The impact of household credit and enterprise credit on economic growth : an empirical study of 24 OECD countries

(1)

1

The Impact of Household credit and

Enterprise credit on economic growth: an

empirical study of 24 OECD countries

Master thesis, academic year 2013-2014 Faculty of Economics and Business Universiteit van Amsterdam Jianqiao Zhao 10647589 Supervisor: prof. Lex Hoogduin Date: July 2014

(2)

2

Abstract

It is important to understand the relationship between credit compositions and economic development, since household credit and enterprise credit influence economic growth through different channels. This study uses empirical techniques to investigate whether household credit and enterprise credit increase economic growth. Previous researches predict a positive impact of enterprise credit on economic growth and an ambiguous effect of household credit on economic growth. My estimated results confirm that enterprise credit improves economic growth and household credit reduces economic growth in 24 OECD countries over the period 1995-2011.

(3)

3

Outline of content

1. Introduction--- 4

2. Literature review--- 6

3. Private sector credit among 24 OECD countries---7

3.1

Private sector credit definition and composition

--- 8

3.2

Historical development: total credit to GDP

--- 8

3.3

Bank credit versus total credit

--- 9

3.4

Development of household and Enterprise credit

--- 9

4. Data description--- 10

4.1

Credit data and calculation

--- 11

4.2

Other macroeconomic variables

--- 12

5. Empirical methodology---13

6. Results of estimation---15

6.1 Estimated

results in 24 OECD countries

---15

6.2

Estimated results in 7 emerging OECD countries

---20

6.3

Estimated results in 17 advanced OECD countries

---23

6.4

Impacts of credits on economic growth: emerging economies versus

advanced economies---25

7. Conclusion

---27

8. References

---29

(4)

4

1. Introduction

This paper attempts to empirically examine whether enterprise credit and household credit improve economic growth. Corporate finance theories highlight that lending to enterprises is able to improve the productivity efficiency and hence economic growth. However, the impact of household credit on economic growth is controversial. Two possible channels reveal the effect of household credit on economic growth. According to Jappelli & Pagano’s (1994) overlapping-generations model, relaxing liquidity constraints on households decrease the saving rate, since the excess sensitivity of household consumption to business cycle variations is lower. The reduction of saving rate1 results in a decline in economic growth. Thus, economic development is negatively affected by household credit. Another channel is the positive impact of household credit on human capital accumulation. While households invest in education, economic growth is improved. In order to specifically investigate the impact of household credit and enterprise credit on economic development, I construct an annual panel data comprising 24 OECD countries over the period 1995-2011.

The structure of this paper is as follows. Section 2 presents related literature. Both theoretical and empirical researches indicate a positive impact of enterprise credit on economic growth. The expansion of enterprise credit improves productivity efficiency and hence economic development. The effect of household credit on economic growth is ambiguous. However, some empirical studies suggest a negative sign of household credit. Section 3 provides a detailed analysis of private sector credit among 24 OECD countries over the period 1995-2011. From the perspective of the demand side, private sector credit can be decomposed into two components: non-financial corporations (enterprises) and households. The growth of credit market has risen dramatically during the past 18 years. The importance of household credit

(5)

5

in credit market development has been increasing among both advanced and emerging OECD economies. Nevertheless, enterprise credit still accounts for a predominant share of total credit in 24 OECD countries. Section 4 describes data sources and the calculation method. Credit data are collected form the Bank for International Settlements (BIS) and most macroeconomic variables data is collected form the World Bank database (WDI). In section 5, empirical methodologies are presented. I use standard OLS and IV regressions to estimate the independent impact of household credit and enterprise credit on annual GDP per capita growth. Two exogenous variables are employed to be instruments of enterprise credit to GDP and household credit to GDP. Legal origin is used to explain the variation of enterprise credit and religious composition is used to explain the variation of household credit. In addition, some explanatory variables are used to control for specific effects on economic growth. The estimated results are discussed in section 6. Except for the aggregate analysis, 24 OECD countries are classified into two groups: 7 emerging and 17 advanced economies. The specific impacts in different economies are analyzed separately. Although aggregates results confirm the positive impact of enterprise credit on economic development and negative effect of household credit on economic growth, emerging and advanced OECD economies face different situations. Section 7 concludes main findings of my empirical studies. A discussion about reverse causality between economic growth and credit compositions is also presented in the final section.

(6)

6

2. Literature review

Most theoretical and empirical studies focus on the impact of enterprise credit on economic growth. According to Levine’s endogenous growth model (1993), financial systems affect entrepreneurial activities that improve productivity. His research states that financial systems impact savings and investment decision and hence economic growth through five channels. (i) Mobilizing and pooling savings. (ii) Allocating recourses and providing ex-ante information of investment. (iii) Facilitating the exchange of goods and services. (iv) Monitoring managers and exerting corporate governance. (v) Facilitating trading, hedging and risk management. Financial institutions are only willing to finance the most productive enterprise. Thus, enterprise finance is able to stimulate economic growth by productivity improvement. Aghion, Howitt & Mayer-Foulkes (2005) use a Schumpeterian growth model with financial constraints to examine the relationship between enterprise credit development and steady-state GDP. Their model shows that countries with developed financial systems will converge to the economic growth rate of the world technology frontier. Except for theoretical models, many empirical studies confirm that enterprise finance has a positive impact on economic growth. Based on the industry-level data, Rajan & Zingales (1998) state that financial development facilitates economic growth by reducing costs of external finance to enterprises. King & Levine (1993) provide cross-country evidence that financial systems can increase economic growth. They emphasize that economic growth is strongly associated with physical capital accumulation and efficiency improvement. This implies that the enterprise credit market has a positive influence on economic development. Moreover, Beck, Demirguc-kunt & Levine (2005) support that credits to SMEs (small and medium sized enterprise) strongly accelerate GDP per capita growth.

(7)

7

However, household credit as an important part of private sector credit has ambiguous effects on economic growth. Jappelli & Pagano (1994) use an overlapping-generations model to demonstrate that the increase in household credit reduces the saving rate and therefore negatively affects economic growth. On the contrary, Galor & Zeira’s (1993) study on income distribution and macroeconomics indicates that individuals’ investment in human capital encourages economic growth. In addition, Gregorio’s (1996) empirical analysis in OECD and developing countries also shows the positive relationship between human capital accumulation and economic growth. He suggests that relaxing the borrowing constraint on individuals will increase educational investment that provides a positive growth effect.

