We have been engaged in decades-long public policy debates on gaps and how best to close them: the income gap, the student achievement gap, gender-linked gaps in employment opportunities. But why do we care so much about gaps? In a land of diversity, why are subgroup differences such a concern?
At a basic level, we care about gaps because (or when) our fundamental assumption is that, on a “level playing field,” there should be no systematic differences among people based on ascribed traits, such as race and gender, that are unrelated to the “game.” It is “ok” if a specific Hispanic kid performs at a lower level than his/her white counterpart or vice-versa. But it’s not ok if, on average, Hispanic students’ test scores systematically lag behind that of similar white children. Why? Because we know intelligence and ability are normally distributed across racial/ethnic groups. So, when groups differ in important outcomes, we know that this “distance” is indicative of other problems.
What problems exactly? That is a more complex question.
My colleague recently wrote a post in which he explained that the raw differences in male/female earnings are not primarily caused by “unequal pay for equal work” by equally-qualified workers. According to the data, women earn around 75 cents on the male dollar. But when other important factors, such as educational levels and seniority, are taken into account – or “controlled for” in statistical jargon – there is only a 5-15 cent difference that can’t be explained. This “residual,” he argues, can be regarded as “a more accurate estimate of the extent of discrimination.”
This is, of course, correct. But, as my colleague also notes in that post, there are forms of gender discrimination in today’s work organizations that are not pay discrimination – not directly anyway – but which are caused by similar underlying factors and can lead to inequalities in access to rewards – including pay. While it is important to be precise in the presentation of detail, it is essential not to miss the big picture.
The 5-15 cent statistic, as correct as it may be, should not obscure a more important reality: men’s and women’s experiences in the workplace are still very different for a variety of reasons. Many of these reasons share a lowest common denominator: existing cultural beliefs about men and women that tend to disadvantage working women.
To illustrate my point, let’s take a look at another gap: The gender birth gap. The natural sex ratio at birth is 105. That is, there are 105 males born for every 100 females. When a country’s sex ratio exceeds the 105-threshold, we begin to worry because this ratio – and its consequences – are not natural.
Through the press, we have become all too familiar with the unfortunate reality – first documented by Das Gupta (1987), Coale (1984), and Sen (1992) – that some societies still show a strong preference for sons (also see here and here). So strong, in fact, that people in these cultures intervene in nature to reduce the number of baby girls, through a variety of different mechanisms, including child abandonment/adoption, selective in vitro fertilization, selective abortion – even infanticide. How much of the gap is explained by the premeditated murder of baby girls? Probably not all of it. In fact, some scholars have pointed out that son preference is so deeply rooted in some cultures that it can result in discrimination against girls in nutrition, as well as preventive and, especially, curative health care, all of which can lead to excess girl mortality.
So while not all the missing girls are victims of infanticide, most of them are still victims of beliefs that relegate females to second class citizenship. The acts to prevent or deny their births are still rooted in the low status that their societies accord to women. These beliefs have led some individuals to commit horrible acts, including infanticide. Accordingly, when we consider the ways that this gap might be remedied, we should maintain focus on the sway of these deeply rooted beliefs, not narrow in on the form that these beliefs may take – e.g., infanticide, neglect, selective abortions, etc.
So, when we say a portion of the gender wage gap “disappears” when we “control for” different factors, we are just “unpacking disadvantage.” Women only earn about 75 cents to the male dollar; but, only 5-15 cents of this gap is “unexplained.” Yes; but this statement should not be interpreted to mean that only this residual is important or “unjustified.” While not all differences in pay between men and women are due to employers rewarding men and women differently for the same work, the gender gap in wages is the result of the systemic inequalities that continue to disadvantage women in today’s work organizations* – for an excellent example of how gender biases can affect employees working under seemingly objective systems see Profesor Janice Fanning Madden’s recent research here and here.
This residual only means that there are sources of disadvantage that we have not yet measured or included in our models that also decrease women’s wages. Similarly, if we “controlled for” baby girl neglect, diminished access to care, malnourishment, adoption, and so forth, we would be disaggregating the relative contribution of these factors to unnatural sex ratios. But never would the implication be that it is just the “residual” – which may or may not be infanticide – that is problematic. Explaining the gap – or breaking it down into pieces – does not mean that either the pieces or the totality cease to be of concern, especially as it is the case here, when the pieces are manifestations of the same primary cause.
- Esther Quintero
* Most of our evidence about the effect of gender beliefs on work-relevant situations comes from laboratory experiments where performance is held experimentally constant and gender is manipulated. Overall, this research has consistently found that cultural beliefs about gender bias the perceptions and evaluations of women and men in ways that tend to favor men. There is some evidence that these types of biases also emerge in more “real world” settings – see Steinpreis, Anders, & Ritzke 1999; Trix and Psenka 2003; Sinclair & Kunda 2000; Wenneras & Wold 1997; Goldin & Rouse 2000.