Employers under-invest in training their employees in general skills. Evidence from rural Burundi shows that while there are large returns to providing training, these are not captured by the employer who invested in it due to short contract durations.
Most labour contracts in low- and middle-income countries are short-term and informal. In rural economies, labour is traded almost entirely via spot labour contracts, with minimal long-term contracting (Rosenzweig 1988, Kaur 2019), while even among urban employers, there is more limited attachment between employers and employees than there is in higher-income countries (Blattman and Dercon 2018, Donovan et al. 2023).
Do employers invest enough in training their employees?
In our research (Cefala, Naso, Ndayikeza and Swanson 2024), we investigate a possible downside of labour markets that rely on short term contracting: there might be too little investment in workers’ general human capital, because employers may not want to invest in training if workers are likely to leave soon after. This hypothesis, if correct, might help to explain why worker productivity is low and grows little with experience in LMICs (Lagakos et al 2018).
Testing this theory is challenging. Employers might not train workers because they often leave for other jobs (limiting the employer’s returns to providing training), or because training only generates low returns for the employer, even if the worker stayed at the employer after being trained.
In our research we focus on the decisions of rural employers who live in Burundi and under-invest in training their workers in a new planting technology. To establish the reasons underlying under-training, we conduct two experiments. In the first, we test whether employers who train workers capture only a small fraction of the total returns from training, as these workers are likely to work elsewhere after being trained. In a second experiment, we test whether employers in a labour market where workers are less likely to leave after training are more willing to invest in training.
Agricultural technologies and labour markets in Burundi
We conduct our experiments in rural Burundian villages in collaboration with the country office of the NGO One Acre Fund (1AF). In these villages, one can characterise farmers as employers (smallholder farmers with relatively more land who farm and hire labourers to work on their farms) and labourers (also smallholder farmers but with less land who farm their own lands and also work for others). This setting enables us to measure the spillovers from training: since villages are geographically isolated and self-contained labour markets, we can measure both the direct and indirect (spillover) returns from training investments.
We focus on farmers’ decisions of whether to adopt row planting - an agricultural planting technique that has been shown to increase yields substantially, and which the NGO 1AF has promoted adoption of for around a decade among its clients in Burundi. Employers in our context are likely to be members of 1AF and, therefore, receive training in the technique by the NGO. Labourers, by contrast, are less likely to be members of the NGO, and, at baseline, many do not know how to row plant and consequently have not adopted row planting.
Adoption, even among the employers who know the techniques, remains limited: while many employers adopt the techniques on some fields, many do not adopt the technique on all of their fields. Time is often cited as a limiting factor for further adoption: row planting is more time-intensive than broadcasting seeds, and farmers perceive high returns to planting quickly after harvest. Given the labour intensity of the techniques, most employers in our context cite a lack of labourers who know how to do the techniques well in their villages as limiting further adoption. Despite this, few employers report ever training labourers. When asked why they do not train workers, most employers state that if they trained a labourer, they wouldn’t get any benefit since she would be in high demand by other employers or become less available, opting to use the techniques on her own fields.
Training generates returns for others
Our first experiment tests whether employers do not capture all of the returns to training, because training generates spillovers to other employers. We work in 80 villages, approximately half of which are assigned to a control and half to a treatment status. In all villages, we recruit around 20 trainer-employers to identify 20 trainees – labourers in the village they would be willing to hire who did not know row planting. In treated villages, we offer financial incentives to these trainer-employers to train their labourers in row planting (whereas in control villages we provide no incentives to train). To measure the spillovers from the training, in every village we also sample around 20 “spillover employers” – farmers in the village who regularly hire labour but who do not train labourers themselves. In total, we sample more than 3600 households falling into these 3 categories (trainer-employers, trainees and spillover employers) to measure who captures the returns to training.
We find that providing financial incentives induces farmers to train, and this training is effective in upgrading the skills of workers. Trainees in treated villages are employed for 3.5 more days doing the trained techniques during the planting season (from a baseline of 0.5 days) but, consistent with the theory that training generates only limited returns for the training employer, they work not only for the farmer who trained them, but also for non-training employers, and also apply the skilled tasks on their own farms. While trainer-employers in treated villages hire 46% more days of labour for row planting and adopt row planting on 19% more fields, they are also more likely than trainer-employers in control village to say that they were unable to hire the trainee whom they invited to the training event. Consistent with this, we estimate a large labour market spillover to other hiring farmers in the same village. Specifically, employers in training villages, uninvolved in the training, also hire 55% more days of labour to work on their fields doing the techniques taught in training and adopt row planting on 24% more fields.
