From Information to Search: The Causal Chain So Far

The Delhi experiment now has enough structure to support a useful judgment. We know the starting problem was not one thing: job seekers held imperfect wage beliefs, and they did not have equal access to usable search channels. We also know the treatment was not one thing: it combined personal assistance, a digital platform, and repeated generic or tailored messages. This checkpoint connects those facts into a causal chain so readers can predict what information might change first, where the effect might stop, and why a later employment result cannot be inferred from an early application count.

The key move is to stop asking whether information “worked” in the singular. Information can alter a belief about pay, reveal an opportunity, lower the effort needed to search, or make a known route more relevant. Those changes can lead to different actions even when the participants receive the same broad intervention. The Delhi design gives us a way to keep the actions separate.

The chain has two different levers

The first lever is a person's prior belief about the labor market. Respondents entered the study with aspirations for salaried work and, on average, expected salaries above the urban occupational benchmarks used by the researchers. The gap was larger for men. If credible information narrows that gap, the expected payoff from a costly search can fall. A person who had been spending heavily to pursue a high-paying ideal job may search less, postpone entry to build skills, or reconsider the timing of a decision.

The second lever is access. Women entered with lower employment exposure and were less likely to apply even when they had searched. Men depended more on relatives, friends, and neighbors; women relied relatively more on institutions and digital routes. If information makes a usable channel available where a personal network was incomplete or poorly matched, the expected cost of finding a relevant vacancy can fall. The response may be a change in route rather than an increase in total effort.

These levers can operate together. A woman may have a realistic idea of what work pays but lack a reliable way to find vacancies compatible with her preferences. A man may know how to find leads through his network but overestimate what a desirable job will pay. The same message stream can therefore change the first person's access and the second person's expected return. Calling both responses “motivation” would erase the mechanism.

This gives the experiment a diagnostic logic. If beliefs are the binding constraint, information should first alter expectations and then change search intensity among people whose initial expectations were most misaligned. If access is the binding constraint, the clearest change may be a substitution between channels, with little movement in total search. If the main obstacle is activation, applications should respond quickly after contact, even if employment is too distant or employer-controlled to respond in the same period. These predictions can coexist, but they point to different evidence. A single average measure of “job-search activity” cannot tell them apart.

The treatment also enters the chain at more than one point. The in-person session can make registration possible. The platform can make vacancies visible and offer a way to apply or contact HR. A message can provide information about the market or prompt someone to return. Personalization can make the information more relevant to the recipient's sector, status, or preferences. None of these changes guarantees that an employer has a suitable vacancy, responds, makes an offer, or that the worker accepts it.

That last boundary is central.