آهنگ مادرانه و سیاست های عمومی: درباره پیشگویی های خود انجامی و شکاف جنسیتی در استخدام و ارتقاء

Consider a model with two types of jobs. The profitability of hiring a worker to a fast-track job depends not only on his or her observable talent, but also on incontractible effort. We investigate whether self-fulfilling expectations may lead to higher hiring or promotion standards for women. If employers expect women to do more household work than men, thereby exerting less effort in their paid job, then women must be more talented to make it profitable to hire them. Specialization in the family will then result in women doing most of the household work. Such self-fulfilling prophecies can be defeated by affirmative action or family policy. However, it is unlikely that temporary policy can move the economy to a symmetric equilibrium: policy must be made permanent. Anti-discrimination policy need not enhance efficiency, and from a distribution viewpoint this is a policy with both winners and losers.

Why does a marked sexual division of labor seem to continue even when comparative advantages, for example based on physical strength, is of diminishing importance? The traditional picture of a breadwinning father and a homemaking mother may be disappearing in modern economies, but there is marked sectoral segregation based on gender and women also tend to hold jobs that are more compatible with bearing the main responsibility for children.1 Self-fulfilling prophecies have been suggested as one possible basis for enduring discrimination in the labor market. If an employer expects women to shoulder the brunt of household work, this will have consequences both for firms’ hiring decisions and investments in female human capital. When the family in turn shall decide on its time allocation, the fact that women tend to earn less makes it rational that the woman actually do carry the main responsibility for children and is less active in the labor market. The key question is to which extent public policy can break this type of discrimination. More precisely, our main concern is if anti-discrimination policies only need to be applied a short while, for beliefs to change, or if policy needs to be permanent, to prevent the economy from sliding back into the previous discriminatory situation. Of course, it is much easier to combat discrimination if policy only needs to be temporary. Unfortunately, we show that under what we claim are plausible assumptions, anti-discrimination policy designed to break self-fulfilling prophecies must be in place permanently to have a permanent effect. Our starting point is the following: suppose there are two types of jobs, “fast track” and “slow track” jobs. A worker is placed on the fast track if a fixed investment, paid by the firm, is undertaken. Future wages are incontractible at the time of investment, and assumed to be set ex post through bargaining. The firm then gets a fraction of the value produced by any given employee. This value in turn depends on the (known) ability of the person, but also on his or her future “effort”, which is to be interpreted broadly – for instance to include the willingness to work irregular hours and not take long parental leaves. The employers only place those employees on the fast track whose expected output is so large that the investment cost is recouped. Suppose that women traditionally have been the ones to assume primary responsibility for child care and that such responsibilities make it more costly to exert effort in the labor market. If employers expect less effort from women than from men of comparable talent also in the future, then women, more often than men, will be put on the labor market's slow track. The decision concerning who will actually assume primary responsibility for the children and who is allowed to concentrate on the outside labor market is then taken within the family. Most family models predict some degree of specialization according to comparative advantage, and perhaps the most important source of a comparative advantage in household production is a low outside wage.2 Can it be that employers’ beliefs in this way turn into self-fulfilling prophecies? This would mean that the slow track in the labor market predominantly would be a “mommy track.” The present paper formalizes this story. We associate self-fulfilling prophecies with situations where the equilibrium outcome is asymmetric even when agents are symmetric. Whenever there is an asymmetric equilibrium in which the female assumes primary responsibility for household production, there will be another asymmetric equilibrium with reversed gender roles. Self-fulfilling beliefs can make one of these equilibria focal. The paper establishes conditions under which discrimination prevails in an imagined world in which men and women are identical in all economic respects (which in our setup basically means that there are no gender differences in preferences or skills). Two conclusions can be drawn: (i) The symmetric model sometimes has only asymmetric stable equilibria, meaning that self-fulfilling prophecies discrimination can be an equilibrium phenomenon. But notice that this is the case only under given conditions. (ii) Under quite general assumptions, if asymmetric equilibria exist, then there is no stable symmetric equilibrium. This implies that successful anti-discrimination policy will have to be permanent to sustain a situation that is not a stable equilibrium – rather than shifting the economy from a discriminatory equilibrium to a nondiscriminatory one. The latter would arguably be the easier task. We argue that many types of policy can, indeed, break a discrimination equilibrium. Affirmative action and family policy are but two examples of policies that might work. Turning to welfare issues, however, it turns out that anti-discrimination policy is a policy with both winners and losers. Perhaps paradoxically, discriminatory hiring standards make the income distribution across families more even. If fair hiring standards are introduced, the likelihood that some families will have two workers in fast-track jobs and some families having none will increase. This could suggest that anti-discrimination policies should be complemented by redistributive measures. Even if these distributional effects are neutralized, it is not certain that anti-discrimination measures will increase welfare. On the positive side, fair hiring standards means that the talent capital in society is put to better use. However, as someone has to do the housework in a family, it can be wasteful to have both parties in a couple in fast-track jobs. Discrimination can be seen as a coordination device that reduces the probability of wasteful investments in both partners. These caveats should not cloud what our analysis reveals: that discriminatory treatment of women can arise even when men and women are completely equal in all economic respects; that policy – if made permanent – can effectuate symmetric treatment of the sexes; and that there are cases where anti-discrimination policy is welfare-improving. However, the correct application of policy measures requires a careful analysis of the situation. A string of existing papers discuss self-fulfilling prophecies and different types of labor market discrimination in various model settings. We mention Coate and Loury, 1993, Francois, 1998, Engineer and Welling, 1999, Moro and Norman, 