In 2017 I published (ungated) a review of active labor market programs in developing countries, in which I concluded that the “accumulated evidence suggests that wage subsidies are unlikely to be very effective in generating additional employment under standard labor market conditions, and may also even not be very effective in playing a distributional role in determining which individuals get to access jobs”. This was based on experiments in Argentina and South Africa in which very few firms hired workers with wage subsidy vouchers, an experiment I had done in Jordan in which firms hired workers, but then had the large impacts disappear as workers quit or firms fired workers after the subsidy period ended, and an experiment I had done with small firms in Sri Lanka, which likewise found temporary impacts that fell off after the subsidy.
This set of experiences had made me pretty pessimistic about the effectiveness of wage subsidies. However, several more recent papers have led me to update my priors a little bit. I thought I’d summarize some of these results, and some of my reflections on these.
Wage subsidies to get firms to hire women in Pakistan – changing who is hired
An experiment by Bussolo et al. 2025 in Pakistan worked with a sample of 1,227 small, medium, and large firms on an online job portal. They recruited firms that had already decided to advertise a job, and had not expressed a gender preference. The treatment group of 657 firms then got a six-month subsidy, using BDM to elicit the subsidy amount firms would need to hire a woman instead of a man - finding on average firms needed a 15% subsidy. This led to a 10.7 percentage point increase in the likelihood of hiring a women in treated firms. Note that only 53% of firms that were eligible for the subsidy actually took it up. There is some evidence that these impacts last: 1 month post-subsidy, 72% are still with the firm. Longer-term impacts are much more tentative: their endline survey at 1.5 years had really high attrition during Covid (only 27% response), with the firms interviewed more likely to still have a woman in the advertised position, but having no significant increase in the share of women in overall employment. Treated firms were more likely to not specify a male preference in subsequent ads.
Reflections: this study differs from previous ones by screening explicitly on firms that had decided to make a hire, and then using subsidies to change who they hire. The study does not show that wage subsidies can create more jobs, but suggests that relatively small subsidies may be needed to change who firms hire in jobs that they didn’t have strong worker preferences for already – and that there can be some learning from doing so that changes preferences over later hiring. But this learning does not increase overall demand for labor.
Wage subsidies for Mexican secondary school graduates – making formal work more likely for those subsidized
An experiment by Abel et al. 2024 worked with 1,924 secondary school graduates in San Luis Potosi, Mexico. 80% of them had attended schools with a vocational focus. A treatment group of 970 students were offered a six-month subsidy of around $45/month (equivalent to 17% of the average formal sector wage for them), on the condition that they worked in a formal sector job. Here the payment goes straight to the worker, rather than the employer. They link these students to the social security data (IMSS) and track them for 2 years. After a year they find a 4.2 percentage point increase in the likelihood that the individual is working in a formal sector job, with this impact the same after 2 years (relative to only 29% of the control group having a formal job). Using an endline survey with 75% response, they find that some of this may come from a reduction in informal employment, but a point estimate of 3 percentage points higher employment. Monthly income is 9.7% higher after two years. The authors argue that a key channel here is that these students underestimate the wage growth possible in formal sector work, and have high discount rates, and so have reservation wages that are too high. The subsidy leads to a reduction in reservation wages for formal work, getting them to gain experience and wage growth in this sector.
Reflections: this study shows how a wage subsidy can help overcome a market failure that arises from two factors – inaccurate expectations that lead to reservation wages being too high, and second, the wedge between formal and informal wages that arises from the design of the tax and social protection system in Mexico. The subsidy helps low-skilled youth who may not try formal employment otherwise to gain experience, and wages rise enough with experience so that a temporary subsidy may have a lasting impact on sector choice. It is less clear that this generates more jobs, rather than just changing who firms hire, although the authors argue that firms in the area have difficulty filling vacancies so it may have helped fill some jobs that might otherwise be vacant.
Big effects in a small sample of long-term unemployed in North Macedonia
Armand et al. (2026) do a small experiment within a program for the long-term unemployed in North Macedonia that provides employers a subsidy covering half the wage payments during the first year of employment. They have 153 unemployed individuals, and 80 of them get matched to interview for one of 22 job vacancies with firms that would use the subsidy. Those randomly selected for an interview are 14 p.p. more likely to be formally employed 3.5 years later.
Reflections: this combines the joint impact of being matched to a job and the subsidy – the impact seems to be only for those who got offered the job (an interview alone doesn’t do much) – but does provide at least some suggestive evidence that impacts can last after the subsidy has ended. This study also does not show any impact on job creation.
Subsidies get firms to try hiring new types of skills in Nigeria, and continue to do so after the subsidy ends
The focus in the above studies is on using wage subsidies to help certain types of workers find jobs. But for wage subsidies to generate new jobs, we need them to help firms expand their hiring. One way this could happen would be if it helps firms learn the value of hiring certain types of skills they hadn’t considered before. Stephen Anderson and I found some support for this in an experiment in Nigeria (blogged about here). One of our treatment arms subsidized firms to use a human resources agency to hire a marketing or accounting worker to join the firm, instead of the firm owner doing these tasks themselves. The subsidy lasted 9 months, with the amount decreasing over time. 97% of firms offered the subsidy hired a worker, and a year after the subsidy ended 37-48% of these workers were still in the firm, and firms were more likely to use their own money to go back and hire further HR services. The sample size of only 200 treated firms yields an imprecise and insignificant impact on total employment- but the results are at least consistent with firms using a subsidy to grow certain categories of employment.
Helping firms and workers smooth large aggregate shocks
A final use of wage subsidies has been to help firms and workers during crises. Bruhn uses DiD to find that temporary wage subsidies in Mexico during the 2008-9 recession helped employment in eligible industries rebound more quickly to their pre-crisis levels. Köhler et al. use DiD to find temporary wage subsidies helped South African workers increase their chance of remaining employed in the early Covid period. And Chatri and Tahir use an RDD in Morocco to find temporary wage subsidies helped Moroccan workers who lost their jobs during Covid get reemployed more quickly.
How much do I update my priors based on these studies?
Let’s consider again the conclusion I previously had “wage subsidies are unlikely to be very effective in generating additional employment under standard labor market conditions, and may also even not be very effective in playing a distributional role in determining which individuals get to access jobs”. The above studies only slightly temper the first part of this statement. Perhaps they helped generate additional employment in the Mexican experiment if firms had trouble filling jobs, and in Nigeria by getting firms to learn about new types of worker skills- but neither study can show significant impacts on overall job creation. This is still an area ripe for more research. The impacts during large aggregate shocks speak to non-standard labor market conditions. I would update a little on their ability to shape who gets hired – especially if the subsidies are explicitly linked to job openings through matching or screening as in North Macedonia or Pakistan.
Join the Conversation