Gender stratification is well documented at the macro and micro levels of analysis. Several theories explain how cultural beliefs about gender shape such inequalities. One such theory – status characteristics theory (Berger et al. 1977; Correll and Ridgeway 2006) – asserts that status characteristics generate self-fulfilling prophesies, whereby those with characteristics that are considered high-status are more influential, are more positively evaluated, and are more highly rewarded for their efforts. The mechanism that generates the self-fulfilling prophesy is purported to be expectation states, i.e., status differences result in differential expectations of competence and worth, which translate into behavioral inequalities. Historically, expectation states have been treated exclusively as theoretical constructs, not being directly measured. In part this is because of social desirability bias – explicit measures of gender status beliefs are likely to be biased because participants are unlikely to state that they view women as less worthy and competent than men (DeMaio 1984). Furthermore, respondents may not even be aware of their biases (Swim and Cohen 1997). Recent efforts, however, have shown progress in measuring status beliefs without explicit questionnaire items. Building on recent work using an implicit association test to measure racial status beliefs (Melamed et al. 2019), this paper develops an implicit gender status beliefs measure.
To overcome the discrepancy between responses and real attitudes, implicit association tests (IAT) measure the linkage between two dimensions by asking participants to sort objects as quickly as possible using computer keys (Greenwald et al. 1988). This paper focuses on two dimensions – gender and status. Participants sort images of men and women, and also words associated with high and low status (e.g., worthy, competent). Over repeated phases, the two dimensions are paired systematically to assess the extent to which a state of one dimension (e.g., male) is associated with a state of the other dimension (e.g., high status). Such an association is manifested by faster responses when “male” and “high status” are paired together than when “female” and “high status” are paired together. Such implicit beliefs are much less susceptible to social desirability bias than are explicit beliefs and cannot be controlled without training (Kim 2003).
In three studies, this paper systematically evaluates each dimension of the implicit gender status beliefs measure. In the first, participants (N = 463) sorted images of 11 different men and 11 different women, with each image presented in color and in grayscale. The goal of Study 1 was to identify whether images presented in grayscale or in color better evoke implicit gender categorization. Results from factor analyses reveal that grayscale is preferred to color images for implicit gender categorization. Moreover, factor analyses identify a subset of grayscale images (5 male and 5 female) that best elicit implicit gender categorization and can be used in subsequent studies on gender status beliefs.
The second study focused on implicit status categorization. Implicit association tests typically establish positive or negative evaluations (Schnabel et al. 2008), and status items are evaluative, e.g., it is better to be competent than incompetent. Thus Study 2 had two objectives, to establish that the content of words associated with relative status differs from mere evaluation and to test particular words that connote status. Participants (N = 467) sorted words over repeated phases. In some phases, they sorted good and bad words (e.g., rotten or pleasure), while in others they sorted high and low status words (11 of each). Importantly, factor analyses of the response latencies find that status and evaluations load on correlated but unique factors. Furthermore, results indicate that the words “capable,” “competent,” “able,” “worthy,” and “skilled” perform the best at distinguishing implicit relative status.
The third study combines the optimal images from Study 1 and the optimal words from Study 2 to create both a long-form and a brief implicit association test (Sriram and Greenwald 2009). Participants (N = 44) completed the IAT and BIAT twice; participants completed part 1 and then completed part 2 at least 24 hours later. Two multilevel linear regressions calculate the unconditional intraclass correlation coefficients for both the long-form IAT and the BIAT; these results represent the retest reliability of both methods (Koo and Li 2016). Results demonstrate that the retest reliability of the long-form IAT is higher than the retest reliability of the BIAT. All three studies combined identify a subset of images for implicit gender categorization, identify a subset of words for implicit status categorization (and show that they are different from mere evaluations), combine them to form a new measure of implicit gender status beliefs, and evaluate the measurement properties of the new measure. This paper concludes with directions for future research, including how to improve the measurement of gender status beliefs, and how the developed measure can be used in applied research settings to explain gender inequalities.