Explore the link between Financial Structures and Economic Growth

Economics Letters 124 (2014) 382–385

Contents lists available at ScienceDirect

Economics Letters

journal homepage: www.elsevier.com/locate/ecolet

The finance and growth nexus revisited✩

Roland Beck, Georgios Georgiadis ∗, Roland Straub European Central Bank, Germany

h i g h l i g h t s

• Credit expansion has a positive output growth effect only up to a point. • Beyond the threshold the impact of finance on growth vanishes. • The non-linearity may stem from the omission of factors not considered so far. • The omitted factors may have negative growth effects in mature financial sectors. • Such factors include financial cycles and banks’ non-intermediation activities.

a r t i c l e i n f o

Article history: Received 3 January 2014 Received in revised form 17 June 2014 Accepted 22 June 2014 Available online 1 July 2014

JEL classification: G1 O4

Keywords: Growth Financial development Dynamic panel data

a b s t r a c t

We find that an expansion of credit has a positive effect on per capita output growth only up to a point. Beyond this threshold the impact of finance on growth is not statistically significant anymore. We show, however, that the estimated non-linear relationship may stem from the omission of factors not considered in the literature so far. These factors may have a negative impact on growth in mature financial systems, and include the magnitude of financial cycles as well as the importance of non-intermediation activities in banks’ business models.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Since the seminal contribution of King and Levine (1993) a large body of empirical literature has shown that financial development exhibits a positive impact on growth.1 Several recent cross-country studies have re-examined this relationship finding that finance indeed fosters growth, but only up to a point beyond which this positive effect vanishes. However, the possible sources of this non- linearity have not been addressed in this literature yet.2

✩ We thank Kalin Nikolov, Alexander Popov and an anonymous referee for useful comments. The opinions expressed in this paper are those of the authors and do not necessarily reflect those of the ECB or the Eurosystem. ∗ Correspondence to: Kaiserstr. 29, 60311 Frankfurt, Germany. Tel.: +49 69 1344

5851. E-mail addresses: jorgo@georgiadis.de, georgios.georgiadis@ecb.int

(G. Georgiadis). 1 See, Rajan and Zingales (1998), Levine and Zervos (1998), Beck et al. (2000), and

Wurgler (2000). For a survey of the literature see Levine (2005). 2 See Favara (2003), Rioja and Valev (2004), Panizza et al. (2012), and Cecchetti

and Kharroubi (2012). Manganelli and Popov (2013) study the underlying channels using industry-level data.

http://dx.doi.org/10.1016/j.econlet.2014.06.024 0165-1765/© 2014 Elsevier B.V. All rights reserved.

Theory and recent experience suggest a number of possible mechanisms through which a non-linear effect of finance on growth may arise. First, intermediaries may build up excessive leverage in mature financial sectors, possibly accentuated by the ample availability of capital market funding (Rajan, 2006). In such an environment, an economy might experience more financial sector-induced fluctuations with protracted recoveries after crises and balance sheet recessions that dampen growth systematically (Rong et al., 2010). Second, large financial sectors may progres- sively rely on business models that are based on proprietary trad- ing and other non-interest income generating activities, which might have a smaller effect on economic growth than the tradi- tional intermediation of savings to productive investment (Turner, 2010; Beck et al., 2013). Third, mature financial systems may de- velop complex financial instruments whose opaqueness allows informed agents to extract rents from investors, leading to an over- allocation of human capital to the financial sector; as a result, re- search and development in the non-financial sector is reduced and growth slows down (Tobin, 1984; Philippon, 2010). Fourth, the progressive share of loans extended to households instead of firms in more mature financial sectors may fail to foster growth, as the

R. Beck et al. / Economics Letters 124 (2014) 382–385 383

Table 1 Descriptive statistics.

Variable # obs Mean Std. dev. Min Max

Private credit rel. to GDP 575 0.47 0.43 0.00 2.23 Stock market cap. rel. to GDP 393 0.44 0.54 0.00 4.81 Bank credit to deposits 573 1.00 0.58 0.10 5.77 Financial sector assets rel. to GDP 575 0.59 0.52 0.00 3.16 Financial reform index 362 0.66 0.25 0 1 Banking crisis dummy 575 0.09 0.20 0 1 Value added share 401 0.05 0.03 0.01 0.27 Researchers per million people 204 6.69 1.58 2.05 8.96 Household credit share 130 0.37 0.14 0.02 0.62

financing is used for consumption rather than investment (Beck et al., 2012). Fifth, while reducing distortionary state intervention generally fosters the development of financial markets, recent ex- periences in many countries suggest that excessive deregulation may increase the frequency of boom and bust cycles (Rajan, 2006).

