whether the effects of the reason for the abortion depend on gestational age. Figure 4 reports marginal means for each combination of reason and trimester treatments (Leeper et al., 2020). Two patterns are apparent. First, respondents were less likely to support permitting abortion later in a pregnancy, regardless of the reason. Second, consistent with the expectation that the reason for the abortion would be more consequential later in a pregnancy, the effects of each of the three reasons that had a statistically significant effect in the main analysis are amplified in the second and third trimesters.

Figure 4.

Reason for Abortion, by Trimester (marginal means). Markers show marginal means—the probability of a respondent saying abortion should be permitted in the circumstance, averaged across all other treatment conditions. Whiskers are 95% confidence intervals. N = 9703 (trimester 1 = 2678; trimester 2 = 4086; trimester 3 = 2939); 1950 total respondents).

For example, when the pregnancy was in the first trimester, the difference between the probability of saying abortion should be permitted when no reason is specified (61%) and when there is a “strong chance of a serious defect in the baby” (68%) is 7 percentage points. This difference more than doubles to 18 points in the second trimester and 15 points in the third trimester. Similarly, the effect of the woman having become pregnant as a result of rape is 13 percentage points in the first trimester, but 25 percentage points in the second and third trimesters. Finally, the 25 point effect of danger to the woman’s own health in the first trimester grows to 31 and 35 points in the second and third trimesters, respectively.⁸ These differences in effects between the first and later trimesters are each statistically significant.⁹

Group Differences

Finally, I present descriptive analysis of whether treatment effects varied with respondents' partisanship or gender. Figure 5 reports marginal means tied to the reason for the abortion and trimester from models estimated separately for Democratic and Republican respondents (treating leaners as partisans). Support for permitting abortion is systematically lower among Republicans than among Democrats. However, the gap between Republicans and Democrats narrows when a reason is presented. For example, Republicans supported permitting abortion in 29% of tasks where no reason for the abortion was given, compared to Democrats, who supported permitting abortion in 59% of these tasks—a gap of about 30 percentage points. This gap narrows to just over 20 percentage points when there is a risk of serious birth defect or the pregnancy resulted from rape, and to only 15 percentage points when the woman’s health is endangered. This narrowing of differences across parties is tied to the fact that the effects of the risk of birth defect, the pregnancy resulting from rape, and threats to the health of the woman (relative to no reason being given) are more pronounced among Republicans (p < .05 in each case).¹⁰

Figure 5.

Marginal Means, by Respondent Party Identification Markers show marginal means—the average probability of a respondent saying abortion should be permitted in the circumstance, averaged across all other treatment conditions. Black markers are estimates among Democratic respondents; gray markers are for Republicans. Whiskers are 95% confidence intervals. Analysis relies on 4712 Democratic cases (947 Democratic respondents); 3640 Republican cases (732 Republican respondents). See Supplemental Figure A1 for estimated effects of all treatments.

The partisan gap also narrows in cases where the pregnancy is in the third trimester. Democrats supported permitting abortion in 78% of cases when the pregnancy was in the first trimester; Republicans said abortion should be permitted in 54% of these cases—a gap of 24 percentage points. This gap narrows to 16 percentage points in the third trimester (the gap in the second trimester is similar to the difference in the first trimester). Here, the narrowing of the gap appears to be rooted in the fact that the effect of the pregnancy being in the third, rather than first, trimester was more pronounced among Democrats (p < .05). The fact that partisan differences narrow under some circumstances is notable. This said, broadly speaking, the patterns that emerge across parties are similar. The characteristics of the woman seeking an abortion do not affect support for permitting abortion among Democrats or Republicans (see Supplemental Figure A1 for a presentation of treatment effects by party akin to that reported in Figure 2); gestational age and the reason for the abortion substantially and significantly affect support among both groups.

Figure 6 shows marginal means from models estimated separately by respondent gender (see Supplemental Figure A2 for a presentation of treatment effects by respondent gender akin to that reported in Figure 2). The results indicate that women were less likely to say abortion should be permitted regardless of the reason or gestational age. Indeed, the negative, statistically significant coefficient on Female in column (1) of Supplemental Table A2 indicates that, on average, women were 10 percentage points less likely to say abortion should be permitted across experimental tasks. Figure 6 offers suggestive evidence that this gender gap is particularly pronounced when the reason for the abortion is the risk of a serious birth defect, and less pronounced when the woman’s health is in danger. However, the difference between these two differences is the only one that reaches conventional thresholds for statistical significance.