2003, Moro and Norman, 2004, Bjerk and Han, 2007 and Albanesi and Olivetti, 2009 and Dolado et al. (2013). Coate and Loury's (1993) seminal paper builds on Arrow's (1973) version of the statistical discrimination model. The essence of Arrow's model is that employers’ negative stereotypes about a group can in themselves weaken incentives for acquiring skills, and in this way become self-fulfilling prophecies.3Coate and Loury (1993) redesign Arrow's model to form a model of job discrimination rather than wage discrimination within one given job category.4 Coate and Loury's most important claim about policy is that temporary affirmative action may move the economy from a discriminatory to a non-discriminatory outcome. When black workers know they have a higher promotion probability, they will invest in productivity, and when employers discover that black workers are more productive than they originally thought they were, they will revise their beliefs. When beliefs have changed, there is no longer need for the original policy. In our setting, it is quite likely that affirmative action must be permanent. This, of course, makes anti-discrimination policy much less attractive. The Coate–Loury model is tailored to a situation with racial discrimination. There are no links between the situations for black and white workers, so in principle black workers could have been studied in isolation. In such a case, there are few constraints on assumptions about beliefs, which makes it rather easy to construct a case with effective temporary anti-discrimination policy. Our case is different, however. Men and women tend to live together in families and have joint children. As we know that someone must take care of the children, any belief we might have about how much effort women will typically exert in the labor market is logically linked to what one must rationally believe about men's behavior. We will show that this tends to produce situations in which the economy reverts to an asymmetric equilibrium once policies that encourage women's careers are lifted.5,6 Getting rid of discrimination in the Coate–Loury setting is always welfare-improving. In a family context, enforcing gender equality when there are substantial gains from specialization turns out, as already suggested, to be a much more complicated issue from a welfare perspective.7,8 In contrast to the Coate–Loury paper, Francois, 1998, Bjerk and Han, 2007 and Albanesi and Olivetti, 2009 and Dolado et al. (2013) explicitly consider gender discrimination and present models that share the following key characteristics: (i) there are no ex ante gender differences with respect to preferences or skills, (ii) labor markets are imperfect, (iii) household decisions (about the allocation of housework) affect either job effort or the probability of quitting the job, and (iv) these household decisions are unobservable to the employer when job contracts are offered. Although the details of the models differ with respect to the type of job contract, the type of labor market imperfection and the types of information asymmetries, they all share the feature that asymmetric (or “gendered”) equilibria, where workers are systematically offered different job contracts based on gender, can arise as a result of self-fulfilling prophecies.9 As such, these four contributions constitute the set of papers most closely related to ours.10 There are, however, several important differences between the present and the above-mentioned papers, particularly since the main contribution of our paper is not the idea of gender discrimination as a result of self-fulfilling prophecies in itself, which is by now a fairly well-known result, but rather to explore more thoroughly how this type of explanation for observed gender differences in the labor market may have implications for public policy. Arguably, a main limitation of the above-mentioned papers (with the exception of Bjerk and Han, 2007) is that all workers are equally talented. This implies, for example, that there is no waste of talent when men rather than women are placed on the fast-track. Thus, models of labor market discrimination without heterogeneous workers seem to miss an important aspect of the problem and makes a welfare discussion of anti-discrimination policies less interesting, in our view. Bjerk and Han (2007) do assume that workers are heterogeneous with respect to labor market productivity, and the authors analyze different kinds of child care subsidy schemes to reduce gender inequality in the labor market. However, although their model is generally characterized by multiple equilibria, only one single (asymmetric) equilibrium is considered, on which the entire policy analysis is based. By contrast, in the present paper we carefully characterize the potential coexistence of symmetric and asymmetric equilibria. As previously discussed, this allows us to tackle the question of whether anti-discrimination policies need to be permanent or not in order to produce a stable outcome without gender discrimination in the labor market.11 With the exception of Francois (1998), permanent versus temporary policies is not an explicit issue in either of the above-mentioned papers. Albanesi and Olivetti (2009) do not analyse policy at all. Dolado et al. (2013) do analyse different types of anti-discrimination policies, but do not discuss stability properties of the multiple equilibria. Thus, in stark contrast to our own findings, their discussion of affirmative action policies implicitly suggests that such policies need only be temporary.12 In contrast to the other papers, Francois (1998) explicitly addresses the issue of equilibrium stability and shows that temporary affirmative action policies can move the economy from a discrimination to a non-discrimination equilibrium (if the latter equilibrium is stable). In a model where firms might discriminate against women when hiring workers to “good jobs”, he characterizes conditions when a temporary policy can permanently shift the economy from a discriminatory situation to one without discrimination. We ask the same question as Francois, but our answer is opposite. The key difference between Francois (1998) and the present paper is that Francois assumes all workers to be homogeneous with respect to skill. With the added assumption that there are always more than enough workers of each gender to fill all the good jobs, a discriminatory equilibrium is one in which all the good jobs in the economy are given to men. It follows that such an equilibrium is always locally stable.13 Temporary anti-discrimination policy can therefore work whenever a symmetric equilibrium exists. In contrast, using a somewhat similar type of model, but where workers differ in skills, the present paper concludes that, under a fairly general set of assumptions, asymmetric (discriminating) equilibria do not co-exist with a stable symmetric (non-discriminating) equilibrium. Thus, we conclude that temporary non-discrimination policies are unlikely to work. This we see as our main policy message. The remainder of the paper is organized as follows. In Section 2 we present our model. Section 3 contains analyses of different means of battling self-fulfilling-prophecy discrimination, whereas welfare issues are addressed in Section 4. In Section 5, we extend the basic model along different dimensions. Finally, some concluding remarks are presented in Section 6.