Credit expansion and the economy

Abstract

Credit expansion has been associated with faster economic growth and with a higher occurrence of financial crises – a pair of results which seem to contradict each other. This paper advances an explanation for these results by separating credit to the private sector into credit to firms and credit to households. The empirical analysis shows that credit to firms is responsible for the positive growth effect, while the higher occurrence of crises is mainly due to credit to households. The events of the last decade, where fast credit expansion led to crises and very little growth, can be understood as a shift in the composition of credit towards its household component.

Keywords:

• credit
• growth
• financial crises

JEL Classification:

• E44
• G21

Notes

¹ Important papers in this literature include King and Levine (1993), Beck et al. (2000), Levine et al. (2000), Benhabib and Spiegel (2000), Rioja and Valev (2004), Aghion et al. (2005) and Badunenko and Romero-Ávila (2013).

² For recent evidence, with an emphasis on the 2008 global financial crisis, see Schularick and Taylor (2012) and Jorda et al. (2011). For earlier evidence, with an emphasis on crises in developing countries, see Kaminsky and Reinhart (1999), Demirguc-Kunt and Detragiache (1998, 2002) and Domac and Peria (2003).

³ The main exception to this is student loans, which are used to accumulate human capital and could therefore be expected to have a positive effect on growth. Student loans are relatively unimportant in most countries other than the United States – and even there they accounted for just 8% of total household credit in 2011.

⁴ See Dembiermont et al. (2013) for a description of the data.

⁵ See Table A1 in the Appendix for more details on time coverage by country.

⁶ Since the country coverage of our data increases over time, plotting the average over all available observations would introduce bias due to composition effects. Figure 1 gets around this by considering a constant set of countries with complete time series over three selected time periods.

⁷ I control for the initial level of GDP per capita, average years of schooling in the adult population, government consumption over GDP, exports plus imports over GDP and the inflation rate. With the exception of GDP per capita and average years of schooling, all variables are averaged over the 5 years of each growth period. All regressors are used in log form.

⁸ Table 1 also reports the Arellano–Bond test for serial correlation of order two in the error term in differences for columns 1 and 3 (the test is not possible in columns 2 and 4 as these use only two time periods). In both instances, the test does not reject the null of no serial correlation. I do not report the Hansen test of overidentifying restrictions as the large number of instruments renders it very weak.