Beck, Buyukkarabacak, Rioja & Valev (2008) decompose bank lending into household credit and enterprise credit. Their empirical result confirms the positive impact of enterprise credit on economic growth whereas no evidence supports that household credit can raise GDP per capita growth. Similarly, an empirical analysis by Sassi & Gasmi (2012) among all European countries shows a positive impact of enterprise credit on economic growth and an inverse effect of household credit on economic growth.

3. Private sector credit among 24 OECD countries

In this section, the specific analysis of private sector credit is presented. It includes the definition and composition of private sector credit, the development of the credit to GDP ratio, bank credit vs. total credit, and the development of household credit and enterprise credit. I briefly summarize the main findings of private sector credit in 24 OECD countries over the period 1995-2011. Firstly, credit has outgrown GDP in most OECD countries. Secondly, the share of bank credit decreases with credit development. Thirdly, the growth of household credit is more rapid than the growth

(8)

8

of enterprise credit. Finally, the importance of household credit in credit market among 24 OECD countries is increasing.

3.1 Private sector Credit definition and composition

To distinguish from Beck, Buyukkarabacak, Rioja & Valev’s (2008) credit definition2, the credit sources for the analysis in this paper include all lending sectors such as commercial banks, saving banks or credit unions. In term of financial instruments, outstanding claims of loans and debt securities are both included. From the perspective of the credit demand, I decompose total credit into two components: non-financial corporation credit and household credit. Non-financial corporation credit or enterprise credit contains credits to both private and public non-financial corporations. Household credit contains credits to household and non-profit institutions serving households.

3.2 Historical development: total credit to GDP

Over the period 1995 to 2011, credit has outgrown GDP in most OECD countries. As shown in figure A1 and A2 (see appendix), credit to GDP ratios in most advanced OECD economies are more than 150% of GDP, whereas these ratios in emerging OECD economies are relatively low ,less than 100% of GDP. The highest credit to GDP ratio occurs in Demark 2009 (approximately 280% of GDP) and the lowest credit to GDP ratio occurs in Mexico 2001 (approximately 18% of GDP). The average credit growth among 24 OECD countries was approximately 3% per year. In some OECD countries, like Demark, Norway, Sweden and Belgium, total credits grew rapidly during the past 18 years. Most advanced economies experience substantial credit expansions with credit to GDP ratios reaching levels close to or above 200% of GDP around the global financial crisis. Compared to advanced economies, total credit to GDP ratios in emerging countries such as Mexico, Poland and Turkey grow slowly or remain at a low level.

2_Beck,_{Buyukkarabacak, Rioja & Valev (2008) use credit to the private sector by deposit money bank} to analyze the impact of household and enterprise credit on economic growth. Besides, he defines business credit as the sum of loans to industry, construction, service agriculture and trade.

(9)

9

3.3 Bank credit versus total credit

On average, 60% of total credits are provided by domestic banks. Nevertheless, this ratio varies across OECD countries and over time. For instance, bank credit in Greece accounts for almost 86% of total credit. In the United States, only 30% of total credits are provided by domestic banks. With the development of the financial system, domestic bank credit becomes less important. A simple regression indicates a negative correlation between the share of bank credit and total credit in figure A4 (see appendix). This result confirms that the share of bank credit decreases with credit development. Interpretations of this trend are the development of the shadow banking system and deregulation of financial systems, which encourage the diversification of funding sources. Bank loans are no longer the largest source of the credit market.

However, in other countries, bank credit is still the most important source. This phenomenon is observed in both emerging and advanced OECD economies. For example, average bank credit to total credit ratios in Germany and Denmark are 75% and 73% respectively. Regarding emerging countries like Greece and Korea, this ratio amounts to 85% and 76% respectively. In general, bank credit in emerging OECD economies is a more significant source than in advanced OECD economies. Moreover, bank credit in bank-based financial systems is the majority of funding sources whereas bank credit share in market-based financial systems is a relatively small amount.

3.4 The Development of Household credit and Enterprise credit

Over the past 18 years, the volume of both household credit and enterprise credit in most OECD countries increases dramatically. The growth of household credit is more rapid than the growth of enterprise credit, which is particular the case for emerging OECD countries. By early 1995, household credit in 7 emerging OECD countries only occupies 10%-30% of total credit. Nevertheless, 2011 data shows that the share of household credit rises to 40%-60%. Additionally, in most advanced economies, the

(10)

10

share of household credit exceeds the share of enterprise credit. For example, by early 1995, the share of household credit is already more than 50% of total credit in USA. Although household credit becomes more important, enterprises or corporations are still major credit takers. On average, the share of enterprise credit is 62% and the share of household credit is only 38% in 24 OECD countries during the period 1995 to 2011.

Table 1：This table shows the correlation between the logarithm of total credit to GDP, the logarithm of Enterprise credit to GDP, the logarithm of Household credit to GDP, Enterprise credit share and Household credit share.

Correlation Totoal credit to GDP Enterprise credit to GDP Household credit to GDP Enterprise credit share Household credit share Totoal credit to GDP Enterprise credit to GDP _0.9081 Household credit to GDP _0.8724 _0.5874

Enterprise credit share _-0.3722 _-0.0108 _-0.7066

Household credit share _0.3722 _0.0108 _0.7066 _-1.0000

Source: (own calculation)

As illustrated in table 1 above, an important finding is that the household credit share is positively correlated with total credit to GDP. This indicates that the importance of household credit in credit markets among 24 OECD countries is increasing from 1995 to 2011.

3. Data description

This section presents data sources and calculation method. I construct an annual balanced panel dataset comprising 24 OECD countries over the period 1995-2011. 17 of them are developed countries and only 7 of them are developing countries. Selected countries are classified below in table 2.

Table 2: This table divides 24 OECD countries into two groups (developed and developing countries).

