Financial development and TFP growth: cross-country and industry-level evidence

Abstract

This article estimates the impact of financial development on industry-level Total Factor Productivity (TFP) growth using a largely unexploited panel of 77 countries with data for 26 manufacturing industries for the years 1965 to 2003. A significant relationship is found between financial development and industry-level TFP growth when controlling for country-time and industry-time fixed effects. The results are both statistically and economically significant. TFP growth can accelerate up to 0.6% per year, depending on the external finance requirement of industries, following a one SD increase in financial development. The results are robust to different samples and specifications.

Keywords:

• financial development
• TFP growth

JEL Classification::

• D24
• E44
• O47

Acknowledgements

This article represents the views of the authors and do not necessarily reflect the views of any institution including the IDB and the IMF, their Executive Directors or the countries they represent. We thank Alejandro Izquierdo, Carmen Pagés and Fabio Schiantarelli as well as other participants at the IDB workshops on productivity, for very useful comments and discussions on previous versions, and Oscar Becerra and Carlos Rodriguez for their comments and research assistance. All remaining errors are our own.

Notes

¹ Most of the studies showing strong links between financial development and growth are based on cross-sectional growth regressions (see, for instance, King and Levine, 1993a, b; Levine, 1997; Levine and Zervos, 1998), others on pooled time series-cross sectional country level data (Beck et al., 2000; Levine et al., 2000). At the macro-level depth of access is negatively correlated with poverty rates (Levine, 1997; Honohan, 2004).

² As discussed by Levine (2005), the channels through which finance operates include higher savings rates, greater investment, technological innovations and productivity gains.

³ See Levine (1997) and Bencivenga et al. (1995) for a general discussion.

⁴ Greenwald et al. (1990), Aghion et al. (2005), Buera and Shin (2008) and Buera et al. (2008) are examples of this literature.

⁵ See, e.g. Fisman and Love (2005) ; Hartmann et al. (2007).

⁶ See, e.g. Greenwald et al. (1990), Aghion et al. (2005), Buera and Shin (2008) and Buera et al. (2008).

⁷ Along similar lines, Behr and Lee (2005) show that when mechanisms to transfer credit risk are in place, productivity enhancement investments are stimulated and the mechanism strengthened.

⁸ According to Rioja and Valev (2003) less-developed countries adopt a capital accumulation strategy in which there is less innovation and productivity growth. In developed countries, there is a stronger incentive for TFP improvement via innovation and technological developments, since firms compete across countries with similar capital stocks. Financial markets in these economies fund these innovations, leading to larger productivity gains.

⁹ See, e.g. Wurgler (2000), Fisman and Love (2003, 2005), Hartmann et al. (2007), Papaioannou (2007), Ciccione and Papaioannou (2006) and Bekaert et al. (2007). These are showing how deeper credit markets allow sectors to exploit technological innovations better. Similarly, Beck et al. (2008) employ a cross-country, cross-industry approach to explore the effect of financial sector efficiency on firm entry.

¹⁰ We transform the series to constant prices using US Consumer Price Index base 2000 taken from the International Monetary Fund's (IMF) International Financial Statistics (IFS). The list of countries, industries and time periods as well as descriptive statistics of the data used in the study are reported in the Appendix.

¹¹ Ideally, we would wish to instrument K and L such that we could overcome the endogeneity problem; due to data limitations, however, we do not follow this course of action.

¹² We apply a double weight procedure. Within a country, we weight every sector by its relative size in the national economy. And between countries, we weight by the country's relative size (in terms of value added) in the regional aggregate. The figures with the unweighted mean and median levels show similar results.

¹³ In our sample, the Latin American region corresponds to Bolivia, Chile, Colombia, Ecuador, Guatemala, Mexico, Panama, Peru, Uruguay and Venezuela.

¹⁴ In our sample. Emerging Asia includes Indonesia, Korea, Malaysia, Philippines and Singapore.

¹⁵ Following Cerra and Saxena (2008), in our sample industrial countries refers to Australia, Austria, Belgium, Canada, Denmark, Finland, France, Greece, Ireland, Italy, Japan, Luxemburg, Malta, the Netherlands, New Zealand, Norway, Portugal, South Africa, Spain, Sweden, Turkey, the UK and the US.

¹⁶ Earlier studies suggested that the East Asian growth came mainly from a capital surge, but Hsieh (2002) shows that productivity growth was an important factor behind the surge.

¹⁷ We also run the regressions using other proxies for financial development used elsewhere in the literature such as the ratio of liquid liabilities to GDP, or the stock market capitalization of listed companies as a ratio of GDP. While the results we obtain are consistent with the baseline, our preferred indicator is the ratio of private credit to GDP because it is the most direct measure of financial intermediation to the private sector.

¹⁸ Although, it is worth pointing out that Rajan and Zingales (1998) is purely cross-sectional, thus they do not exploit the time dimension as we do here.

¹⁹ The inclusion of country-time fixed effects in Equation 11 accounts for all possible observable and unobservable determinants of TFP growth that vary at the country-time dimensions. In Equation 12, we forgo the fixed effect in order to be able to identify . To the extent that all the other determinants of industry-level TFP growth that vary across country and time are either not observable, or may be missing from , then Equation 12 is possibly misspecified due to the omitted variable bias.

²⁰ See, e.g. Barro (1997).

²¹ See, e.g. Cerra and Saxena (2008) and Rancière et al. (2008).

²² We compute inflation volatility as the absolute value of the coefficient of variation in inflation for each country over a 5-year period. We also try other measures of macroeconomic volatility, such as real exchange rate volatility, terms of trade volatility and the volatility of capital flows. The results are qualitatively similar. We opt to report the results based on the inflation measure because it is the one with the largest data availability for our sample of countries.

²³ In our sample, , but it is sufficiently small in absolute value such that the RHS of Equation 16 is still an increasing function of .

²⁴ The results are qualitatively the same for any other levels of RZ_{i .}

²⁵ While the positive slope suggests that financial development is relatively more important in a volatile environment, the effect is imprecisely estimated.

²⁶ The emphasis in added in the word ‘may’ as we cannot rule out the possibility that the widening of the SEs at higher levels of volatility in Fig. 4 is a purely statistical phenomenon.

²⁷ For a discussion on the relatively high incidence of macroeconomic and financial volatility in Latin America and the Caribbean, see Hausmann and Gavin (1996), De Ferranti and Perry (2000) and IDB (2005).