In this paper, we complement the existing empirical literature by examining whether the failure to control for such features of countries’ financial systems in cross-country panel regressions may be responsible for the finding of a non-linear effect of finance on growth. Our results suggest that the non-linear relationship may partly be explained by the omission of financial variables reflecting the magnitude of financial cycles and the importance of non-intermediation, but not so much by the increase in household relative to corporate lending and the extent of financial sector deregulation.

2. Empirical strategy

We estimate dynamic panel regressions of output on a number of standard growth determinants including financial system characteristics. The model we estimate is given by

yit = αyi,t−1 + βxit + π1wit + π2w 2 it + γ1zit + µi + λt + uit , (1)

where the dependent variable yit is the logarithm of real GDP per capita in PPP-adjusted US dollars, and the explanatory variables xit include the rate of inflation, years of secondary schooling, government consumption and trade (imports plus exports) relative to GDP.3 Our explanatory variable of interest wit is the ratio of private credit by deposit money banks relative to GDP, which is entered in linear as well as in squared terms in order to allow for a non-linear effect of finance on growth. We also include time and country-fixed effects in order to account for common factors and unobservable, time-invariant, country-specific effects on output. Finally, we include additional financial system characteristics zit as controls in order to address possible omitted variable bias that may be driving the finding of a non-linear effect of finance on growth.4

Specifically, we consider the inclusion of (i) stock market capitalisation, bank credit relative to deposits and a banking crisis dummy in order to examine whether the existence of more pronounced financial cycles in mature financial systems may account for the non-linear effects of finance on growth; (ii) the financial sector’s share in total value added and financial sector total assets in order to test whether financial firms increasingly engaging in non-intermediation activities reduces their growth-promoting potential; (iii) the number of researchers per million people to address that the existence of rents in more complex financial systems might drain human capital from other sectors in the economy and thereby hamper productivity growth; (iv) an index of financial reform to capture whether excessive

3 See Barro (1998) for a discussion of these standard growth regressions. 4 A description of the data sources can be found in the Appendix.

deregulation and risk-taking may be responsible for the non- linearity in the finance–growth nexus; (v) the share of household credit in total credit to investigate whether the declining growth effects stem from finance becoming less a facilitator for the accumulation of productive physical capital.

Our sample features data on up to 132 countries and spans the time period from 1980 to 2005. As is standard in the empir- ical growth literature, we purge business cycle frequencies from the data by using five-year averages. Table 1 provides summary statistics of the financial system characteristics we use in the re- gressions. Due to the resulting small time-series dimension and the presence of the lagged dependent variable in Eq. (1), it is well known that the dynamic fixed effects estimator is biased. There- fore, we resort to system GMM estimation (see Blundell and Bond (1998)).5 The GMM approach also allows us to account for the pos- sibility that the explanatory variables might not be exogenous or predetermined. Importantly, we pay careful attention to limit the instrument count in order to avoid problems stemming from ex- cessive instrument proliferation, such as over-fitting, weakened tests for over-identifying restrictions, biased two-step variance es- timators and imprecise estimates of the optimal weighting matrix (see Roodman, 2009a,b).

3. Results

Table 2 presents our results for the estimation of the coefficients π1 and π2 from Eq. (1) for our baseline model as well as a number of alternative specifications with one additional explanatory variable included at a time. Table 2 also reports the associated coefficient estimates for γ1. In addition to coefficient estimates, Table 2 also reports the number of instruments that were used, the p-values for the tests of second-order residual serial correlation, and the Hansen test for the validity of over-identifying restrictions. Finally, Table 2 also reports the values of the implied thresholds beyond which a further expansion of credit does not foster growth any longer. Notice that we also calculate the implied thresholds of the baseline model specification using the restricted sample of the extended model. Overall, the diagnostics suggest that the models are correctly specified. Moreover, the number of instruments is reasonably small, mitigating concerns that excessive instrument proliferation could compromise the reliability of the estimation results.

All coefficient estimates are consistent with a non-linear effect of finance on growth across specifications. In the baseline specification, we find that an expansion of credit beyond 109% of GDP does not foster growth anymore.6 While the estimates in the specifications with additional financial variables are not

5 The dependent variable, the logarithm of GDP per capita, is highly persistent so that the GMM estimator introduced by Arellano and Bond (1991) is likely to suffer from weak instrument problems. 6 Our threshold estimate is very similar to that in Panizza et al. (2012) of 100% of

GDP.

384 R. Beck et al. / Economics Letters 124 (2014) 382–385

Table 2 Estimation results.