(11)

11

Austria, Australia, Belgium, Canada, Denmark, France, Finland, Germany, Italy, Japan, Netherlands, Norway, Portugal, Spain, Sweden, United Kingdom, United States

Czech Republic, Greece3, Hungary, Korea republic, Mexico, Poland, Turkey

Source: IMF country classification 4.1 Credit data and calculation

The credit data are provided by the Bank for International Settlements (BIS) from 1995 to 2011. The original credit data use quarterly information with national currency. In order to standardize the formation, I select credits at the end of the year as the annual credits, including household credit, non-financial corporation credit, total credit and domestic bank credit. Additionally, GDP data from OECD statics are used to calculate the credit to GDP ratio. The method is the following:

𝐓𝐨𝐭𝐚𝐥 𝐜𝐫𝐞𝐝𝐢𝐭 𝐭𝐨 𝐆𝐃𝐏 =𝐂𝐫𝐞𝐝𝐢𝐭 𝐟𝐫𝐨𝐦 𝐚𝐥𝐥 𝐬𝐞𝐜𝐭𝐨𝐫𝐬 ,𝐧𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐜𝐮𝐫𝐫𝐞𝐧𝐜𝐲,𝐜𝐮𝐫𝐫𝐞𝐧𝐭 𝐩𝐫𝐢𝐜𝐞

𝐆𝐃𝐏 ,𝐧𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐜𝐮𝐫𝐫𝐞𝐧𝐜𝐲,𝐜𝐮𝐫𝐫𝐞𝐧𝐭 𝐩𝐫𝐢𝐜𝐞 (1)

Then, I use the distribution of household credit and enterprise credit to obtain the household credit to GDP ratio and enterprise credit to GDP ratio respectively. Table 3 below summarizes key variables of household credit to GDP, enterprise credit to GDP, total credit to GDP, domestic bank credit to GDP, household credit share and enterprise credit share.

Table 3: description and definition of credit variables

Variable Definition Observations Mean Std.Dev Min Max

_{Total credit to GDP (%)} 0.5940373

Domestic bank credit to GDP (%) 0.3794625

HCG Household credit to GDP (%) 0.3074515

ECG Enterprise credit to GDP (%) 0.3588097

HCR Household credit share (%) 0.1254634 0.3868588 0.9494427

ECR _{Enterprise credit share (%)} 0.377457 0.1254634 0.0505573 0.6131412

3

Greece had been deemed a developed market since May 2001, however it recently returned to an emerging market.

(12)

12

4.2 Other macroeconomic variables

Most macroeconomic variables are extracted from the World Bank database (WDI), including annual GDP per capita growth, trade openness to GDP, initial GDP per capita, the annual inflation rate and government spending to GDP. Recession indicator as a dummy variable captures the business cycle effect. It was extracted from The National bureau of Economic Research (NBER) and Organization of Economic Development (OECD) Composite leading indicators: Reference Turning Points and Component series data. Legal origin and religious composition as dummy instrumental variables are extracted from La Porta al. (1999) database. The legal origin includes UK legal system, Germany legal system, French legal system and Scandinavian legal system. The religious composition includes Catholic, Protestant and Muslim. Specifically, previous empirical studies4 suggest that legal origin explains the variation of enterprise credit and religious composition explains the variation of household credit.

Table 4: description and definition of macroeconomic variables

Variable Definition Observations Mean Std.Dev Min Max

_{Annual GDP per capita growth (%)} 2.680291

Trade openness to GDP (%) 0.3367007

Gsg Government spending to GDP (%) 0.0403379

Inf Inflation (%) 0.1038763

BCI _{Recession indicator} 0.4345679 0.4963133 0 1

Inig _{Initial GDP per capita (million)} 408 23999.11 11847.56 5197.597 52213.96

Table 4 above summarizes the main macroeconomic variables. In general, there are 408 observations for most macroeconomic variables. Nevertheless, the recession indicator (BCI) only has 405 observations due to lack of data in Poland from 1995 to

4_{Beck & Levine’s (2005) suggests British common law countries have better financial systems than} other legal origin countries. Stulz & Willianmson (2003) suggest that Protestant population have stronger credit rights.

(13)

13

1997. The average annual GDP per capita growth rate among 24 OECD counties is 1.9% over the period 1995-2011. According to the definition of World Bank indicators, General government final consumption expenditure denotes government spending. Trade openness denotes the combination of export and import.

5. Empirical methodology

This section presents empirical methodologies that are employed to estimate the impact of household credit and enterprise credit on economic growth. The general panel growth regression is organized in the following form:

it = β1ln H it+ β2ln E it + δ ontrolsit+ εit

Where g is the GDP per capita growth rate, ln H it is the logarithm of household

credit to GDP, ln E it is the logarithm of enterprise credit to GDP, Controls are the

explanatory variables that might influence economic growth and εit is the error

term of the regression function. Control variables include dummy country variables, the logarithm of government spending to GDP, the logarithm of initial GDP per capita, the logarithm of trade openness to GDP, the inflation rate and recession indicator. The logarithm term is employed to measure the non-linear relationship. Moreover, the legal origin and religious composition are adopted to be instrumental variables of household credit to GDP and enterprise credit to GDP.

Many empirical and theoretical works indicate that enterprise credit has a positive impact on economic growth，whereas the effect of household credit on economic growth is ambiguous. To empirically examine the independent effect of credits to households and enterprises on economic growth among 24 OECD countries, I use different estimation techniques. Firstly, a standard OLS regression is employed. The random-effect estimation is more compatible than fixed-effect estimation, since time invariant variables can be included. Secondly, two instrumental variables are used to avoid endogenous problems. The legal origin and religious composition are employed as instruments of household credit and enterprise credit to avoid endogenity bias and omitted variables. According to Beck & Levine’s (2005) empirical research, British

(14)

14

common law countries have better financial systems than other legal origin countries. On the other hand, Stulz & Willianmson (2003) suggest that Protestant population have stronger credit rights. Thus, the instrument of legal origin is employed to explain the variation of enterprise credit and the instrument of religious composition is employed to explain the variation of household credit. Finally, 24 OECD countries are classified into two groups: emerging economies and advanced economies. The aggregate analysis is not able to reflect specific impacts of household credit and enterprise credit on economic growth in emerging and advanced OECD economies. Thus, the impact of household credit and enterprise credit on economic growth in emerging and advanced OECD economies are analyzed separately. In addition, although this paper mainly focuses on the impact of household credit and enterprise credit on economic growth, effects of total credit and domestic bank credit on economic growth are also examined.

