Contestable Credit Markets and Household Welfare: Panel Data Evidence from Ethiopia

Abstract

This paper explores the impact of credit constraints on household welfare in Ethiopia. We use a three-wave panel dataset for rural and small-town households to estimate the effects of household borrowing constraints on two alternative indicators of household welfare: consumption expenditure and asset ownership. The presence of a constraint is treated as an endogenous regressor, using an instrumental variable based on Baumol’s theory of contestable markets. We find that credit constraints have a significantly negative effect on both outcomes. These results are robust to several alternative specifications of the model.

Disclosure statement

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

Supplementary material

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2020.1826447.

Notes

1. A few studies have presented measures of the level of contestability in formal sector financial markets in industrialised and transitional economies: see for example De Bandt and Davis (2000) for the Eurozone, Heffernan (2002) for the UK, Molyneux, Thornton, and Lloyd-Williams (1996) for Japan, Nathan and Neave (1989) for Canada, and Yildirim and Philippatos (2007) for Eastern Europe. There is also evidence from other countries that the distance between formal-sector lenders and low-income borrowers is positively associated with moral hazard problems and therefore the interest rate that these lenders charge (Ergungor, 2010; Pedrosa & Do, 2011; Presbitero & Rabellotti, 2014), but our paper explores a mechanism that is completely different to this.

2. However, not all cross-sectional studies report a uniformly positive association. For example, Liverpool and Winter-Nelson (2010) find a positive association only for wealthier households in Ethiopia, and Ali and Deininger (2014) find a great deal of heterogeneity in the strength of the association across different areas in Ethiopia.

3. Li et al. (2013) use Chinese panel data. In a related study, Li, Gan, and Hu (2011) apply a difference-in-difference estimator to a Chinese panel dataset.

4. The ESS is a collaborative project between the Central Statistical Agency of Ethiopia and the World Bank’s Living Standards Measurement Study. The first wave of the ESS in 2012 covered rural areas and small towns (with fewer than 10,000 inhabitants), while the second wave in 2014 and the third wave in 2016 also covered large towns and cities. Data were collected using a two-stage sampling strategy. At the first stage, enumeration areas (EAs) were randomly selected, the probability of selection being proportional to population size. At the second stage, individual households were chosen at random from each selected EA. The data can be accessed at http://microdata.worldbank.org/index.php/catalog/2783/get_microdata.

5. A Stata do-file to produce the results in this table is included in the Supplementary Material; the data file is available on request.

6. The zone is an official administrative division of Ethiopia; there are 68 zones in total.

7. The 2016 round of the survey also included 1,254 households from large towns and cities, but these observations do not constitute a panel, so they cannot be included in our fixed-effects estimates.

8. An alternative is the ‘indirect’ approach, which is based on a lifecycle/permanent income hypothesis model (Diagne, Zeller, & Sharma, 2000). See Ibrahim, Kedir, and Sebastián (2007) for a discussion of this approach.

9. Recent papers using similar outcome variables include Akotey and Adjasi (2016) and Li et al. (2016).

10. This variable is based on the following question: ‘Did you or members of your household receive any of the following types of assistance in the past 12 months from the government or a non-governmental institution (such as a church)?’ The types of assistance are as follows: Productive Safety Net Programme payment; free food; Food-for-Work/Cash-for-Work/Inputs-for-Work Programme payment; other.

11. In order to increase the efficiency of the Mundlak estimates, the regression equations also include a set of indicator variables for the zone in which the community is located.

12. Stata do-files to produce the results in Tables 5–6 are included in the Supplementary Material; the data file is available on request.