In both product and labor markets, when quality is hard to predict prior to purchase, new entrants often struggle to break in because buyers and employers don't know how good they are. This "cold-start" phenomenon is especially true in online labor markets, where hundreds of millions of workers from low- and middle-income countries compete for digital jobs through online freelancing platforms. While these platforms expand LMIC workers' access to global opportunities, most struggle to land work without an established reputation. This challenge is made even more severe by the fact that workers’ credentials, such as education, are less recognizable to foreign employers.
Reputation theory offers a seemingly simple solution: novice workers without previous online job history can offer low wages initially to get jobs and raise them later once they build a reputation. But when we look at job application data, very few novices do this despite intense competition on these platforms. Why don't novices offer low initial wages to gain market entry?
A novice’s wage decision depends on two key beliefs:
- Beliefs about employer demand. Do employers interpret low wages as a signal of low quality?
- Beliefs about their own ability. Can they deliver high-quality work and earn positive reviews that make early investment worthwhile?
In my job market paper, I test these two mechanisms through two field experiments on a leading global freelancing platform and examine what they imply for market efficiency and policy.
What Novices Believe — and Why It Matters
To understand what's driving workers' decisions, we hired 481 novice freelancers from 37 LMICs through the platform. Two things stood out.
First, 44% of our sample believed that bidding below the posted job budget would signal low quality to employers. This is a reasonable concern: extensive evidence from product markets shows that consumers infer lower quality from lower prices when evaluating unfamiliar goods. Novices may therefore avoid the low-wage strategy because they believe it would backfire.
Second, when we hired these novices to complete a standardized data entry task, we found substantial variation in their performance: some did well, others struggled. When asked to guess their performance relative to other novices, very few got it right. If workers are uncertain about whether they can deliver good work and earn positive reviews, they may be hesitant to make the upfront investment of lowering wages to build a reputation.
Demand Experiment: Does Lowering Wages Help Novices?
Motivated by novices' baseline beliefs, we first examined whether employers penalize low wages from novices as low-quality signals. If so, lowering wages should boost demand less for novices than for experienced workers. To directly test this, we created experimental profiles for novice and veteran freelancers with comparable characteristics, except for their reputation (a combination of job history and ratings). We then submitted applications from these profiles to 703 online jobs, randomly varying the wage offers, and tracked employer responses.
We find that although employers consistently favor veterans, lowering wage offers boosts callback rates substantially more for novices: bidding at 80% of the budget increases novice callbacks by 50% (base rate: 3 p.p.) and veteran callbacks by 16% (base rate: 6 p.p.). This shows that employers do not treat low wages from novices as low-quality signals, contrary to what novices believe. In fact, lowering wages improves their chances.
Novices may still avoid this strategy, however, if the loss in immediate earnings outweighs the gain from building a reputation. To quantify the net returns, we conducted a back-of-the-envelope calculation using results from the experiment and platform data. The returns are high: novices who consistently bid 20% below the budget could secure 50% more jobs and increase their first-year earnings by 2.5 times. The demand experiment thus tells us that lowering wages is both an effective and profitable entry strategy, which makes it more puzzling why workers refrain from it.
Freelancer Experiment: How Do Novices' Beliefs Influence Wage Offers?
To test whether correcting workers' beliefs changes their behavior, we worked with the same set of novice freelancers. After they completed the initial data entry task, we cross-randomized two treatment arms at the individual level, stratified by baseline performance.
First, we gave half the sample private feedback on their performance relative to other novices – the feedback treatment.
Second, using a separate employer account, we posted a new job on the platform and referred it to all workers. We then informed half the novices that this employer would not judge their quality based on wage bids – the employer info treatment.
We find strong evidence that both belief frictions affect workers' wage offers, but that they matter differently for high- and low-ability workers. As shown in Figure 1, correcting workers' misperceptions about employer demand (employer-info arm) roughly quadrupled the likelihood of low wage offers for both types. But once workers also received information about their own performance (feedback and combined arms), the treatment effect was driven by the high-performing novices. This is consistent with the prediction of the reputation model: low initial prices can signal high quality because high-ability workers have more to gain from investing in reputation, since they are more likely to be rehired in the future.
Figure 1. Heterogeneous Treatment Effects by Worker Performance
What does this imply for market efficiency? Once these belief frictions are removed, high-ability workers invest in reputation by offering lower initial wages, and the market moves toward an efficient equilibrium where talent is discovered more quickly. To examine whether talent was being left undiscovered at baseline, we compared the performance of novice and veteran freelancers by hiring in-sample and out-of-sample applicants for the referred job. Although the sample is small (N=62), novices on average outperformed veterans. This suggests inefficient talent allocation at baseline, and that correcting novices' misbeliefs could meaningfully improve matching quality in these markets.
Policy Simulation: Can Novices Correct Misbeliefs through Experience?
Given these findings, one might think that novices could simply try different pricing strategies over time and gradually resolve their belief frictions without any external intervention. But the key question is: how costly would it be for novices to learn on their own?
To answer this, we consider the ideal case of a perfectly rational worker and simulate learning about true demand through systematic experimentation across different bidding strategies and Bayesian updating. We find that learning is very slow because employer callbacks are rare, meaning most applications provide little information for workers to update their priors.
Figure 2 shows the distribution of applications needed to learn true demand under different prior beliefs. To be 80% confident in the true demand, a median worker with moderately inaccurate beliefs would need around 200 applications and incur roughly $270 in application fees, equivalent to about a quarter of annual income in low-income countries. This high cost of learning helps explain why belief frictions persist and continue to block market entry.
Figure 2. Worker Learning about True Demand
Recommendations for Policymakers
Many LMIC governments view online platforms as a promising avenue for addressing local job creation challenges by connecting their workers to global opportunities. Most current government initiatives focus on upskilling freelancers, but these have shown limited impacts on long-run success.
Our results suggest that even when novices from LMICs have sufficient skills, a lack of understanding of market fundamentals and their own abilities still prevents them from learning and adopting effective entry strategies. External information interventions, such as mentorship and performance feedback, are therefore crucial to help novices navigate and break into these markets.
About the Author
Ruoxuan (Rebecca) Wu is a PhD candidate in Economics at the University of Chicago.
Her work is the intersection of development and labor economics, with a focus on digital employment. To learn more about her research, visit: https://www.ruoxuan-rebecca-wu.com/
Social Media Handles:
