|Title||Behavioral Effects of Descriptive Representation|
|C1 Background and Explanation of Rationale||
How can diverse societies guarantee inclusiveness? Descriptive representation – that is, whether an individual shares a social characteristic such as ethnicity or gender with her elected representative – is advocated as an effective part of the solution. It has been shown to foster political stability, trust, participation, and efficacy, particularly among groups that have been historically marginalized (Gay, 2002; Mansbridge, 1999; Schwindt- Bayer and Mishler, 2005; West, 2017). The effects of descriptive representation have been said to go beyond the political realm and may empower disadvantaged social groups in all aspects of life.
Empowerment arises with the existence of role models, a change in the level of expected discrimination, and altered beliefs about how much governmental institutions enforce anti-discrimination policies (among others, Marx, Ko and Friedman (2009)). We propose a study that explicitly investigates empowered behavior outside of politics rooted in descriptive political representation. Specifically, we ask whether descriptive representation changes the beliefs of members of a disadvantaged group in their own ability as well as with respect to the level of discrimination they expect from others. Further, we test whether descriptive representation increases labor market participation among those who previously may have been dissuaded from such participation. Such effects of descriptive representation would allow a diverse society to increase the economic well-being of historically marginalized groups, thereby countering the pernicious effects of historical discrimination that have led to current wage, employment, and achievement gaps.
Our study (1) experimentally varies whether women’s descriptive representation in U.S. Congress is framed as positive achievement or reason for concern and, then, (2) elicits subjects’ beliefs about discrimination in society as well as measures their willingness to compete in the labor market. We employ a survey experimental research design. The survey experiment provides an ecologically valid instantiation of descriptive representation in the real political system.
Our study focuses on the effect of gender-based representation on women's beliefs about gender discrimination in society and women's behavior in the labor market. Women are under-represented in the legislatures of most Western democracies (Paxton et al 2007) and face a substantial wage gap (Altonji/Blank (1999). Fearing discrimination, they are likely to select out of competition for political office (Kanthak/Woon 2015) and the labor market (Niederle/Vesterlund 2007). Thus, empowerment through descriptive representation may have economic welfare-enhancing effects both for women in particular and for society in general.
|C2 What are the hypotheses to be tested?||
We elicit two kinds of average treatment effects: 1)Preferences over jobs: Women who receive the positive frame treatment are more likely to apply to and feel more qualified for a job that is of high status, highly paid, most competitive, and in a discriminatory industry (i.e. tech) than women who receive the negative frame treatment or no frame at all. 2) Influence of job attributes: Status, pay, and competitiveness of a job are less likely to negatively affect willingness to apply to or feel qualified for a job for women who received the positive frame treatment than for those who received the negative frame treatment or no frame at all. That is, for those in the treatment, if they have in fact been empowered in some way, should exhibit less hesitancy towards high status, high paying, competitive jobs relative to women in the control. In other words, the negative effect of discriminatory nature of an industry on the willingness to apply to or feel qualified for a job is smaller for women who received the positive frame treatment than for women who received the negative frame treatment or no frame at all.
Evidence of both kinds of treatment effects support our main hypothesis of an Empowerment effect of descriptive representation on women’s behavior in non-political domains.
|C3 How will these hypotheses be tested? *||
We recruit a sample of partisan women representative of the US voting-age population through Survey Sampling International (SSI). The survey experiment follows the following sequence:
After a range of questions on demographics, political attitudes, and political efficacy, respondents receive a battery of questions about gender roles in today’s society (i.e. a modern sexism scale), feelings of marginalization due to the fact that one is a woman, and about collective self-esteem. We expect modern sexism, feelings of marginalization and collective self-esteem (along with partisanship and political efficacy) to moderate our effects (details in hypotheses below).
We then give respondents an attention check. We kick out respondents who fail this check, which comes before treatment is given.
Respondents will then be presented with general information about the U.S. Congress as well as our survey experimental treatments. The treatments vary (1) whether respondents see additional information about the partisan, racial, and gender composition of the current U.S. Congress, or whether they receive no such information (information treatment). Further, treatments vary (2) whether respondents are given a positive prime highlighting the achievement that is the current gender composition of the US Congress, whether they see a negative prime surrounding this same information about the current gender composition, or whether they see no prime (frame treatment). Thus, our experimental treatments result in a 2x3 partial factorial design, where two of the cells are omitted (we omit the “no information, positive frame” and “no information, negative frame” conditions because while we want to be able to parse the effect of information, we do not need to identify the independent effect of the frame in the presence of information given that the frame itself has information about the number of women in Congress. See section 3.4 below for further details about sampling.). This results in 4 populated experimental cells (see section below on Experimental Treatments and Study Flow for further details). The positive frame treatment also includes a picture depicting the growth in the number of women in Congress from 1999 to 2019, whereas the negative frame has a picture depicting the large number of men in US Congress in 2019. These images are to make the treatments stronger, and are further detailed in the Experimental Treatments section.
We then ask manipulation check questions. Specifically, we ask for subjects’ best guess on the number of women in the US Congress, whether the number of women in Congress is higher or lower than they expected, and whether the number of women in Congress is an achievement or a reason for concern.
Respondents are then asked a range of questions to pin down different mechanisms by which those treatment conditions may affect the outcome measure in distinct ways. Specifically, subjects are asked their feelings towards a host of well-known public figures (half of which are women), names of members of Congress that inspired subjects recently, how much subjects’ believe they’d be discriminated against because of their gender if they were to apply for jobs right now, as well as their belief in institutional protection against discrimination, and questions measuring their self-efficacy and self-esteem.
After being told that this survey on political attitudes is now finished (a fake end screen), respondents are asked to begin another survey (which appears to subjects to be separate from the survey they just finished). This second survey is presented as asking them to evaluate five job descriptions according to a variety of job attributes. These job profiles (modeled off of advertisements of positions on Monster.com) randomly vary in a conjoint experimental design: the position’s status (executive, assistant, manager, administrative aide, etc), whether the position is among the highest paid of its kind, whether the job is the most competitive of its kind, and which industry the job is in (education or tech industry, previous pre-test data shows that the former is stereotyped to be feminine, and the latter is stereotyped to be masculine). Our subjects are asked how likely they are to apply to a job as described in the shown profile and whether they feel qualified for such a job.
|C4 Country||United States|
|C5 Scale (# of Units)||3,000|
|C6 Was a power analysis conducted prior to data collection?||Yes|
|C7 Has this research received Insitutional Review Board (IRB) or ethics committee approval?||Yes|
|C8 IRB Number||PRO18050064 (University of Pittsburgh)|
|C9 Date of IRB Approval||07/16/2018|
|C10 Will the intervention be implemented by the researcher or a third party?||Researchers, The intervention is embedded in a survey, which will be implemented on the Qualtrics platform (and subjects are invited to participate through Dynana—formerly Survey Sampling International (SSI)).|
|C11 Did any of the research team receive remuneration from the implementing agency for taking part in this research?||No|
|C12 If relevant, is there an advance agreement with the implementation group that all results can be published?||No|
|C13 JEL Classification(s)||not provided by authors|