After finding that gifts of capital had no impact on the growth of female-owned subsistence firms in Sri Lanka and Ghana, in a recent paper (with Chris Woodruff and Suresh de Mel) I test whether business training can help poor women in Sri Lanka start and grow their businesses. A new 2-page impact note summarizes the results of the full paper. The even shorter version is that the ILO’s Start and Improve Your Business Training:
- has no significant impact on the survival or growth of existing subsistence enterprises run by women in either the short or the medium run.
- Speeds up entry into business of women out of the labor force who would like to start businesses, although the control group catches up after a year.
- Leads to less analytically skilled women starting new businesses.
- Makes the new businesses started by women more profitable
You may ask whether this is just a Sri Lanka effect. In a recent review paper, Chris and I examine all existing experimental evaluations of business training programs in developing countries (Markus Goldstein provides a nice summary in the Development Impact blog). Similar findings of a lack of significant impact of business training on the profits or sales of existing female-owned firms are found by Berge et al. in Tanzania, Gine and Mansuri in Pakistan, and Karlan and Valdivia in Peru. Lack of statistical power is a problem, but given the low profits earned by these subsistence firms, the percentage gain from training has to be pretty large to justify the costs of training. One exception is a study by Calderon et al. in rural Mexico, who claim to find significant positive impacts on profits and sales, although 50 percent of their non-attriting sample has closed by the time of their second follow-up survey.
So is training futile?
Let’s think about how business training might affect an existing subsistence business. We can express the production output (Y) of an enterprise as a function of its capital (K), labor (L), and efficiency at using this capital and labor (A), so that:
Y = f(A, K, L)
f(.) here is the underlying production technology, which will differ by the type of business, and will help determine the efficient scale of operations. When the outside options for jobs such as wage labor are limited and/or women are choosing self-employment as a way to balance family time responsibilities with the need to earn income, many of the women operating these businesses will be of rather limited entrepreneurial aptitude and have chosen businesses with a very small optimal efficient scale. Examples include many typical businesses run by women that serve a very small neighborhood market, such as some tailoring, cooking, and small-scale retail businesses.
The typical business training courses consist of a few days trying to teach business owners how to keep better records, plan ahead, cut costs, undertake more marketing, etc. For existing businesses, we should think of these as primarily affecting the A term – helping business owners try and get more out of their existing capital and labor inputs, but facing the same underlying production technology. But the problem then is if the underlying production technology f(.) is one with a very small optimal production scale, changing A (or even A and K) is underlying to have that much impact. For example, consider the following example from Poor Economics (p. 218) “Walking down the main street of the biggest slum in the town of Guntur at 9:00am, it is hard to miss the long line of women selling dosas, the rice-and-lentil pancakes…by our count, one particular morning, there was one dosa seller for every six houses”. In such a business, so undifferentiated from those around it, adding more capital to keep producing the same thing is unlikely to do much for sales. One could imagine that somehow training allows a firm to cut costs and take over more of this market, but if customers are mostly a set of regulars and competition for outside customers is fierce, keeping better records and planning ahead may again not change the fundamental nature of these businesses.
We need an f change
What’s needed for real growth are for these women to fundamentally change how and what they produce – that is, changing the f(.) function. However, this seems to be much harder in practice that making marginal improvements to an existing business. In many countries and contexts, other constraints are going to bind and limit the ability of women to make this change. As an extreme example, Gine and Mansuri talk about how the custom of purdah and other cultural restrictions in rural Pakistan limits many women to home-based manufacturing. In countries where female labor force participation is very low, it may be acceptable to run a subsistence business, but not to compete with male business owners in other sectors. Or a lack of child or elderly care facilities may restrict many women from operating businesses that take them away from home. The solution to these problems are likely to lie outside of training and finance interventions.
Nevertheless, there may be more hope with interventions for new firm owners. It may be much easier to change the choice of sector before it is made than to get women to switch from one type of production to another. But even here the evidence is pretty limited – our experiment does finds some positive impact, as do Field et al. (2010) who find that upper caste Hindu women who took their training were 19 percentage points more likely to report income, whereas the training had no effect on lower caste Hindu women or on Muslim women. They attribute the lack of impact on these other groups to social restrictions, arguing that training helped women whose businesses had been held down by social restrictions, but that women who faced more extreme restrictions could not respond to training. But in both cases, the businesses being created are still very much of the subsistence sort.
I have heard many times of efforts to combine business training with vocational or technical training in order to try and change the sectors women go into (e.g. The Lady Mechanic Initiative). But I’ve yet to see any rigorous evidence that any of these programs work and have heard of problems with customer acceptance in some cases. This is an area where we definitely need more work, but anyone know of any good evidence about these types of programs?