In order to estimate the independent effect of private sector credit on economic growth, I control for some variables that might influence economic growth. The dummy recession indicator captures the business cycle effect on economic growth. The recession period is one and booming period is zero. Thus, this indicator is expected to show a negative sign. The initial GDP per capita reflects the convergence speed in different countries. According to conditional convergence theory, poor economies typically grow faster than richer economies. Therefore, emerging counties have high economic growth rates and advanced economies have low economic growth rates. The sign of initial GDP per capita is expected to be negative. The dummy country variables are used to control for the country specific effect, which allows me to extract accurate coefficients of explanatory variables. Barro (1995) suggests that the inflation rate is expected to have an inverse effect on economic growth in the long-run. The trade openness to GDP contains both export to GDP and import to GDP. This explanatory variable measures the openness of economies. Theoretical and empirical studies supports that open economies grow faster than closed economies. Thus, the coefficient of this variable is expected to be positive.

(15)

15

The government spending to GDP is used to measure government size. The relation between government expenditure and economic growth is not certainly confirmed. Whereas government spending in human capital and infrastructures are likely to improve economic growth, inefficient public consumption may reduce economic growth.

6. Results of estimation

This section discusses results of the empirical studies. There are four sub-sections. Firstly, the effects of household credit, enterprise credit, total credit and domestic bank credit on economic growth in 24 OECD countries are presented. Secondly, I only focus on impacts of different credits on economic growth in seven emerging OECD countries. Thirdly, the same analysis in 17 advanced OECD economics is discussed. Finally, the comparison of estimated results between emerging OECD economies and advanced OECD economies is analyzed. In general, my estimated results confirm the positive impact of enterprise credit on economic growth. Nevertheless, total credit, domestic bank credit and household credit reduce economic growth. Although household credit shows a negative sign in emerging OECD counties, the coefficient is not significant. Lending to non-financial corporations in advanced OECD economies may be not optimal for economic development, since the IV regression indicates a zero impact of enterprise credit on annual GDP per capita growth.

6.1 Estimated results in 24 OECD countries

Correlations in table B5 (see appendix) indicate that household credit to GDP, total credit to GDP and domestic bank credit to GDP are negatively correlated with annual GDP per capita growth rate. It is unexpected that enterprise credit also has a negative correlation with economic growth. To specifically investigate impacts of different credits on economic growth, the standard OLS regressions and IV regression are illustrated in table 6 below.

(16)

16

Table 6: Growth results of panel regression in 24 OECD economies over the period 1995-2011, random effect.

(Excluding dummy country variables)

Regression OLS (1) OLS (2) OLS (3) OLS (4) OLS (5) IV (6) Initial GDP per capita

Total credit to GDP

Domestic bank credit to GDP

Household credit to GDP

Enterprise credit to GDP

Business cycle indicator

Inflation rate Government spending to GDP Trade openness to GDP Constant Observations Adj.𝑹 Wald chi2(28) Prob> chi2 Sargan-Hansen test P-value 0.000*** (5.83) -0.020*** (-4.70) -0.013*** (-5.63) -0.0003* (-1.93) -0.082*** (-6.59) 0.055 *** (5.69) -0.157 *** (-5.59) 405 0.274 142.98 0 0.000*** (4.46) -0.010*** (-2.66) -0.014*** (-5.72) -0.0002 (-1.46) -0.076*** (-5.93) 0.044** (4.54) -0.136*** (-4.84) 405 0.246 123.40 0 0.000*** (6.21) -0.018*** (-5.87) - -0.014*** (-6.01) -0.0007*** (-3.83) -0.089*** (-7.15) -0.052*** (5.91) -0.182*** (-6.39) 405 0.296 159.09 0 0.000*** (4.25) -0.011*** (-2.72) -0.013*** (-5.52) -0.0002 (-1.24) -0.078*** (-6.11) 0.046*** (4.58) -0.147*** (-5.14) 405 0.247 123.78 0 0.000*** (5.25) -0.021 ** (-5.31) 0.006 (1.24) -0.014*** (-6.13) -0.0008*** (-4.02) -0.091*** (-7.26) 0.048*** (4.94) -0.182*** (-6.40) 405 0.300 160.84 0 0.000** (2.18) -0.026*** (-4.49) 0.039 *** (4.90) -0.016 *** (-6.61) -0.001*** (-4.22) -0.100 *** (-7.26) 0.023** (1.98) -0.179 *** (-5.42) 405 0.244 139.36 0 *significance at 10% level **significance at 5% level ***significance at 1% level

(17)

17

In column (1), the estimation measures the impact of total credit on economic growth. The result suggests that total credit to GDP has a significantly negative effect on economic growth at the 1% level. Column (2) reports the impact of domestic bank credit to GDP on economic growth. The result is similar with total credit to GDP, which is negatively and significantly associated with GDP per capita growth at the 1% level. The absolute value of the coefficient of domestic bank credit to GDP is lower than the absolute value of the coefficient of total credit to GDP. This implies that the negative effect of domestic bank credit on economic growth is smaller than the negative effect of total credit on economic growth. In Column (3-4), household credit to GDP and enterprise credit to GDP are estimated separately. The result of estimation suggests that household credit to GDP and enterprise credit to GDP enter negatively and significantly at the 1% level. The Column (5) regression indicates that household credit has a negative effect on economic growth and enterprise credit has a positive impact on economic growth. The coefficient of household credit is significant at the 5% level. Enterprise credit enters insignificantly.