Baseline Financial cycles Non-intermediation activities Misallocation of talent

Excessive deregulation

Household vs. corporate lending

Financial system characteristic measure

Stock market capitalisation relative to GDP

Bank credit to deposits

Banking crisis dummy

Share of value added by financial sector

Financial sector total assets relative to GDP

Researchers per million people

Index of financial reform

Share of household credit in total credit

Credit/GDP 0.74*** 1.10*** 0.64 0.60** 0.63** 1.44 0.57 1.14** 0.72 (Credit/GDP) 2 −0.34*** −0.44*** −0.34* −0.23** −0.35** −0.52** −0.27 −0.59*** −0.32 Additional financial variable

– 0.07 0.04 −0.06 2.77 −0.01 −0.16 −0.83 −0.37

# of observations 575 393 569 575 401 575 204 362 130 # of countries 132 101 131 132 117 132 83 79 34 # of instruments 15 17 16 17 17 15 12 16 15 Hansen p-value 0.47 0.98 0.85 0.32 0.37 0.40 0.43 0.93 0.12 AR(2) p-value 0.21 0.28 0.32 0.13 0.65 0.20 – 0.04 0.59 Threshold 109% 124% 95% 130% 91% 138% 106% 96% 111% Threshold baseline model with restricted sample

– 127% 105% 109% 94% 109% 80% 103% 110%

The threshold is calculated as x∗ = −0.5γ1/γ2. The models are estimated with collapsed set of instruments in order to avoid issues resulting from excessive instrument proliferation as discussed by Roodman (2009a,b). We use the two-step variance matrix estimators and robust standard errors as suggested by Windmeijer (2005). We employ the xtabond2 routine in Stata 12 with the options collapse and pca (see Roodman, 2009a,b). In the regressions with bank credit relative to bank deposits we drop Vietnam which features implausibly high values of this financial system characteristics. As there are at most three consecutive time-series observations in the case of researchers per million people, the test for second-order residual serial correlation cannot be carried out.

* Indicates statistical significance at the 10% significance level. ** Indicates statistical significance at the 5% significance level. *** Indicates statistical significance at the 1% significance level.

always statistically significant, in all specifications do the point estimates suggest that there is a threshold beyond which credit does not foster growth anymore. However, several specifications feature considerably higher point estimates for the threshold. For example, when controlling for the incidence of banking crises the threshold for credit rises from 109% of GDP in the baseline model to 130% of GDP.7 Likewise, controlling for total financial sector assets – which include assets other than loans, such as trading receivables – relative to GDP lifts the threshold to 138% of GDP. Importantly, in both cases the increase in the threshold does not stem from differences in the samples. Finally, the threshold also rises considerably when controlling for the number of researchers in the population: while the threshold is 80% of GDP when the baseline specification is estimated on the restricted sample, it rises to 106% when the number of researchers is controlled for. However, the coefficient estimates for credit relative to GDP are not statistically significant in this case.

4. Conclusions

This paper extends the existing empirical literature on the non- linearity in the finance and growth nexus in cross-country panel regressions by controlling for structural features of countries’ fi- nancial systems which – if omitted – may give rise to the finding of such a non-linearity. We find that even after controlling for a num- ber of structural features finance continues to display a positive effect on growth only up to a point; beyond this critical threshold, the positive marginal effect of finance on growth vanishes. How- ever, some specifications lead to considerably higher point esti- mates for this threshold, suggesting that controlling for certain fi- nancial variables – such as the frequency of financial cycles and

7 It should be stressed that the effects of banking crises might be rather persistent so that they are not purged from the data by five-year averages (see Rong et al., 2010).

the importance of non-intermediation activities in banks’ business models – tends to ‘‘push out’’ the point beyond which finance no longer fosters growth. Our results thus do not call for an indiscrim- inate downsizing of financial sectors. They do suggest, however, that policies aimed at limiting excessive leverage and risk-taking as well as requiring banks to refocus their business models towards the provision of credit could ensure that financial deepening has positive growth effects even in mature financial systems. The non- linearity in the finance and growth nexus which remains after con- trolling for the financial variables discussed in this paper may stem from other factors not considered here. For example, finance may foster growth by supporting investment that enables economies to catch up to the global technological frontier. Once economies have reached that point, they may no longer benefit from further finan- cial deepening (Aghion et al., 2005).

Appendix

Data on per capita output, inflation, government consumption, trade and researchers per million people are taken from the World Bank’s World Development Indicators. Data on financial variables are taken from the World Bank’s Global Financial Development data base (Čihák et al., 2012). Educational attainment data stem from Barro and Lee (2010). Data on banking crises originate from Laeven and Valencia (2012). The value added share data are taken from the European Union’s KLEMS database, which provides data for the 27 EU countries plus the US, Australia and Japan from 1970 to 2007 (but for shorter time periods for some countries) and from the United Nations, which cover more than 100 countries from 1970 to 2011 (but for most countries only for a shorter time period); we merge the data from these two sources to obtain a broad country coverage. The data on financial reform are taken from Abiad et al. (2008). Data on the share of credit allocated to households is taken from the BIS.