Figure 6.

Marginal Means, by Respondent Gender Markers show marginal means—the average probability of a respondent saying abortion should be permitted in the circumstance, averaged across all other treatment conditions. Black markers are estimates among respondents who identified as male; gray markers are for female respondents. Whiskers are 95% confidence intervals. Analysis relies on 4581 male cases (921 male respondents); 5122 female cases (1029 female respondents). See Supplemental Figure A2 for estimated effects of all treatments.

There is clearer evidence that women were more sensitive to the gestational age treatment than men. Respondents who identified as male were approximately 8 percentage points less likely to say abortion should be permitted when the profile indicated that the pregnancy was in the second (rather than first) trimester; the comparable effect was 13 percentage points among female respondents (tests of equality of effects: p = .056). Similarly, men were 17 percentage points less likely to say abortion should be permitted when the pregnancy was in the third trimester, compared to the 26 percentage point effect among women (p < .01).¹¹ This said, broadly speaking, men and women responded similarly to the treatments.

Discussion

Conjoint experiments are increasingly common in political research (Bansak et al., 2021). Here, I assessed what these designs can tell us about the American public’s abortion attitudes. Some of the patterns that emerge are consistent with patterns found in standard batteries of survey questions. For example, support for permitting abortion varies substantially depending on the reason for the abortion. That said, the evidence also offers insights into the contours of the public’s attitudes about abortion that standard survey questions do not provide. I find little evidence of the “stair step” relationship between gestational age and support for permitting abortion that elite discourse regarding trimester and viability thresholds suggests should emerge—the relationship between gestational age and support for permitting abortion is linear. I also find that the reason for the abortion becomes more consequential after the first trimester. Notably, the experiments yield little evidence that the characteristics of the woman seeking an abortion directly affect attitudes. These null findings are encouraging in that they suggest that, at least on average, people do not view the right to an abortion as contingent on women’s demographic characteristics.

This said, the effects of pregnant women’s race and ethnicity approach conventional thresholds of statistical significance. Given the complexity of potential connections between these characteristics and abortion, further research is warranted. In additional exploratory analysis, I consider the possibility that these treatment effects (from Study 2) varied based on respondents’ self-reported ethnoracial identity by interacting all variables in my core model with indicators for this respondent characteristic. The results are reported in Supplemental Appendix Figure A5. I find suggestive, but inconclusive, evidence that they did. Respondents who identified as white were almost 4 percentage points more likely to say abortion should be permitted when the woman’s name signaled that she was Latina (p = .038). I also find suggestive evidence that Latinx respondents were more supportive of permitting abortion when the woman’s name signaled she was not white. However, caution is warranted given that these estimates are noisy and none of the differences in effect sizes between white, Black, and Latinx respondents reach conventional thresholds of statistical significance.

I also presented descriptive evidence that the partisan gap in support for abortion narrows when there is a risk of serious birth defect, when the pregnancy resulted from rape, and when the health of the mother is in danger. It also narrows in cases where the pregnancy is in the third trimester. Consistent with evidence that the effects of questions wording on reported abortion attitudes can vary across subgroups (e.g., Singer & Couper, 2014), this narrowing of the partisan gap is tied to differences in how Democrats and Republicans respond to the “reason” and gestational age treatments. This said, in line with recent work that finds substantial consensus in terms of how partisans respond to particular features of policy proposals (Hainmueller & Hopkins, 2015; Doherty et al., 2020; Coppock, 2021) these are differences in effect magnitudes, rather than differences in which considerations “matter” to respondents.