Except for examining impacts of different credits on economic growth, estimations of control variables are also presented in table 6. The recession indicator successfully captures the business cycle movement. As discussed in the section on empirical methodology, the recession indicator as a dummy variable negatively affects economic growth. Column (1-6) reports that the recession indicator has a negative and significant sign at the 1% level. According to conditional convergence theory, poor economies typically grow faster than richer economies. Therefore, the coefficient of initial GDP per capita is expected to be negative. However, my estimated result shows a zero impact on economic growth. The effect of the inflation rate on annual GDP per capita growth is negative and significant, which is in line with my expectation. The positive and significant impact of trade openness on GDP per capita growth confirms that open economies grow faster than closed economies. According to my estimations, government spending among 24 OECD countries significantly reduces economic growth at the 1% level. This may imply that

(18)

18

government expenditure is inefficiently allocated and hence economic growth is harmed.

Column (6) presents the estimated result of the Instrumental variables regression. Two exogenous variables of legal origin and religious composition are used as instrumental variables of household credit and enterprise credit to avoid omitted variables and endogenous bias. The test of over identifying restriction indicates that my regression equation is exactly identified. Therefore, it is impossible to use the Sargan-Hansen test to approve the validity of my instruments. However, Results in IV regression confirm that the impact of enterprise credit on economic growth is positive and significant at the 1% level. Oppositely, the coefficient of household credit is negative and significant at the 1% level. In an instrumental variables regression, the coefficient of enterprise credit to GDP increases to 0.039 percentage point. This positive impact on GDP per capita growth approves the theoretical prediction that credit to non-financial corporations improves economic growth. The negative effect of household credit on economic growth may imply that the rapid growth of household credit is not optimal for economic development.

As we see in the IV regression (6), one percentage point increase in household credit to GDP ratio reduces GDP per capita growth by 0.026 percentage points. For instance, the growth of household credit to GDP in USA is 0.32 percentage points from 1995 to 2011. My results estimate that the growth of household credit to GDP in USA leads to 0.008 percentage points reduction in GDP per capita growth from 1995 to 2011. On the other hand, one percentage increase in enterprise credit to GDP ratio raises GDP per capita growth by 0.039 percentage points. For instance, the growth of enterprise credit to GDP in USA is 0.38 percentage points from 1995 to 2011. My results estimate that the growth of enterprise credit to GDP in USA raises GDP per capita growth by 0.015 percentage points from 1995 to 2011. Those results above approve that the impact of household credit and enterprise credit on economic growth is not only statistically, but also economically significant. Lending to

(19)

19

productive enterprises encourages productivity improvement. With the development of technologies and innovations, economic growth is accelerated. Regarding to the impact of household credit on economic growth, my estimated results revel a negative growth effect. Therefore, the prediction of households’ investment in education is not a reasonable explanation. Alleviating credit constraints on households will reduce the savings rate and hence economic growth. The negative effect of household credit on economic growth shows an inefficient use of household credit.

To summarize this sub-section, the positive impact of enterprise credit on economic growth can be confirmed in 24 OECD countries over the period 1995-2011. My estimated results in column (1-4) reveal that total credit, domestic bank credit, enterprise credit and household credit significantly reduce economic growth. However, the result in IV regression (6) confirms the prediction that enterprise credit significantly and positively affects economic growth. The impact of household credit on economic growth depends on how households decide to use credits. My estimated results show a negative effect on economic development. This may imply that credit is inefficiently spent by households. Most control variables are in line with my expectations. The only exception is initial GDP per capita. According to conditional convergence theory, the increase in initial GDP per capita reduces economic growth. My estimations suggest that the coefficient of initial GDP per capita is zero among 24 OECD countries over the period 1995-2011.

(20)

20

6.2 Estimated results in 7 OECD Emerging countries

Table 8: Growth results of panel regression in 7 emerging OECD economies over the period 1995-2011, random

effect. (Excluding dummy country variables)

Inflation rate Government spending to GDP Trade openness to GDP Constant Observations Adj.𝑹 Wald chi2(28) Prob> chi2 Sargan-Hansen test P-value -0.0001 (-1.50) -0.036*** (-3.27) -0.020*** (-3.31) -0.0003 (-1.43) -0.042 (-1.40) 0.037 ( 1.51) 0.599 (1.50) 116 0.302 45.07 0.0000 -0.0001 (-1.44) -0.017*** (-2.70) -0.022*** (-3.63) -0.0002 (-1.09) 0.040 (1.30) 0.016 (0.71) 0.587 (1.44) 116 0.268 38.12 0.0001 -0.0001 (-1.12) -0.019*** (-2.93) - -0.023*** (-3.80) -0.0005** (-2.13) 0.041 (1.35) 0.033 (1.32) 0.449 (1.09) 116 0.281 40.63 0.0000 -0.0001* (-1.72) 0.028** (-2.42) -0.020*** (-3.24) -0.00001 (-0.80) 0.046 (1.49 ) 0.017 (0.73) 0.690* (1.70) 116 0.272 38.78 0.0001 -0.0001 (-1.29) -0.012 (-1.55) -0.015 (-1.02) -0.021*** (-3.39) -0.0004 (-1.58) -0.042 (-1.35) 0.033 (1.30) 0.523 (1.25) 116 0.288 41.70 0.0000 -0.0001 (-0.57) -0.033 (-1.63) 0.034** (3.22) -0.024*** (-3.99) -0.001* (-1.90) -0.037 (-1.50) 0.051 (1.07) -0.007 (-0.11) 116 0.217 31.70 0.000 0.517 *significance at 10% level **significance at 5% level ***significance at 1% level

(21)

21

Table B7 (see appendix) shows that total credit to GDP, domestic bank credit to GDP and household credit to GDP are negatively correlated with GDP per capita growth in 7 emerging OECD countries over the period 1995-2011. Enterprise credit to GDP has a positive correlation with GDP per capita growth. This may imply that the positive impact of enterprise credit on economic growth could be observed. To specifically investigate impacts of different credits on economic growth, the standard OLS regressions and IV regression are illustrated in table 8 above.

Results in column (7-8) report that total credit to GDP and domestic bank credit to GDP are negatively and significantly associated with GDP per capita growth in 7 emerging OECD economies from 1995 to 2011. Column (9) reports the significant and negative effect of household credit on GDP per capita growth at the 1 % level. The result in column (10) reveals that enterprise credit to GDP significantly improves economic growth at the 5% level. The OLS regression in column (11) combines the impact of enterprise credit to GDP and household credit to GDP on economic growth. The estimated result shows that both household credit and enterprise credit insignificantly reduce GDP per capita growth.