R. Beck et al. / Economics Letters 124 (2014) 382–385 385

References

Abiad, A., Detragiache, E., Tressel, T., 2008. A new database of financial reforms, IMF Working Paper 08/266.

Aghion, P., Howitt, P., Mayer-Foulkes, D., 2005. The effect of financial development on convergence. Q. J. Econ. 120 (10), 173–222.

Arellano, M., Bond, S., 1991. Some tests of specification for panel data: monte carlo evidence and an application to employment equations. Rev. Econom. Stud. 58, 277–297.

Barro, R., 1998. Determinants of Economic Growth: A Cross-Country Empirical Study, first ed.. In: MIT Press Books, vol. 1. The MIT Press.

Barro, R., Lee, J.-W., 2010. A new data set of educational attainment in the world, 1950–2010, NBER Working Paper 15902.

Beck, T., Büyükkarabacak, B., Rioja, F., Valev, N., 2012. Who gets the credit? and does it matter? household vs. firm lending across countries. B.E. J. Macroecon.

Beck, T., Degryse, H., Kneer, C., Is more finance better? disentangling intermediation and size effects of financial systems, European Banking Center Discussion Paper 2012-016.

Beck, T., Levine, R., Loayza, N., 2000. Finance and the sources of growth. J. Financ. Econ. 58 (1–2), 261–300.

Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. J. Econometrics 87 (1), 115–143.

Cecchetti, S., Kharroubi, E., 2012. Reassessing the impact of finance on growth, BIS Working Paper 381.

Čihák, M., Demirgüç-Kunt, A., Feyen, E., Levine, R., 2012. Benchmarking financial systems around the world, Policy Research Working Paper 6175.

Favara, G., 2003. A reassessment of the relationship between finance and growth, IMF Working Paper 03/123.

King, R., Levine, R., 1993. Finance and growth: schumpeter might be right. Q. J. Econ. 108 (3), 717–737.

Laeven, L., Valencia, F., 2012. Systemic banking crises database: an update, IMF Working Paper 12/163.

Levine, Ross, 2005. Finance and growth: theory and evidence. In: Aghion, P., Durlauf, S. (Eds.), Handbook of Economic Growth, Vol. 1, first ed. Elsevier, pp. 865–934. chapter 12.

Levine, R., Zervos, amd S., 1998. Stock markets, banks and economic growth. Am. Econ. Rev. 88 (3), 537–558.

Manganelli, S., Popov, A., 2013. Financial dependence, global growth opportunities, and growth revisited. Econom. Lett. 120 (1), 123–125.

Panizza, U., Arcand, J.-L., Berkes, E., 2012. Too much finance, IMF Working Paper 12/161.

Philippon, T., 2010. Engineers vs. financiers: should the financial sector be taxed or subsidized? Am. Econ. J.: Macroecon. 2, 158–182.

Rong, Q., Reinhart, C., Rogoff, K., 2010. On graduation from default, inflation and banking crisis: elusive or illusion? NBER Working Papers 16168, National Bureau of Economic Research, Inc., forthcoming in D. Acemoglu M. Woodford (Eds.), NBER Macroeconomics Annual 2010, vol. 25, Cambridge, Mass, MIT Press.

Rajan, R., 2006. Has finance made the world riskier? Eur. Financ. Manag. 12, 499–533.

Rajan, R., Zingales, L., 1998. Financial dependence and growth. Am. Econ. Rev. 88 (3), 559–586.

Rioja, F., Valev, N., 2004. Does one size fit all? a re-examination of the finance and growth relationship. J. Dev. Econ. 74, 429–447.

Roodman, D., 2009a. A note on the theme of too many instruments. Oxf. Bull. Econ. Stat. 71 (1), 135–158.

Roodman, D., 2009b. How to do xtabond2: an introduction to difference and system GMM in stata. Stata J. 9 (1), 86–136.

Tobin, J., 1984. On the efficiency of the financial system. Lloyds Bank Rev. 153, 1–15. Turner, A., 2010. What do banks do? Why do credit booms and busts occur and what

can public policy do about it?, in: The Future of Finance: The LSE Report. Windmeijer, F., 2005. A finite sample correction for the variance of linear efficient

two-step GMM estimators. J. Econometrics 126 (1), 25–51. Wurgler, J., 2000. Financial markets and the allocation of capital. J. Financ. Econ. 58

(1-2), 187–214.