Creating jobs by helping small and medium firms understand labor law


This page in:

If you are going to go with a one-word title for your paper or talk, Chang-Tai Hsieh’s keynote entitled “Jugaad” is definitely a much more interesting and memorable one than most. Jugaad is a colloquial Hindi term that he defined as “an innovative fix or workaround by bending rules”, and he used it to describe how firms would get around burdensome regulations, particular those surrounding labor law.  My impression with talking with firms in several countries has often been that they indeed find workarounds to regulations that on paper look like they could be costly. For example, I remember talking with the owner of a garment firm in Sri Lanka, and asking how he dealt with regulations that, starting at a firm size of 15 employees, made it harder to fire workers. The owner replied “I never need to fire them, if there is someone I want to have leave, I assign them to clean the toilets and do other such tasks, and they always quit”. This possibility of often lax enforcement of the law, and creative workarounds, made me wonder whether such regulations really had much impact on the hiring and growth of small firms.

A recent paper by Marianne Bertrand and Bruno Crépon that is forthcoming in the AEJ Applied (ungated)  has led me to update my priors on this, and shows that in addition to policy efforts to help improve the laws and make regulations simpler, there can also be important benefits for firm growth of helping firms better understand and deal with the existing laws.

Helping firms understand the labor laws in South Africa

The setting is South Africa, which has a very unusual labor market, with high levels of unemployment and surprisingly low levels of self-employment and informality. The authors note that “there is widespread perception (particularly amongst employers and within the media) that it is more difficult to dismiss an employee in South Africa than virtually anywhere else in the world” and that firms, especially SMEs, cite labor market regulation as a key obstacle to business and constraint to hiring. But they note that these perceptions might be misplaced, in part because labor laws had changed relatively recently, and in part because relatively rare and complex labor cases would get the most publicity, while those resolved easily would be less visible.

They conducted a randomized experiment in 2013 with 1,824 SMEs with between 10 and 300 workers (mean 78 workers, 47% with <50 workers) to test the effect of helping firms to better understand the labor law. Baseline knowledge of labor law was low, with only 18% knowing how much salary they would need to pay to workers for unfair dismissal, and knowing conditions needed for validity of an employment contract. They were divided into:

·       a control group of 912 firms, and

·       a treatment group of 912 firms. The treatment group received a 21-week membership to a labor law club that normally charged $40/month. This club, run by labor law experts, sends out twice-weekly newsletters that focused on labor law and human resources management, and also provides access to a website with details, videos, and a discussion forum focused on South African labor law.

They then conduct an endline survey 6 months later, which had a 83% response rate.

One nice feature from a measurement perspective is that because the intervention is online, lots of details of take-up can be observed – opening rates for each newsletter, whether firms logged into the website, watched video tutorials, posted on the discussion forum etc. 70.6% of employers opened at least one newsletter and 36.7% did at least one such action on the website. In the endline, the authors include 7 quiz questions related to this content, and find no treatment impact on knowledge according to this measure. This is in line with what is sometimes seen in business training and financial education, where knowledge-based test questions are hard to improve. What should matter more is whether firms update their beliefs about how much trouble labor regulations could cause for them. On this, the authors find some changes, although relatively small in magnitude:  treated firms are 3 to 5 percentage points less likely to say that labor is a constraint to hiring at endline.

The treatment effect on employment is stunning to me: a 11.8 worker increase in employment, or 12-15% increase relative to the control mean. Almost all of this increase seems to be for the categories of permanent and fixed term workers, rather than casual workers. The CDFs show increases across the distribution in both endline levels (left panel of Figure 1) and changes (right panel).

Figure 1: CDFs of Endline Employment and Employment Changes.

CDFs of endline employment and the change in employment

However, when the firm is asked about the number of workers hired and number dismissed in the past 6 months, the coefficients are only around 1 worker, and not significant. Figure 2 shows that there is only a weak correlation between the change in total employment from baseline to endline (x-axis) and the net increase in employment as calculated by reported hiring in past 6 months less reported firing. The authors chalk this up to measurement error in the hiring and firing data, and firms being unable to recall timing. They also note the hiring and firing will not include voluntary quits – so perhaps treated firms are behaving like my Sri Lankan interviewee, and now they know the labor law, feel more confident in “encouraging” workers to quit. The authors are to be commended for presenting these conflicting measurements, but nevertheless, this should definitely caveat the size of the estimated effect.

Figure 2: A low correlation between the changed in total employment, and the net reporting hiring.

Change in hiring minus firing is not very strongly correlated with change in total employment

A second point to note is that this figure shows how large the changes in employment can be in firms – even over this short period, we see lots of firms losing 50 or more workers or gaining 50 or more workers in total employment. This is the big challenge for detecting employment effects – a treatment-induced change of 2 or 3 workers would be really hard to detect amongst all this variation – having a large sample size and big treatment effect is then crucial here.

The endline survey was done online and by phone, and so the response rate of 83% is pretty good for this form of survey. But a downside is that it prevents any checking of how many employees can actually be observed at the firm and thus to better understand the reporting differences. The authors note that good administrative data on employment is with the South African Treasury, but they were not able to get access to these data. Another issue is that six months is pretty short-term – perhaps the intervention just accelerated hiring the firms were planning on doing anyway, and so it would be good to have been able to track employment over longer time horizons – something else the administrative data would be helpful for.

The cost per firm of the intervention was only $200, so that the cost per job created is amazingly low (less than $20). This would likely make the intervention worth doing even if the impacts are only transitory. It definitely seems worthwhile to replicate this study in other contexts as a result, and to see whether these results generalize elsewhere, or whether firms are indeed usually able to Jugaad their way around labor regulatory concerns.



David McKenzie

Lead Economist, Development Research Group, World Bank