Tax Innovations and Public Revenues in Sub-Saharan Africa

Abstract

We study the effect of two tax innovations – value added taxes (VAT) and autonomous revenue authorities (ARA) – on tax revenues in sub-Saharan Africa. The dataset consists of 47 countries over 1980–2010. We find that VATs have no effect on total tax revenues, neither in the short- nor in the long-run. ARAs lead to higher tax revenues in the short- and medium-run, but the effect dissipates over time. The main conclusion is that tax innovations are not a panacea to overcome the revenue shortages in African countries, but they are helpful in the short- and medium-run.

Acknowledgements

We are also grateful for comments by participants of the Twelfth Nordic Conference in Development Economics in Bergen, Norway. Funding by Sida-SAREC within the project Fiscal Capacity and Democracy in Developing Countries is gratefully acknowledged. We also thank an anonymous referee for helpful comments and suggestions. All data used in the paper is available upon request.

Notes

1. According to our data, many African countries had average tax revenues to GDP ratios of less than 10 per cent of GDP during the sample period. Countries like Chad, Nigeria, or Angola even had ratios of only 5–6 per cent.

2. Another recent study that explores the link between VATs and public revenues is Baskaran (2014). However, he neither focuses exclusively on sub-Saharan Africa nor allows for effect heterogeneity. Other related contributions are Keen and Lockwood (2006) who show that the VAT is associated with higher revenues in OECD countries, Ebeke and Erhart (2011) who find that VATs reduce tax revenue instability in developing countries, and Desai and Hines (2005) who show that the VAT is associated with lower exports.

3. Baskaran (2014) conducts a rudimentary quantitative assessment of how ARAs affected revenues. However, as for the VAT he neither focuses on sub-Saharan Africa only nor allow for heterogeneous effects. Fjeldstad and Moore (2009) provide a comprehensive qualitative description of the ARA model.

4. There is some ambiguity in the literature about the date of the French VAT introduction. The VAT in its original form was introduced in 1948 (Bird & Gendron, 2007); the modern variant was introduced in 1954.

5. A theoretical discussion of various design elements of a VAT is offered by Agha and Haughton (1996).

6. This specification is the regression analogue of the difference-in-difference approach.

7. To be more specific, country fixed effects cleanse out systematic differences between, say, Francophone and Anglophone countries. This is essential considering that francophone countries were early adopters of VATs and Anglophone countries where early adopters of ARAs.

8. Note that given the incidental parameters problem in probit models, we do not include country and year dummies in the first stage of either model.

9. The results from the first stage are collected in Table A5 in the Online Appendix. They show, in particular, that the exclusion restrictions perform reasonably: The number of contiguous countries with a VAT is positively related to the probability that a given country introduces a VAT but has no effect on the probability for an ARA, and vice versa. An alternative approach to account for self-selection would be to follow Ebeke and Erhart (2011) and use the exclusion restrictions as instruments in an IV approach. This is a less desirable approach in our setting because we are interested in effect heterogeneity, which would require additional instruments for the endogenous interaction variables.

10. Results are available upon request.

11. Another way to evaluate the economic significance of the effect is to compare standardised beta coefficients. This would give an impression of the relative importance of the tax innovations compared to other controls. Standardised beta regressions suggest that ARAs are in the short to medium run as effective in increasing revenues as IMF non-crisis programmes. More important covariates for tax revenues, however, are the share of agriculture in GDP and GDP per capita. These results are available upon request.