I also show that, when presented with concrete examples of women seeking abortions, women are significantly and substantially less supportive of permitting abortion than men. Specifically, across experimental tasks, women were approximately 10 percentage points less likely to say the woman in the profile should have legal access to abortion (see coefficient on Female in column [1] of Supplemental Table A2). A similar gender gap does not emerge in models specifying a standard 6-item additive abortion attitude index as the outcome variable using either the 2018 GSS or the subset of my respondents assigned to answer the GSS battery instead of completing the conjoint tasks (see Supplemental Table A5). Women’s abortion attitudes are also more sensitive to gestational age than men. These patterns suggest promising avenues for future work. For example, perhaps being presented with concrete profiles of pregnant women considering abortion—as opposed to being asked about broad rules regarding when abortion should be permitted—triggers distinctive cognitive or emotional responses in women. Similarly, researchers could assess whether women’s greater sensitivity to gestational age is driven by women respondents who have experienced carrying a pregnancy to term. One possibility is that direct experiences with pregnancy (e.g., the sensations that accompany “quickening”) increase sensitivity to and awareness of how fetuses develop and women’s bodies change through the course of pregnancy.

As with all research designs, the conjoint approach demonstrated here has strengths and limitations. Conjoint designs allow researchers to consider how an array of dimensions and levels of those dimensions each independently affect attitudes. They may also reduce social desirability bias when respondents are asked about sensitive topics like abortion (Sedgh & Keogh, 2019; Pelzer, 2019; Horiuchi et al., 2021). This said, the multitude of dimensions varied in these experiments, multitude of potential interactions between treatment dimensions, and avenues for subgroup analyses may set the stage for reporting of false positives. Additionally, although respondents who were randomly assigned to answer the GSS battery provided responses that closely track those found in the GSS, online surveys—which are often used to field conjoint studies—may not be representative of those of the broader public on unmeasured characteristics (e.g., although my surveys measured frequency of religious attendance, they did not measure religious affiliation—a characteristics that may have affected response to the treatments). Perhaps most importantly, although the designs I used offer new insights into the contours of aggregate abortion attitudes, they do not yield an individual-level measure of abortion attitudes that could be used as a covariate or outcome measure in analysis that treats individuals as the unit of analysis.

Supplemental Material

sj-pdf-1-apr-10.1177_1532673X211053211 – Supplemental Material for What Can Conjoint Experiments Tell Us about Americans’ Abortion Attitudes?

Supplemental Material, sj-pdf-1-apr-10.1177_1532673X211053211 for What Can Conjoint Experiments Tell Us about Americans’ Abortion Attitudes? by David Doherty in American Politics Research

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

David Doherty

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Adamczyk

Kim

Dillon

(2020). Examining public opinion about abortion: A mixed-methods systematic review of research over the last 15 years. Sociological Inquiry, 90(4), 920–954. https://doi.org/10.1111/soin.12351.

Adamczyk

Valdimarsdóttir

(2018). Understanding Americans’ abortion attitudes: The role of the local religious context. Social Science Research, 71(2), 129–144. https://doi.org/10.1016/j.ssresearch.2017.12.005.

Adams

G. D.

(1997). Abortion: Evidence of an issue evolution. American Journal of Political Science, 41(3), 718–737. https://doi.org/10.2307/2111673.

Arceneaux

(2002). Direct democracy and the link between public opinion and state abortion policy. State Politics & Policy Quarterly, 2(4), 372–387. https://doi.org/10.1177/153244000200200403.

Bansak

Bechtel

M. M.

Margalit

(2019). “Why Austerity? The Mass Politics of a Contested Policy.”. Mimeo. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3486227.

Bansak

Hainmueller

Hopkins

D. J.

Yamamoto

Druckman

J. N

Green

D. P

(2021). Conjoint survey experiments. Advances in Experimental Political Science, 3(1), 19–41. https://doi.org/10.1017/9781108777919.004.

Barkan

S. E.

(2014). Gender and abortion attitudes: Religiosity as a suppressor variable. Public Opinion Quarterly, 78(4), 940–950. https://doi.org/10.1093/poq/nfu047.

Berinsky

A. J.

Huber

G. A.

Lenz

G. S.

(2012). Evaluating online labor markets for experimental research: Amazon.com’s Mechanical Turk. Political Analysis, 20(3), 351–368. https://doi.org/10.1093/pan/mpr057.

Bumpass

L. L.

(1997). The measurement of public opinion on abortion: The effects of survey design. Family Planning Perspectives, 29(4), 177–180. https://doi.org/10.2307/2953382.

10.

