Published on Development Impact

Targeting which informal firms might formalize and bringing them into the tax system

This page in:
I have worked for a while with different attempts to get informal firms to register their businesses and become formal. We have tried giving them information and actually paying them to formalize, lowering the cost of registering to zero, offering them accountants and increasing enforcement. At the end of all of this, in strong contrast to the De Soto view that the informal sector is full of firms that would love to become formal if only burdensome regulations didn’t stop them, these studies have found most informal firms stay informal unless you strongly enforce them or pay them to formalize.

This is true for the average, but perhaps better targeting might help. I teamed up with Najy Benhassine, Victor Pouliquen and Massi Santini to help evaluate a new program the government of Benin was doing to make it easier for firms to formalize. In a recent working paper and impact note we provide the results of this evaluation. I thought I’d briefly summarize the key findings, but focus on a point of more broader interest in impact evaluations – even when we expect average impacts might be low, can we ex ante target or pre-specify groups for whom the intervention may work a lot better than average?

A quick summary of the intervention and experiment
Benin (along with 16 other OHADA countries in West and Central Africa), revised their commercial law to introduce a new status called the entreprenant. This status was designed for micro and small businesses, and registering with this new status is easy, free of charge and takes only one business day. Formalizing means firms gain access to many of the key benefits of formal status, but also makes the firm registered for tax purposes.

We used a listing survey of businesses around Cotonou in early 2014 to obtain a sample of 3,596 informal businesses. We then tested several interventions of increasing assistance to see if enhancing the supposed benefits of formalization would draw more firms into formality. The sample was therefore randomized into four groups:
  1. A control group of 1,197 firms
  2. Treatment group 1: 301 firms received an in-person visit explaining the benefits of formalizing, and help with registering if needed.
  3. Treatment group 2: 899 received the in-person visit, and also facilitated access to government training programs, and support to open a business bank account designed for the entreprenant.
  4. Treatment group 3: 1,199 firms in addition to the services provided to groups 1 and 2, were also offered support in dealing with the tax authorities including tax mediation services.
We then later added a fourth treatment, which was to provide information alone.

Headline result
Information alone has almost no effect on formalization. Adding the supplementary services did lead to large increases in take-up of the new status. The most comprehensive package of services (Group 3) boosted the formalization rate by 16.3 percentage points (Figure 1).

Figure 1: Some success at formalizing firms, but most stay informal
This rate of tax formalization is high relative to what other studies have found, and was likely helped by the fact that firms received a tax holiday to begin with. But even the most comprehensive treatment left more than 80 percent of firms remaining informal. Moreover, since these additional services were costly, the most cost-effective method was treatment 1, which cost $1,237 per additional firm formalized, compared to $2,217 for treatment 2 and $1,678 for treatment 3.

Would targeting help?
While most informal firms might want to stay informal, perhaps we can identify in advance a subset that are more likely to want to formalize and then target programs at them. We pre-specified several ways of doing this. The approach I had highest hopes for was a species classification approach ( de Mel et al, 2010, Bruhn, 2013). This involved collecting data in our listing survey for already formal firms, as well as the informals, and then running a model that used location, gender, age, education, sector, firm age, business practices, sales, etc. to predict formal status. Using this approach, we classified ex-ante 18 percent of the informal firms in our sample as “looking like the formal species”, with the idea that the interventions should work better for such a group.

Figure 2 shows the results. The formal species firms are more likely to formalize even in the control group, and the treatment has a significantly larger impact for them – we get 27% of the formal-like firms formalizing in treatment group 3.

Figure 2: Targeting formal-like firms can have more success
This suggests some promise for this targeting approach. However, it required a lot of data on the firms to be able to identify which firms are the formal species. We find that a few much cruder targeting tools also work as well in separating responsiveness to treatment: women are less likely to formalize than men, those without secondary education are less likely to formalize than those with secondary education, and firms operating outside the main Dantokpa market more likely to formalize than those within the market. Using any of these criteria for targeting can drive down the cost per firm formalized to $500-600. However, this is still expensive relative to business profits, and governments may have concerns about targeting a program only to men or the more educated.

This work also points to the importance of testing different targeting strategies experimentally – other targeting criteria that we thought might explain responsiveness to treatment were initial business size, and how frequently they were visited by tax inspectors, yet there was not significant heterogeneity according to these criteria.

Targeting, program effectiveness, and external validity
I think this issue of how to best target programs is going to continue to be a fruitful area for research, especially in the firm literature. There are a couple of related issues I want to note here:
  • Targeting on treatment impacts is harder to do than targeting on levels: in my example, the “formal-like species” was more likely to formalize anyway, but additionally also had larger treatment effects. You could then imagine that the interventions might have had more success with the less formal-like firms, who needed that extra push. This wasn’t the case, but it is often hard to know if our interventions will be substitutes for or complements to our targeting criteria. Similar issues appear when trying to target firms which you think will grow fastest, etc. See also this old post on targeting high growth firms.
  • Possible trade-offs between maximizing program effects and external validity: suppose I got my targeting even better, and could identify a small subset of the population with really large treatment effects. If I just did my experiment/program on that subgroup, I could show really large impacts, but this wouldn’t tell me anything about how well the program works for the majority of firms. But conversely, if I don’t target at all, I might find an average impact that is small, even though the intervention could be of great benefit to some firms. Bruce Wydick has a nice post on this when thinking about microfinance impacts.


David McKenzie

Lead Economist, Development Research Group, World Bank

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000