Estimating the impacts of affirmative action policies in public service hiring in Pakistan: A regression discontinuity approach
This project examines the impact of affirmative action in public service hiring in Pakistan. Governments adopt affirmative action policies to help disadvantaged communities achieve development in line with other groups in society. However, these policies are often criticized because of efficiency losses that occur when a merit-based rule for resource allocation is superseded by affirmative action. This results in performance losses compared to a system without affirmative action.
In Pakistan, recruitment into the civil services occurs through competitive examinations and interviews. Though applicants must score a minimum number of marks in order to qualify for the service, in a typical yearly hiring cycle, there are more qualified candidates than available positions. As a result, a merit list is drawn up and candidates are selected based on their merit score. Affirmative action is practiced through the introduction of quotas for candidates belonging to particular domiciles. In addition, there are also quotas for women and minorities.
There exists limited academic research on a) the consequences of affirmative action on efficiency, and b) causal identification of the effects of such recruitment policies. Efficiency estimation is of crucial importance to debates regarding affirmative action[1], especially in the case of public service.[2] Similarly, work on the effects of recruitment into government has studied local level bureaucrats[3] and their effect on policy. We focus on senior bureaucrats (grade 17+), and consider impacts on a broad range of socioeconomic outcomes.
In this project, we will aim to answer two key questions in assessing the ``success'' of affirmative action. First, what are the redistributive versus the efficiency consequences of the quotas policy per se? And second, what are the effects of public service employment on the societal integration of underrepresented groups versus the potential disaffection of members of groups displaced by the policies (the subject of the second paper)? Our study will involve a mix of descriptive and causal analyses.
The descriptive analysis will follow the lead of past studies such as Bertrand, Hanna and Mullainathan (2009), Bowen and Bok (2000), and Weisskopf (2004) in analyzing the characteristics of those non-quota domicile groups ``placed out'' by the quotas as compared to those ``placed in.'' This first step allows us establish whether the quotas are effective in targeting those who otherwise have fewer opportunities due to less educated parents, lower household income, non-access to privileged networks , and other markers of lower socio-economic status.
The causal analysis will use a regression discontinuity design, motivated by the deterministic aspects of public service recruitment process. Recruits are selected on the basis of their examination score, which serves as the so-called ``forcing variable'' for this analysis (Imbens and Lemieux, 2008; Lee and Lemieux, 2010). For any group subject to admission on the basis of their scores, the relevant ``cut point'' is the mid point between the lowest score among such group members admitted as a result of the examination process and the highest score among such group members not admitted as a result of the examination process.
[1] See, for instance, Holzer and Neumark (2000) and Weisskopf (2004)
[2] Studies have mostly addressed this in the context of education (See Bertrand et al (2008))
[3] See Chattopadhyay and Duflo (2004)