To understand whether the previous level of training in the economy is inefficient, we estimate treatment effects on farm profitability and labour market earnings and compare this to the cost of providing training. Training is a high return investment: we estimate that a dollar of training investment generates more than two dollars of returns. However, consistent with employers capturing only a small proportion of the total surplus, trainer-employers capture only one-quarter of the total returns we estimate, with the remainder captured by non-training employers and workers.
Longer contracts facilitate training investments
In our second experiment, we test whether employers choose to train a worker when it becomes more likely that she would return to work for the employer after training. We conduct a second field experiment – the “Labour Guarantee Experiment” – with 200 other farmers. In this experiment, we utilise our relationship with 1AF to create variation in contract structure, by promising workers credible future payments if they return to work for a particular employer during the planting season, creating more certainty for some employers that a particular worker will come to work for them during the planting season. To implement this treatment, we offer conditional cash payments to labourers which make it more likely that they will return to work for a farmer who provided training at planting time. We then observe how likely it is that farmers with and without this “labour guarantee” are to train workers before the planting season. Consistent with the hypothesis that farmers’ investments are limited by weak attachment, farmers and their hired labourers who receive the experimental contract are more than 50 percentage points more likely to train for 3 hours or more (the estimated minimal amount of time needed to train.
Policy implications
Our research provides evidence that training investments can have large returns, but that these investments may be constrained by labour market frictions. On its own, this an important finding given that the literature on training has generally found very mixed results (Kondylis et al 2017, Aker and Jack 2023). This evidence also contributes to a debate on how to spur technology diffusion in agricultural settings in low-income countries. In line with a growing body of work, our paper suggests that labour market frictions might limit the adoption of new technologies (Jones et al. 2022). Finally, we provide further evidence that the transmission of new technologies may not be seamless, and show that the incentives of those who hold information should be considered when designing policies seeking to diffuse technologies (BenYishay and Mobarak 2019, Chandrasekhar et al 2022).
References
Aker, J C, and B K Jack (2023), "Harvesting the rain: The adoption of environmental technologies in the Sahel", Review of Economics and Statistics, 1-52.
BenYishay, A, and A M Mobarak (2019), "Social learning and incentives for experimentation and communication", The Review of Economic Studies, 86(3): 976-1009.
Blattman, C, and S Dercon (2018), "The impacts of industrial and entrepreneurial work on income and health: Experimental evidence from Ethiopia", American Economic Journal: Applied Economics, 10(3): 1-38.
Cefala, L, P Naso, M Ndayikeza and N Swanson (2024), "Under-training by Employers in Informal Labor Markets: Evidence from Burundi ", Working Paper.
Chandrasekhar, A G, E Duflo, M Kremer, J F Pugliese, J Robinson, and F Schilbach (2022), "Blue spoons: Sparking communication about appropriate technology use", National Bureau of Economic Research Working Paper No. 30423.
Donovan, K, W J Lu, and T Schoellman (2023), "Labor market dynamics and development", The Quarterly Journal of Economics, 138(4): 2287-2325.
Jones, M, F Kondylis, J Loeser, and J Magruder (2022), "Factor market failures and the adoption of irrigation in Rwanda", American Economic Review, 112(7): 2316-2352.
Kaur, S (2019), "Nominal wage rigidity in village labor markets", American Economic Review, 109(10): 3585-3616.
Kondylis, F, V Mueller, and J Zhu (2017), "Seeing is believing? Evidence from an extension network experiment", Journal of Development Economics, 125: 1-20.
Lagakos, D, B Moll, T Porzio, N Qian, and T Schoellman (2018), "Life cycle wage growth across countries", Journal of Political Economy, 126(2): 797-849.
Ma, X, A Nakab, and D Vidart (2021), "Human capital investment and development: The role of on-the-job training", University of Connecticut, Department of Economics Working Paper.
Rosenzweig, M R (1988), "Labor markets in low-income countries", Handbook of Development Economics, 1: 713-762.