The Hidden Economics of a Job Search
The Delhi experiment is easiest to understand if we begin before anyone received a message or opened a job platform. This chapter examines what job seekers already believed about their prospects, what those beliefs were measured against, and how access to information shaped the routes they could use. The aim is to distinguish ambition from expectation and lack of interest from lack of usable information—distinctions that let us predict why the same intervention might lower search for some people while redirecting it for others.
A job search is often described as if it were a simple expression of motivation: people who want work search, and people who search eventually find it. That picture leaves out the economic calculation underneath. A person chooses how much time and money to spend looking partly from an estimate of what opportunities are available, what they pay, how likely a response is, and whether the route to those opportunities is open to them. Those estimates can be wrong. The cost of being wrong is not merely intellectual. It can mean spending months pursuing unlikely jobs, abandoning a viable search route, or failing to look because the available options never become visible.
Aspirations are targets, not forecasts
The respondents in the study had high aspirations. About 69% said they were already in, or moving toward, their ideal career path. Salaried work was the dominant career goal: 80% of women and 67% of men selected it. The gender contrast within that category was also meaningful. Women were more likely to prefer public-sector or government salaried work, while men were more likely to prefer self-employment.
These answers describe desired destinations. They do not tell us what a respondent expected to earn next month, how likely they thought a particular job was, or whether they had a realistic route to it. Someone can want a stable salaried job while knowing that obtaining one will be difficult. Someone else can believe that a desirable job is nearly within reach while having little evidence for that belief. Treating both people as equally “optimistic” because they name the same goal would confuse aspiration with expectation.
The distinction matters for interpreting information. An aspiration is a standard against which a person evaluates work. An expectation is a forecast about what is likely to happen. An observed outcome is what did happen. The experiment measures all three kinds of thing at different points, but none can stand in for the others. A high preference for salaried work does not prove that the labor market offers enough salaried positions. A downward change in the belief that one is on track for an ideal job does not prove that the person has abandoned the goal. And a later employment outcome cannot tell us, by itself, what the person believed when they chose how hard to search.
What “upward-biased” means here
To study expectations, the researchers asked respondents to estimate salaries for a range of jobs listed on the partner platform. They compared those answers with average weighted monthly earnings for comparable occupational categories in urban India, using the 2022–23 Periodic Labour Force Survey as a benchmark. The comparison covered salaried roles such as accounts keeper, primary-school teacher, data-entry operator, and hospital attendant or nurse, as well as casual roles such as garment-factory worker and electrician or fitter.
Expected salaries were higher than the benchmarked actual earnings on average for both salaried and casual work. The average gap across the listed occupations was INR 6,714 for women and INR 7,878 for men. For salaried work, the corresponding gaps were INR 7,489 and INR 8,608. The differences between men and women were statistically significant, with men's expectations farther above the benchmark.
In this study, an upward-biased expectation therefore means that a respondent's average expected salary exceeded the average weighted earnings used as the comparison. It does not mean that the person was wrong about every vacancy, or that a higher-paying employer could not exist. The benchmark is an occupational average, not an offer made to a particular worker. It is useful because it reveals systematic miscalibration across the sample, but it cannot measure the exact probability of one individual's success.
That limitation is also what makes the measure analytically valuable. The researchers do not need to prove that every respondent had the same error. They need to establish that search decisions began from different beliefs about the payoff to work. If a person believes a preferred occupation will pay substantially more than it typically does, the expected return from intensive search may look high. If information brings that estimate closer to the benchmark, the person may decide that another day of applications is less worthwhile than previously thought.
A reduction in search can follow without any reduction in the person's desire for work. It may reflect a revised forecast, a decision to acquire skills first, a choice to remain in education, or acceptance that a current job is better than the imagined alternative. Whether that response is beneficial depends on what happens next. Less wasted effort and a better plan are different from discouragement and withdrawal, even if both first appear as fewer applications.
The flowchart separates two inputs that are often folded into the vague idea of “motivation.” One is what a person thinks the market will reward. The other is what information routes are actually available. The first can influence how attractive search seems; the second can influence how costly it is to carry out. A person can have high expected returns but poor access to vacancies, or realistic beliefs but no convenient way to act on them. The intervention was positioned at the intersection of both problems.
The diagnosis is miscalibration, not ambition
The baseline pattern is best described as a combination of aspiration, imperfect information, and unequal exposure. It would be too simple to say that respondents merely expected too much. Their goals also reflected the kind of security and progression they wanted from work. Salaried employment, especially in the public sector, can represent stability, predictable compensation, and social status. A comparison with average wages tells us that those goals may be difficult to reach, not that they are unreasonable to hold.
