Impact Evaluation of Business Incubators

In developing countries, micro, small and medium-sized enterprises (MSMEs) provide nearly 90% of all employment. MSMEs therefore represent a key part of the solution to unemployment and poverty. However, these enterprises exhibit low levels of survival and productivity. Business incubators are deployed as a vehicle for improving enterprise survival and growth. As a policy option, incubators focus on enterprises that already have growth potential, and aim to support their survival, growth and productivity.[1] But do these incubators achieve the desired impact? What can be done to intensify their impact?

Rigorous analyses on these two questions require relevant data on incubated and non-incubated firms which is mostly unavailable in low-income countries (LICs). Thus, this project seeks to enable the impact evaluation of business incubators by creating a rich dataset from administrative and survey data. This will be done in Nigeria, an archetypal developing country in sub-Saharan Africa.[2] The data from the project will be widely relevant to research and policy discussions on how to support the emergence and evolution of high-growth enterprises in LICs. Particularly, an impact evaluation of incubators is anticipated as a follow-up on the data generation exercise.

Quantitative impact evaluations of incubators are hard to implement for two methodological reasons: one, it is difficult to define a universally acceptable set of performance criteria to assess because incubators vary in goals and expected outcomes (Akçomak, 2011) and two, constructing a valid control group is challenging (Sherman and Chappell, 1988). These problems are deepened in LICs by the lack of appropriate data. By generating a rich dataset that will include information on relevant performance indicators, and that will enable valid counterfactual analyses, this project has the potential to address all three concerns. Presently, the first stage focuses on data generation; the counterfactual analyses are left for a follow-up stage for which funding is currently being sought.

Since its inception in the mid-90s, the incubation programme in Nigeria now includes over 36 incubators, 25 of which were already in operation by 2010. All of these are managed by the NBTI. The NBTI maintains an analog record of all enterprises that apply to any incubator. The amount of information available in this record varies by year and incubator. In some incubator-years, only the names of applicant firms are recorded while in others, basic information such as number of employees and main products are also recorded. Selected firms spend an average of 5 years in incubation.

The first step in the project, which has been completed, is to digitize the administrative record to obtain a roster of all enterprises that ever applied, irrespective of whether they were admitted or not. Based on the roster, a census of firms in the 25 oldest and largest incubators will take place. Admittedly, this seems to be a large number of incubators to start with. However, the Nigerian incubators are generally small - the oldest and largest one has only 30 incubation spaces. Thus, starting with a total of 25 incubators will ensure that the minimum sample size required for a sufficiently powered impact evaluation is realised. The census, which will primarily rely on an enumerator-administered questionnaire, will serve three purposes: one, it will help to establish the survival status of the firms; two, it will allow to update the roster with any missing data on firm-level attributes at baseline[3]; and three, it will allow to collect current data on the relevant firm-level attributes, to general a pseudo-endline.

For the purpose of this project, the outcomes of interest are standardized measures of labour productivity (revenue, sales and profits per employee) and growth (change in number of employees, revenue, sales, profits and growth). These variables are the most relevant to the long-term desired outcome of incubators (that is, enterprise productivity and growth) and they are easier to recollect even when missing from administrative data. Data on these and other relevant firm-level attributes will therefore be gathered. The key immediate deliverables will be available soon.

In a follow-up, the dataset from stage 1 will elicit the causal impact of incubators on firm growth and productivity, by comparing a set of incubates with carefully selected controls and incorporating a randomized control trial (RCT) of an incubator services innovation.[4] Funding for this stage is currently being sought, and partners are actively welcome.

The current stage of project was funded by the Foreign and Commonwealth Development Office under its PEDL initiative, a joint initiative with the Centre for Economic Policy Research.


[1] Some incubated firms in Nigeria have grown to become highly productive. Spectra Industries Limited (graduated from the oldest incubator in 1998) is a good case in point. To date, success claims about the incubator programme use such examples. An important outcome of this project will be to inform a shift away from these ‘outliers’ to a systematic overview of incubator impact.

[2] Nigeria is an illustrative case, being the largest economy in sub-Saharan Africa and having an extensive incubation programme.

[3] Reconstructing baseline by recall is a recommended strategy when data was not collected at the start of a programme (Bamberger, 2014). Consider a firm i that applied to the earliest incubator in 1993. The roster may have only the name of this enterprise and no more. In the survey, I will be able to tell whether the firm is surviving. If so, I can ask the owner or current manager about the average revenue and number of employees, for instance, in 1993. I can also ask about the revenue and number of employees at the time of survey. Thus, I will have the baseline and pseudo-endline data.

[4] Methodological details of the proposed impact evaluation and RCT, as well as the associated sample and power calculations, will be provided later. Suffice it to mention at this stage that, based on preliminary calculations, a total sample of 128 firms suffices to detect a medium effect size in a two-tail hypothesis test with statistical significance of 0.05, statistical power of 0.80 and 10% of variance explained by covariates. Thus, a target sample of 300 firms is adequate, even at 50% response rate.