Burnett

C. M.

Kogan

(2015). When does ballot language influence voter choices? Evidence from a survey experiment. Political Communication, 32(1), 109–126. https://doi.org/10.1080/10584609.2014.894160.

11.

Butler

D. M.

Homola

(2017). An empirical justification for the use of racially distinctive names to signal race in experiments. Political Analysis, 25(1), 122–130. https://doi.org/10.1017/pan.2016.15.

12.

Carsey

T. M.

Layman

G. C.

(2006). Changing sides or changing minds? Party identification and policy preferences in the American electorate. American Journal of Political Science, 50(2), 464–477. https://doi.org/10.1111/j.1540-5907.2006.00196.x.

13.

Cook

E. A.

Jelen

T. G.

Wilcox

(1992). Between two absolutes: Public opinion and the politics of abortion. Routledge.

14.

Coppock

(2021). Persuasion in Parallel. In Chicago studies in American politics. University of Chicago Press. Forthcoming.

15.

Coppock

McClellan

O. A

(2019). Validating the demographic, political, psychological, and experimental results obtained from a new source of online survey respondents. Research & Politics, 6(1), 2053168018822174. https://doi.org/10.1177/2053168018822174.

16.

Craig

S. C.

Kane

J. G.

Martinez

M. D.

(2002). Sometimes you feel like a nut, sometimes you don’t: Citizens’ ambivalence about abortion. Political Psychology, 23(2), 285–301. https://doi.org/10.1111/0162-895x.00282.

17.

Craig

M. A.

Rucker

J. M.

Richeson

J. A.

(2018). Racial and political dynamics of an approaching ‘majority-minority’ United States. The ANNALS of the American Academy of Political and Social Science, 677(1), 204–214. https://doi.org/10.1177/0002716218766269.

18.

de la Cuesta

Egami

, & Imai

(2022). Improving the external validity of conjoint analysis: The essential role of profile distribution. Political Analysis, 30(1), 19-45. https://scholar.princeton.edu/negami/publications/improving-external-validity-conjoint-analysis-essential-role-profile.

19.

Djupe

P. A.

Lewis

A. R.

Jelen

T. G.

Dahan

C. D.

(2014). Rights talk: The opinion dynamics of rights framing. Social Science Quarterly, 95(3), 652–668. https://doi.org/10.1111/ssqu.12083.

20.

Doherty

Bryan

A. C.

Hanania

Pajor

(2020). The public’s foreign aid priorities: Evidence from a conjoint experiment. American Politics Research, 48(5), 635–648. https://doi.org/10.1177/1532673x20939925.

21.

Doherty

Stancliffe

(2017). Interpreting and tolerating speech: The effects of message, messenger, and elite framing. American Politics Research, 45(2), 224–255. https://doi.org/10.1177/1532673x16667090.

22.

Gallup Organization (2003). Gallup news service poll: George W. BushMiddle East/terrorism/abortion/death penalty/assisted suicide/homosexuality. Cornell University, Roper Center for Public Opinion Research.

23.

Gallup Organization . 2018. Gallup poll social series: Values and beliefs

24.

Granberg

B. W.

(1980). Abortion attitudes, 1965-1980: Trends and determinants. Family Planning Perspectives, 12(5), 250–261. https://doi.org/10.2307/2134868.

25.

Hainmueller

Hopkins

D. J.

(2015). The hidden American immigration consensus: A conjoint analysis of attitudes toward immigrants. American Journal of Political Science, 59(3), 529–548. https://doi.org/10.1111/ajps.12138.

26.

Hainmueller

Hopkins

D. J.

Yamamoto

(2014). Causal inference in conjoint analysis: Understanding multidimensional choices via stated preference experiments. Political Analysis, 22(1), 1–30. https://doi.org/10.1093/pan/mpt024.

27.

Hans

J. D.

, & Kimberly

(2014). Abortion attitudes in context: A multidimensional vignette approach. Social Science Research, 48, 145-156. https://doi.org/10.1016/j.ssresearch.2014.06.001.

28.

Hertel

B. R.

Russell

M. C.

(1999). Examining the absence of a gender effect on abortion attitudes: Is there really no difference? Sociological Inquiry, 69(3), 364–381. https://doi.org/10.1111/j.1475-682x.1999.tb00876.x.