Nor does a higher expectation automatically make a person a better or worse job seeker. Expectations are useful when they help someone compare the cost of search with the likely benefit. They become a problem when they are systematically detached from available evidence. Someone who expects a high salary may reject lower-paying but viable work, spend longer waiting for a narrow type of vacancy, or apply with less urgency because success seems assured. Someone who expects too little may stop looking before discovering a better option.
The paper's later analysis uses this distinction to interpret treatment effects. Among people with upward-biased expectations, the intervention could plausibly reduce the perceived gain from searching. Among people with downward-biased expectations, the same information might raise the attractiveness of looking, although the source finds that this response is not uniform. The important prediction is conditional: information changes behavior through the belief it revises, not through a universal rule that accurate information always energizes or always discourages.
That is why the study's baseline survey is more than background description. It establishes the priors with which participants entered the intervention. Without those priors, a fall in search would be ambiguous. It could be a harmful loss of confidence, a rational response to a lower expected payoff, or a change in priorities. The wage-mismatch measure does not settle the question on its own, but it gives the later patterns an observable starting point.
Access to information is a separate constraint
Beliefs are only one side of the diagnosis. The other is whether a person can obtain usable information at a reasonable cost. Women in the sample were much less likely to be employed at baseline—about 25%, compared with 50% for men. Among respondents who had searched for work, 69.4% of women had applied for a job, compared with 74.3% of men. These differences do not establish lower demand for work. They show that the path from wanting a job to making an application was not the same for everyone.
The available search channels differed as well. Men were more likely to depend on family, friends, and neighbors for job information and referrals. Women relied relatively more on formal educational institutions and digital platforms. Among people who had ever applied, 61.6% used some offline mode and 34.5% used an online or digital mode. Offline routes therefore remained important, but they were not equally accessible or equally useful to every respondent.
A network is not a neutral piece of infrastructure. Its value depends on who is in it, what they know, whether they can make a credible referral, and whether the information reflects the job seeker's own preferences. A relative who knows about a delivery job may not know about a public-sector opening, a vacancy with safe working conditions, or a position compatible with household responsibilities. The paper discusses the possibility that information women received through social networks was often generic and shaped by the experiences and preferences of men in those networks. This is a proposed explanation, not a separately randomized finding, but it clarifies why “just use your network” is not a neutral solution.
Digital search can help by widening the set of visible opportunities and reducing some costs of repeated searching. It can also introduce new frictions: registration, profile creation, skills tests, and the effort of sorting through listings. The partner platform attempted to lower some of those barriers through in-person assistance and features such as location filters, direct contact with human-resources staff, and employer verification. But the experiment is not a catalog of platform features. The relevant question is whether the intervention made a usable route to relevant work more available, and for whom.
Women's stronger interest in the platform is consistent with that interpretation. When the treatment groups were offered registration, women expressed more interest and registered at higher rates than men. That does not prove that women valued digital work more, or that the platform was equally suitable for all of them. It does suggest that an opportunity to obtain information through a different channel met a constraint they faced at baseline. The same intervention could therefore lower the cost of search for women without increasing the total time they spent searching.
Two predictions from one starting point
Putting beliefs and access together produces two different predictions. For a person whose expected salary is well above the market benchmark, credible information may lower the expected return to costly search. Search intensity can fall, especially if the person had been pursuing a narrow ideal job with confidence that it was attainable. The relevant outcome is not the number of applications alone but whether the revised plan leads to training, a better existing job, or disengagement.
For a person with fewer reliable information routes, credible vacancy information may lower the cost of finding opportunities. Search intensity need not rise if the person replaces one route with another. Offline contacts can decline while digital search grows; time and money can be reallocated without increasing the total search effort. In that situation, a stable amount of search may become more accessible or more closely aligned with the person's preferences.
The Delhi sample makes these predictions especially consequential because women began with lower employment and less reliance on traditional networks, while men began with stronger upward wage expectations and greater use of social ties. Gender is not the mechanism by itself. The mechanism is the interaction between gendered exposure, prior beliefs, and available channels. The paper also notes possible differences in marriage, children, and time available for search, but these are contextual explanations rather than individually identified causes.
This diagnosis sets a demanding standard for the intervention. It should not be judged only by whether people registered or by whether applications increased. A successful response could be belief calibration, a change in search route, a decision to invest in education, or a better occupational match. Conversely, more activity could simply mean more effort spent on poor leads. To understand which interpretation fits, we first need to see exactly what information and assistance the experiment supplied, and how its generic and tailored versions differed.