Compared to estimated results in 24 OECD countries, most control variables in 7 emerging OECD countries are relatively insignificant. The inverse effect of initial GDP per capita on economic growth in column (7-12) confirms the prediction of conditional convergence theory. Nevertheless, the coefficient of initial GDP per capita is only significantly at the 5% level in column (10). The inflation rate reduces economic growth. The coefficient is only significant at the 5% level in column (9) and at the 10% level in column (12). Government spending to GDP in column (8-10) indicates an insignificantly positive growth impact. Although trade openness in 7 emerging economics is positively associated with economic growth, the t-test rejects the significance of the relevant coefficient. The recession indicator significantly captures the business cycle movement at 1 the % level.

(22)

22

The instrumental variables in 7 emerging OECD economies exclude the UK legal system, since it does not appear in any selected emerging OECD countries. The application of UK legal system leads to the collinearity of regression. The Sargan-Hansen test ensures the validity of instruments and rejects the over-identifying restriction. The estimated result in column (12) approves a positive and significant impact of enterprise credit on economic growth at the 5% level. On the contrary, household credit enters negatively and insignificantly. Compared with estimated results in 24 OECD countries, the absolute value of the coefficient of enterprise credit is lower. One percentage point increase in enterprise credit to GDP ratio raises annual GDP per capita growth by 0.034 percentage points. For instance, the growth of enterprise credit to GDP in Turkey is 1.24 percentage points from 1995 to 2011. My results estimate that the growth of enterprise credit to GDP in Turkey increases GDP per capita growth by 0.042 percentage points from 1995 to 2011.

In this sub-section, effects of different credits on economic growth in 7 emerging OECD counties are similar to estimated results in 24 OECD counties. Total credit, domestic bank credit and household credit are negatively associated with economic growth. The IV regression indicates that the coefficient of household credit is not significant whereas enterprise credit significantly accelerates annual GDP per capita growth at the 5% level. Most control variables are in line with my predictions. Government spending and trade openness in 7 OECD emerging economies might improve economic growth. The initial GDP per capita and inflation rate reduce economic growth. However, OLS and IV regressions cannot ensure the significance of those control variables. The only exception is the recession indicator. All regressions in column (7-12) report a significant and negative sign of the coefficient of business cycle indicator.

(23)

23

6.3 Estimated results in 17 advanced OECD countries

Table 10: Growth results of panel regression in 17 advanced OECD economies over the period 1995-2011,

random effect. (Excluding dummy country variables)

Inflation rate Government spending to GDP Trade openness to GDP Constant Observations Adj.𝑹 Wald chi2(28) Prob> chi2 Sargan-Hansen test P-value -0.0001*** (-3.73) -0.051*** (-8.38) -0.014*** (-7.12) 0.002** (2.46) -0.113*** (-5.98) 0.027*** ( 3.04) 0.319*** (3.34) 289 0.522 292.04 0.0000 -0.0002*** (-6.48) -0.041*** (-7.14) -0.014*** (-6.88) 0.002* (1.90) -0.131*** (-7.00) 0.021** (2.33) 0.522*** (5.62) 289 0.493 260.14 0.0000 -0.0000* (-2.19) -0.038*** (-7.05) - -0.014*** (-7.18) 0.002** (2.02) -0.112*** (-5.55) 0.022** (2.42) 0.138 (1.23) 289 0.491 258.09 0.0000 -0.0002* (-6.24) -0.051*** (-8.32) -0.013*** (-6.61) 0.003** (2.47) -0.132*** (-7.35 ) 0.028*** (3.12) 0.493*** (5.44) 289 0.521 290.51 0.0000 -0.0001*** (-3.73) -0.012* (-1.57) -0.041*** (-4.36) -0.013*** (-6.80) 0.003** (2.41) -0.119*** (-6.09) 0.029*** (3.20) 0.366*** (3.02) 289 0.525 294.59 0.0000 -0.000*** (-4.68) -0.049*** (-6.41) 0.000 (0.23) -0.015*** (-7.16) 0.002* (2.04) -0.027*** (-4.04) 0.027*** (-4.53) 0.005 (1.03) 289 0.485 475.02 0.000 0.229 *significance at 10% level **significance at 5% level ***significance at 1% level

(24)

24

Correlations in table B9 (see appendix) demonstrate that total credit, domestic bank credit, enterprise credit and household credit are negatively associated with GDP per capita growth in 17 advanced OECD countries from 1995 to 2011. To specifically investigate impacts of different credits on economic growth, the standard OLS regressions and IV regression are illustrated in table 10.

Estimated results in column (13-14) confirm the significant and negative effect of total credit and domestic bank credit on economic growth among 17 advanced OECD countries from 1995 to 2011 at the 1% level. Estimated results in column (15-16) show that both household credit and enterprise credit enter negatively and significantly at the 1% level. The standard OLS regression in column (17) indicates that enterprise credit significantly reduces economic growth at the 1% level and household credit is negatively and significantly associated with GDP per capita growth at the 10% level. The absolute value of the coefficient of enterprise credit is higher than the absolute value of the coefficient of household credit.

In 17 advanced OECD countries, most controls variable are significant and in the line with my expectations. The initial GDP per capita, business cycle indicator and government spending significantly reduce economic growth at least 10% level. Trade openness is significantly and positively associated with economic growth at least 5% level. However, the inflation rate positively and significantly influences GDP per capita growth in advanced economies at least 10% level. The explanation is a reverse causal relationship between economic growth and inflation rate. In the long-run, the effect of inflation rate on economic growth is negative. However, economic development also influences the inflation rate. Therefore, the relationship between economic growth and inflation rate could be positive. For instance, the aggregate demand (AD) framework indicates a positive correlation between the inflation rate and economic growth. When economic development is improved, the inflation rate rises too.