29.

Holman

Podrazik

Silber Mohamed

Mohamed

H. S.

(2020). Choosing choice: How gender and religiosity shape abortion attitudes among Latinos. Journal of Race, Ethnicity, and Politics, 5(2), 384–411. https://doi.org/10.1017/rep.2019.51.

30.

Horiuchi

Markovich

Z. D.

Yamamoto

(2021). Can conjoint analysis mitigate social desirability bias? MIT Political Science Department Research Paper.

31.

Jelen

T. G.

Wilcox

(2003). Causes and consequences of public attitudes toward abortion: A review and research agenda. Political Research Quarterly, 56(4), 489–500. https://doi.org/10.1177/106591290305600410.

32.

Jerman

Jones

R. K.

Onda

(2016). Characteristics of U.S. abortion patients in 2014 and changes since 2008. Technical report Guttmacher Institute.

33.

Krupnikov

Nam

H. H.

, & Style

(2021). Convenience samples in political science experiments. Advances in Experimental Political Science, 165. https://doi.org/10.1017/9781108777919.012.

34.

Leeper

T. J.

Hobolt

S. B.

Tilley

(2020). Measuring subgroup preferences in conjoint experiments. Political Analysis, 28(2), 207–221. https://doi.org/10.1017/pan.2019.30.

35.

Loll

Hall

(2019). Differences in abortion attitudes by policy context and between men and women in the world values survey. Women & Health, 59(5), 465–480. https://doi.org/10.1080/03630242.2018.1508539.

36.

Luker

(1984). Abortion and the politics of motherhood. University of California Press.

37.

Luna

(2018). Black children are an endangered species’: Examining racial framing in social movements. Sociological Focus, 51(3), 238–251. https://doi.org/10.1080/00380237.2018.1412233.

38.

McClelland

S. I.

Dutcher

Crawford

(2020). In the fabric of research: Racial and gender stereotypes in survey items assessing attitudes about abortion. Journal of Social Issues, 76(2), 239–269. https://doi.org/10.1111/josi.12367.

39.

Mulligan

Habel

(2011). An experimental test of the effects of fictional framing on attitudes. Social Science Quarterly, 92(1), 79–99. https://doi.org/10.1111/j.1540-6237.2011.00758.x.

40.

Munson

(2018). Abortion politics. Polity Press.

41.

Pelzer

(2019). The potential of conjoint analysis for communication research. Communication Research Reports, 36(2), 136–147. https://doi.org/10.1080/08824096.2018.1559138.

42.

Schuman

Presser

Ludwig

(1981). Context effects on survey responses to questions about abortion. Public Opinion Quarterly, 45(2), 216–223. https://doi.org/10.1086/268652.

43.

Sedgh

Keogh

S. C.

(2019). Novel approaches to estimating abortion incidence. Reproductive Health, 16(1), 44–10. https://doi.org/10.1186/s12978-019-0702-0.

44.

Singer

Couper

M. P.

(2014). The effect of question wording on attitudes toward prenatal testing and abortion. Public Opinion Quarterly, 78(3), 751–760. https://doi.org/10.1093/poq/nfu028.

45.

Tuller

H. M.

Bryan

C. J.

Heyman

G. D.

Heyman

G. D.

, & Christenfeld

N. J. S.

(2015). Seeing the other side: Perspective taking and the moderation of extremity. Journal of Experimental Social Psychology, 59, 18-23. https://doi.org/10.1016/j.jesp.2015.02.003.

46.

Wallace

D. S.

DesforgesDesforges

D. M.

Thomas

J. C.

Lord

C. G.

SiaSia

T. L.

(2001). Effects of perceived consensus on selective reinforcement of other people’s stated opinions. Basic and Applied Social Psychology, 23(2), 75–85. https://doi.org/10.1207/s15324834basp2302_1.

47.

Walzer

(1994). The role of gender in determining abortion attitudes. Social Science Quarterly.

48.

Zigerell

L. J.

Barker

D. C.

(2011). Safe, legal, rare? And early: Gender and the politics of abortion. Journal of Elections, Public Opinion and Parties, 21(1), 83–96. https://doi.org/10.1080/17457289.2010.537341.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.23 MB