Technical Change in the Peruvian Regulated Microfinance Sector

Abstract

This study evaluates and quantifies the technical change in the Peruvian regulated microfinance sector during the period 2003–2015 estimating a translog cost function. We found technical deterioration between 2005 and 2015 due to the differentiated effects that the technical innovations, implemented by the MFIs, had on their production costs. In the analysis by groups of MFIs, all groups converged at a technical deterioration rate between 8.6 and 9.13% in the last year. Problems in the implementation, execution, and/or management of the new technologies as well as the difficulty in achieving reductions in the variable costs may explain these results.

RESUMEN

Este estudio evalúa y cuantifica los cambios tecnológicos ocurridos en el sector de microfinanzas peruano durante el período 2003–2015 y estima una función de costo translog. Notamos un deterioro técnico entre los años 2005-2015 causado por los efectos diferenciados, provocados en los costos de producción por las innovaciones tecnológicas implementadas por las instituciones de microfinanzas (IMFs). Los análisis realizados en grupos de IMFs mostraron que, el año pasado, todos los grupos convergieron a una tasa de deterioro técnico entre 8,6% a 9,13%. Las causales de estos resultados pueden residir en los problemas enfrentados durante la implementación, ejecución y gestión de estas nuevas técnicas y las dificultades en reducir los costos variables.

RESUMO

O presente trabalho avalia e quantifica as mudanças técnicas no regulado setor microfinanceiro peruano, no período de 2003 a 2015, estimando a função de custo translog. Observou-se deterioração técnica entre 2005 e 2015 devido aos efeitos diferenciados que as inovações técnicas, implementadas pelas IMFs (MFIs na sigla em inglês), provocaram nos custos de produção. Nas análises por grupos de IMFs, todos os grupos convergiram para um índice de deterioração técnica entre 8,6% e 9,13% no ano passado. Problemas na implementação, execução e/ou gestão de novas tecnologias, assim como a dificuldade em reduzir custos variáveis podem explicar esses resultados.

Keywords:

• Microfinance
• technical change
• translog cost function

Notes

Acknowledgments

We would like to thank Jorge Rojas and three anonymous referees for comments and suggestions.

Notes

1 There is a non-regulated and unsupervised microfinance sector made up of cooperatives and non-governmental organizations (NGOs) with microcredit programs. There is no official record for these entities that provides their exact number or total loan portfolio. However, some of them (likely, the most important entities) voluntarily report to the MIX Market database and, according to this information, in 2015, they were responsible only for 7% of the total microcredits offered by the entire Peruvian microfinance sector. Therefore, given the impossibility of knowing official information about the unregulated sector and its limited participation in the total supply of microcredits, our analysis considers only the regulated microfinance sector.

2 For eight consecutive years, between 2008 and 2015, the Global Microscope, prepared by the Economist Intelligence Unit (EIU) (2016), chose Peru as the best environment for the development of microfinance in the world.

3 See, for example, Hunter & Timme (1991); Humphrey (1993); Webster (1997); Stiroh (1997), Maudos, Pastor, & Quesada (1997); Altunbas, Goddard, & Molyneux (1999); Carbo et al. (2003).

4 The correlation coefficient between total loans and the number of borrowers is 0.85. In addition, Table A1 shows the average number of clients per group of MFIs.

5 According to information published by the SBS. www.sbs.gob.pe

6 Apoyo & Asociados (2018). Informe anual sobre Financiera Crediscotia S.A.

7 Caja Municipal de Ahorro y Crédito Ica (CMAC Ica). Memoria Anual 2011.

8 The minimum efficient scale is the production level that minimizes the average long-term cost of production.

9 We employ the World Bank methodology that preserves the growth rates exhibited in the constant local price series. See https://datahelpdesk.worldbank.org/knowledgebase/articles/114943-what-is-your-constant-u-s-dollar-methodology

10 Table A2 in the Appendix shows the number of MFIs throughout the period analyzed.

11 Table A3 in the Appendix shows the descriptive statistics by groups of MFIs.

12 We employ xtserial command in Stata to perform Wooldridge’s test for serial autocorrelation. See Drukker (2003) for further details about this command.

13 We employ xtcsd command in Stata to perform Pesaran’s cross-sectional dependence test. See De Hoyos and Sarafidis (2006) for further details about this command.

14 We employ xtscc command in Stata to perform the fixed effects model with Driscoll and Kraay standard errors. This command allows the specification of the lag to be considered in the autocorrelation structure and provides robust standard errors to very general forms of cross-sectional and temporal dependence. Their small sample properties are significantly better than those of the alternative covariance estimators when cross-sectional and temporal dependence are present. See Hoechle (2007) for further details about this command.