(25)

25

The instruments in 17 advanced OECD countries are different from the previous two sub-sections. The exogenous variable of religious composition results in invalidity of instruments. The Sargan-Hansen test shows a high P-value and rejects the validity of instrumental variables, when the instrument of religious composition is included. Thus, the instrumental variable of household credit and enterprise credit in this sub-section is legal origin. The column (18) reports that household credit significantly reduces economic growth at the 1% level. One percentage point increase in household credit to GDP ratio reduces GDP per capita growth by 0.049 percentage points. For instance, the growth of household credit to GDP in Netherlands is 1.16 percentage points from 1995 to 2011. My results estimate that the growth of household credit to GDP in Netherland decreases GDP per capita growth by 0.057 percentage points from 1995 to 2011. The impact of enterprise credit on economic growth in advanced OECD economies is almost zero and insignificant.

To conclude results in 17 advanced OECD countries, the significant and negative effect of total credit, domestic bank credit and household credit on economic growth can be confirmed. Although the coefficient of enterprise credit is positive, t-test rejects the significance. I cannot draw a conclusion that enterprise credit stimulates economic development in advanced OECD countries over the period 1995-2011. Most control variables show the expected coefficients. The only exception is the inflation rate. In advanced OECD countries, estimated results show that the inflation rate significantly and positively influences economic growth. One reasonable interpretation is the reverse causal impact of economic growth on the inflation rate.

6.4 Impacts of credits on economic growth: emerging economies versus advanced economies

The coefficients of total credit to GDP and domestic bank to GDP in emerging OECD counties are -0.036 percentage points and -0.017 percentage points respectively. In advanced OECD economies, those coefficients are -0.051 percentage points and -0.041 percentage points respectively. Different estimated results in emerging and

(26)

26

advanced OECD economies imply that the negative effect of total credit and domestic bank credit on economic growth in advanced OECD economies is stronger than in emerging OECD economies.

In 7 emerging OECD economies, the IV regression presents a significant and positive impact of enterprise credit on economic growth. However, enterprise credit in advanced OECD counties has almost no influence on economic growth. According to theoretical models, credits to enterprises accelerate innovation and thereby economic growth. My empirical results in emerging and advanced OECD economies suggest that credit to non-financial corporations in developing OECD countries significantly improves economic growth and the positive impact of enterprise credit on economic growth in developed OECD counties is almost zero. This may imply that lending to enterprises in advanced OECD economies is not efficient. Enterprise credit includes both public and private non-financial corporations. Because of lack of data, I cannot examine whether lending to public non-financial corporations results in inefficient allocations of credit in developed OECD economies.

The coefficient of household credit in emerging and advanced OECD economies is negative. This is in line with Jappelli & Pagano’s (1994) findings. The increase in household credit reduces the saving rate and hence leads to a decline in economic growth. However, the negative sign of the coefficient of household credit in emerging OECD economies is insignificant. Therefore, lending to households in emerging counties may not hurt the economy.

This sub-section intends to compare estimated results in advanced and emerging OECD economies. The significant and negative signs of coefficients of total credit and domestic bank credit are confirmed in both developing and developed OECD countries. However, the impact of enterprise credit on economic improvement can only be confirmed in emerging OECD economies. One percentage point increase in enterprise credit to GDP ratio increases annual GDP per capita growth by 0.034

(27)

27

percentage points. Estimated results in advanced OECD economies only indicate that lending to household results in a decline in economic growth. One percentage point increase in household credit to GDP ratio reduces annual GDP per capita growth by 0.049 percentage points.

7. Conclusion

This paper mainly discusses independent impacts of different credit compositions on economic growth, using different specifications and estimation techniques on a sample of 24 OECD countries over the period 1995-2011. The multiple OLS and IV regressions are employed to estimate the effect of household credit and enterprise credit on economic development. In contrast to early growth regressions, the dummy recession indicator is employed to capture the business cycle movement and the specific country effect is controlled. In addition, the impact of household credit and enterprise credit on economic growth in advanced and emerging economies is analyzed separately.

Estimated results confirm the positive impact of enterprise credit on economic growth. The increase in household credit, total credit and domestic bank credit significantly reduce annual GDP per capita growth. In addition, results in 7 emerging OECD economies suggest that lending to enterprises supports economic growth. In 17 advanced OECD economies, the increase in household credit results in a reduction in economic growth. My empirical results on a sample of 24 OECD countries confirm the theoretical model that enterprise credit has a positive impact on economic growth. The negative effect of household credit on economic growth shows that credits are not efficiently spent by households. Moreover, emerging and advanced OECD economies face different situations. Estimated results only indicate a positive impact of enterprise credit on economic development in emerging OECD economies. The increase in household credit may not decrease economic growth in 7 emerging OECD economies. Oppositely, the household credit significantly reduces economic

(28)

28

growth in 17 advanced OECD countries. Enterprise credit has no influence on economic growth in advanced OECD economies.

With the development of financial systems, the importance of credit market is increasing. Compared with enterprise credit, the growth of household credit is more rapid. My empirical results suggest that supporting household credit development leads to a reduction in economic growth. Therefore, the dramatic increase of household credit should be controlled. Although my aggregate results are in the line with Sassi & Gasmi (2012), the impact of enterprise credit on economic growth in advanced OECD countries is not obvious. This implies that lending to enterprises in advanced OECD economies may not be efficient. However, my empirical study confirms the significant and positive impact of enterprise credit on economic growth in emerging OECD economies. The implication is that supporting enterprise credit market is better than supporting household credit market, especially in emerging OECD countries. Banks or other financial institutions should lend to productive enterprises, which improves economic development.

Although my empirical studies confirm the theoretical models that different credit compositions have significant influences on economic growth, a reverse causality between the credit market and economic growth cannot be ignored. The “finance-led growth nexus” among economists has been debated for many years. As mentioned in section 2 of the literature review, many theoretical and empirical papers support that the development of financial systems encourages economic growth. However, the other hypothesis indicates a direction from economic growth to financial development. For instance, the demand-following hypothesis postulates that the growth of real economy needs efficient financial services. Thus, the development of the credit market is promoted by economic growth. My empirical study does not investigate that possible reverse causality. The reverse causality between credit compositions and economic growth is an interesting topic, which need to be addressed in future research.

