Power Outages, Firm Productivity, and Generator Ownership in East Africa

ABSTRACT

Unreliable electricity supply is a widespread problem in developing countries. This paper examines the impact of power outages on firm productivity in East Africa, considering the effect of generator ownership in general and during blackouts. The results show large negative effects of power outages on productivity and a mitigating effect of self-generating during power outages of approximately the same size. However, self-generation is also found to generally reduce productivity due to the idle capital effect of unused generators during blackout-free times. This finding highlights that self-generation is only profitable for firms if the experienced power outages are severe enough.

KEYWORDS:

• Blackout
• self-generation
• total factor productivity
• labor productivity
• Africa

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 In some sectors, firms may not experience material spoilage from power outages, and it may be plausible to assume fully-flexible materials like in Allcott et al. (2016). However, this sample contains firms from the chemicals sector which often experience such material spoilages (see, Allcott et al., 2016). Furthermore, choosing material to be semi-flexible increases the understandability of the model. Changing the model to fully-flexible materials is, however, straightforward by differentiating between $M_{itS}^{{\alpha }_m}$ and $M_{itG}^{{\alpha }_m}$ and following the same computations.

2 The characteristics of power outages do not differ substantially from the uncleaned sample.

3 32 observations are lost when estimating TFP since they report zero in one of the production inputs.

4 The drought possibly negatively affected productivity in certain industries, e.g. the food industry. The industry fixed effects also capture these effects.

5 Excluding the control variables changes the significance of the self-generating dummy while the impacts of power outages and the interaction term remain almost unchanged. However, including the interaction terms of the self-generating dummy and each firm characteristic increases the significance and magnitude of the self-generating dummy in all estimations (see Table A1 in Appendix A).

6 The analysis was also conducted including the interaction term ${\mathrm{S}}_{\mathrm{i}\mathrm{j}\mathrm{c}}\times \mathrm{S}\mathrm{e}{\mathrm{c}}_{\mathrm{j}}\times Ge{n}_{ijc}$ with relatively similar results.

7 The positive, statistically significant coefficients in column (4) in Table 6 result in an approximate net-zero effect for other manufacturing firms due to the negative main effect of power outages. The same holds for Tanzanian firms in column (2) in Table 7 .

8 Increasing the threshold to an average outage duration between 5.04 and 7.04 hours still refers to 46 firms that falsely self-generate and 37 firms that falsely do not self-generate. Out of these 37 firms, 36 are SMEs.