The Second-Best of Specific Worlds

Chattopadhyay and Duflo (2004) (hereafter C & D) contains a formal model with a striking normative implication: “welfare” (in the paper’s narrow sense) can rise when you ban half the people from running for office. As they write, “...reservation may make the median voter… better off...”

The model is an interesting thought experiment in the citizen-candidate tradition (Osborne–Slivinski; Besley-Coate). But like most thought experiments, it earns its insights by spending a lot of realism. The key question is not “is the model true?” but “what assumptions does the provocative welfare result actually require?”

The Model

In the standard citizen-candidate setup, citizens care about implementing their preferred policies and can do so by running for office. Entry is costly. Utility falls with the distance between implemented policy and a citizen’s ideal point. Once elected, politicians implement what they want: they do not care about rents, re-election, party advancement, reputation, or anything besides policy.

The model also assumes a one-dimensional policy space, rational and fully informed voters, and no credible commitment to campaign platforms. Selective entry therefore matters because electoral competition does not force candidates toward the median.

C & D add three ingredients:

1. Gendered entry costs. Women face higher costs of running than men.
2. Gendered policy types. Women’s ideal points lie on [0,W], while men’s lie on [M,1]. The authors themselves call this assumption “quite extreme”; their descriptive evidence also suggests substantial overlap in men’s and women’s preferences.
3. Right-biased capture. Political influence pulls implemented policy toward μ, with μ > m, where m is the median voter’s ideal point.

Given the setup, three results almost immediately fall out:

1. Without reservation, women may not run even if the constituency has “pro-women” preferences, because the private cost of entry is too high relative to the benefit of winning.
2. Reservation shifts policy left by forcing the candidate pool to be female (a direct implication of the gendered-support assumption).
3. Reservation can raise the median voter’s utility when the candidate closest to the median is more likely to be female, including cases where women outnumber men, and cases where reservation provides a “counterweight” to right-biased capture.

Criticism and the Growth of Knowledge

1. Gender Must Change the Candidate Menu. C & D assume women’s ideal points lie on [0,W] and men’s on [M,1]. The authors themselves describe this as “quite extreme,” and their descriptive evidence suggests substantial overlap in men’s and women’s preferences.
   
   If men and women instead draw from the same distribution of ideal points, the identity channel disappears: reserving the office for a woman does not systematically change the policy type selected.
   
   A more plausible case is overlapping distributions with different average priorities. Reservation could still change policy in that world. But the welfare result becomes less automatic. If men can also occupy the left side of the policy space—and face lower entry costs—then some men can supply the same counterweight to rightward capture. The separation of supports therefore does important work: it rules out precisely those male substitutes.
   
2. Entry Costs Must Keep Women OutThe second ingredient is the entry-cost wedge. It is not enough that women have different preferences; the women whom the median voter would like to elect must also fail to appear under GEN.
   
   Proposition 1 provides sufficient conditions under which no woman runs under GEN. One of those conditions is:
   
   $\delta_w - \sqrt{5} \delta_m > \mu - m$
   
   Here μ is the policy preferred by the local elite, and the model assumes μ > m, so μ − m measures the rightward pull of capture relative to the median voter. As that distortion grows, the inequality becomes harder to satisfy. Stronger rightward capture makes a leftward counterweight more valuable, which also gives women stronger incentives to enter under GEN. Thus, the very force that makes reservation valuable as a counterweight makes the model’s “no woman runs” condition harder to sustain.
   
   Treating the entry-cost wedge as fixed is also questionable in a system with repeated reservation. A woman who has campaigned or held office may acquire name recognition, administrative experience, political networks, and knowledge of how to run. Female candidacy may also become less socially novel. Reservation itself should therefore tend to change δw.
   
   A natural dynamic extension would make women’s entry costs depend on prior political exposure rather than treating each election as if the barrier began anew. If reservation erodes
3. Capture Must Be Right-Biased and Fixed. Capture does much of the remaining work. Because μ > m, political influence moves policy to the right of the median voter, creating room for a female politician to serve as a leftward counterweight.
   
   But μ is treated as fixed and exogenous. C & D themselves note that reservation could change who is able to influence the Pradhan. If electing a woman makes it easier for women to exert influence, reservation may shift μ itself. The model is then no longer comparing different leaders under the same capture process.
   
   There is also a useful linguistic point here. The paper says lobbying “moderates” women while making men “more extreme.” But lobbying pushes everyone toward the same right-biased μ. It is called moderation for women because women are assumed to begin to the left of the median, and extremism for men because men begin to the right. The language follows from the assumed geometry.
4. Candidate Preferences Must Translate into Policy. The citizen-candidate framework assumes politicians cannot credibly commit to platforms and, once elected, implement policies close to their own ideal points. That assumption is what makes candidate identity translate so directly into policy.
   
   Real political systems contain at least partial commitment mechanisms: repeated elections and career concerns, party reputation and discipline (Cox and McCubbins 1993; Rahn 1993), family political reputations, and punishment by voters for visible departures from promises. C & D themselves note that candidates are generally nominated by political parties and report that 17% of women Pradhans in reserved GPs had a spouse previously elected to the Panchayat.
   
   If politicians can make even partially credible commitments, electoral competition can pull policy away from the candidate’s private ideal point and toward voters’ preferences (Alesina 1988). The stronger these mechanisms are, the less policy depends mechanically on selecting exactly the “right” citizen-candidate type.
   
   The model also treats voter preferences as fixed. Politics is therefore largely a matching problem: voters and potential candidates have given ideal points, and institutions determine who enters. But politicians also try to change voters through persuasion, framing, and information.

Taken together, the assumptions show how specific the "welfare" result is. Gender must predict policy type; entry costs must keep the relevant women off the ballot under GEN; capture must pull policy in the right direction; and candidate preferences must translate sufficiently directly into policy.

Under those conditions, reservation can improve the median voter’s outcome by introducing candidates who would otherwise not enter and whose preferences counteract an existing distortion. This is a second-best result: an additional restriction on political choice can offset distortions already present in the system.

But that immediately raises the more natural policy comparison. Why correct high female entry costs by excluding men from the ballot rather than reducing the entry-cost wedge itself? Why counteract capture through candidate restrictions rather than reducing capture? Why rely on selecting the right policy type rather than strengthening accountability?

Related: https://www.gojiberries.io/first-assume-no-elections/