(29)

29

8. References

Aghion Philippe, Howitt Peter, Mayer-Foulkes David (2005). “The effect of financial development on convergence: theory and evidence”. Quarterly Journal of

Economics 23. P 173-222

Barro Robert J. (1995). “Inflation and Economic Growth”. NBER Working Paper No. 5326. P 1-22

Banerjee Abhijit V. (2001). “Contracting Constraints, Credit Market and Economic Development”. Working paper, MIT Dept. of Economics, No. 02-17. P 1-55

Beck Thorsten, Demirguc-Kunt Asli, Levine Ross (2005). “SMEs, Growth, and Poverty”. Working paper 11224, National Bureau of Economic Research. P 1-43

Beck Thorsten, Ross Levine (2005). “Legal Institutions and Financial Development”. In: Menard,C., Shirley,M.(EDs.), Handbook of New Institutional Economics. Kluwer Dordrecht.

Beck Thorsten, Buyukkarabacak Berrak, Rioja Felix, Valev Neven (2008). “Who gets the credit? And Does it matter? Household vs. Firm lending across countries”, The B.E. J. Macroecon., De Gruyter 12(1). P 1-46

Bencivenga Valerie R., Smith Bruce D. (1990). “Financial Intermediation and Endogenous Growth”. Review of Economic Studies (1991) 58. P 195-209

Bose Niloy (2002). “Inflation, the credit market, and economic growth”. Oxford Economic paper 54. P 412-434

(30)

30

Dembiermont Christian, Mathias Drehann, Muksakunratana Siripon (2013). “How much does the private sector really borrow? A new database for total credit to private non-financial sector”. BIS Quarterly Review March 2013. P 65-81

De Gregorio Jose (1996). “Borrowing constraints, human capital accumulation, and growth”. Journal of Monetary Economics, Vol 37, 1. P 49-71

Galor Oded, Zeira Joseph (1993). “Income Distribution and Macroeconomics”. The Review of Economic Studies, Vol 60, 1. P 35-52

Hung Fu-Shen, Cothren Richard (2001). “Credit market development and economic growth”. Journal of Economics and business 54 (2002). P 219-237

Jappeli Tullio, Pagano Marco (1994). “Saving, Growth, and Liquidity Constraints”. Quarterly Journal of Economics, Vol 109, 1. P 83-109

King Robert G., Levine Ross (1993). “Finance and Growth: Schumpeter Might Be Right”. Quarterly Journal of Economics, Vol 108, 3. P 37-717

King Robert G., Levine Ross (1993). “Finance, entrepreneurship, and growth: Theory and evidence”. Journal of Monetary Economics 32. P 513-542

La Porta , Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer (1999). “The Quality of Government”. Journal of law, Economics and Organization 15. P 222-279

Levine Ross (1997). “Financial Development and Economic Growth: Views and Agenda”. Journal of Economic Literature, 35. P 688-725

(31)

31

Odhiambo Nicholas M. (2007). “Supply-leading versus Demand-following

Hypothesis: Empirical Evidence from Three SSA countries”. African Development Review, Vol 19, 2. P 257-280

Sassi Seifallah, Amira Gasmi (2012). “The effect of enterprise and household credit on economic growth: New evidence from European union countries”. Journal of Macroeconomics, Vol 39. P 226-231

Stulz Rene M., Williamson Rohan (2003). “Culture, openness, and finance”. Journal of financial Economics, Vol 70, 3. P 313-349

Rajan Raghuram G., Zingales (1998). “Financial Dependence and Growth”. Journal of American Economic Review, Vol 88, 3. P 559-586

(32)

32

9. Appendix

Figure A1 ： this graph illustrates the evolution of credit to GDP in 7 advanced OECD countries over the period 1995-2011. (Source: own calculation)

Figure A2: this graph illustrates the evolution of credit to GDP in 7 emerging OECD countries over the period 1995-2011. (Source: own calculation)

0 0.5 1 1.5 2 2.5 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 cr e d it to G D P ratio %

Advanced economies

Canada Germany United Kingdom USA France Austria Japan 0 0.5 1 1.5 2 2.5 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 cr e d it to G D P ratio %

Emerging economies

Czech Republic Greece Hungary Korea Mexico Poland Turkey

(33)

33

Figure A3: this graph illustrates the evolution of highest credit growth in 4 OECD countries over the period 1995-2011. (Source: own calculation)

Figure A4：This graph shows the negative correlation between bank credit and total credit. A simple regression indicates that bank credit to total credit ratio has a significantly negative effect on total credit (the coefficient is -0.62). But the R-square is very low (only 0.0688). (Source: own calculation)

0 0.5 1 1.5 2 2.5 3 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 cr e d it to G D P ratio %

Economies with high credit growth

Denmark Norway Sweden Belgium

average credit of 24 OECd countries 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.5 1 1.5 2 2.5 3 b an k cr e d it/ to to al c re d it % total credit/GDP %

The relationship between bank credit and total

credit

(34)

34 Table B4：This table shows the correlation between annual GDP per capita growth, the logarithm of total credit to GDP, the logarithm of domestic bank credit to GDP, the logarithm of enterprise credit to GDP and the logarithm of household credit to GDP in 24 OECD countries from 1995 to 2011.

Correlation Annual GDP per capita growth Total credit to GDP Domestic bank credit to GDP Enterprise credit to GDP Household credit to GDP

Annual GDP per capita growth

Totoal credit to GDP _-0.2430

Domestic bank credit to GDP _-0.2259 _0.9151

Enterprise credit to GDP _-0.2437 _0.9067 _0.8635

Household credit to GDP _-0.2027 _0.9469 _0.8457 _0.7348

Source: (own calculation)

Table B7：This table shows the correlation between annual GDP per capita growth, the logarithm of total credit to GDP, the logarithm of domestic bank credit to GDP, the logarithm of enterprise credit to GDP and the logarithm of household credit to GDP in 7 OECD Emerging countries from 1995 to 2011.

Enterprise credit to GDP _0.0181 _0.9524 _0.8877

(35)

35 Table B9：This table shows the correlation between annual GDP per capita growth, the logarithm of total credit to GDP, the logarithm of domestic bank credit to GDP, the logarithm of enterprise credit to GDP and the logarithm of household credit to GDP in 17 advanced OECD countries from 1995 to 2011.

Enterprise credit to GDP _-0.2703 _0.8003 _0.4026