Credit Is What Credit Does:

A Functional Differentiation of the Credit-Growth Relation

Dirk Bezemer1∗, Maria Grydaki2 and Lu Zhang1

1
Faculty of Economics and Business, University of Groningen
2 Management School, University of Stirling

ABSTRACT

We use newly collected data to re-examine the credit-growth nexus over 1990-2011 in 46

economies. We distinguish between the uses of credit into ‘nonfinancial business and

consumption’ and ‘financial and real estate’ credit, and between stock and flow effects. We

observe strong growth of credit stocks relative to GDP over 1990-2011. The share of

nonfinancial-sector credit in total credit decreased substantially. The mixed growth effects

of total-credit flows are due to lumping together credit supporting income growth and

credit supporting asset market growth. Distinguishing between these yields robust

positive growth effects of credit flows, with the growth effect of nonfinancial-sector credit

flows several times larger than the growth effect of mortgage and financial business credit

flows. But we find insignificant or negative ‘financial development’ effects of credit stocks

and the positive liquidity effect diminishes at higher levels of financial development. All

this suggests that many of the economies in our sample suffer from ‘too much finance’

(Arcand et al. 2012). The results are robust to using the Rajan and Zingales’ (1998)

methodology and GMM specifications. We discuss implications.

Key Words: credit, growth, stocks, flows, dynamic panel analysis

JEL codes: E44, O16, O40, C33

∗
E-mail addresses: d.j.bezemer@rug.nl (Dirk Bezemer, corresponding author), maria.grydaki@stir.ac.uk
(Maria Grydaki) and lu.zhang@rug.nl (Lu Zhang). We thank seminar participants in the EWEPA June 2013
conference in Helsinki and the September 2013 2nd IES seminar in Prague for helpful comments on earlier
drafts of this paper. Any remaining errors or omissions are ours. We acknowledge financial support from the
Equilibrio Foundation and the Institute for New Economic Thinking (grant INO11-00053).
1
 Credit Is What Credit Does:

A Functional Differentiation of the Credit-Growth Relation

1. Introduction

A large empirical literature has established the positive effects of the growth in bank credit

on output growth.1 Recent research however shows that above a threshold level, high

credit-to-GDP growth may slow down growth (Arcand et al., 2012; Cecchetti and

Kharroubi, 2012) or even that “banking development has an unfavourable, if not negative,

effect on growth” (Shen and Lee, 2006:1907). Credit growth correlations which were

positive until the 1990s are insignificant or negative since (Valickova et al., 2013; Rousseau

and Wachtel, 2011). One reason is that high credit-to-GDP growth precipitates crisis

(Schularick and Taylor, 2012; Jorda et al., 2012). But even beyond crisis, there may be a

changing relation between credit and growth over the long financial cycle (Borio, 2012).

Cecchetti and Kharroubi (2012:15) conclude that “there is a pressing need to reassess the

relationship of finance and real growth in modern economic systems. More finance is

definitely not always better.”

We use newly collected data to re-examine the credit-growth nexus over 1990-2011 in

46 economies. We make two contributions. First, we distinguish between stock and flow

effects of credit on growth (Biggs et al., 2010). Credit flows increase agents’ ability to

finance expenditures and they have an immediate positive effect on recorded output. We

call this the ‘liquidity effect’ of credit. Credit stocks increase agents’ ability to reallocate

production factors and may so support growth. This is the traditional focus of the credit-

growth literature. We call this the ‘financial development effect’ of credit. But credit stocks

are also debt stocks, which may depress growth through the burden of debt servicing or

1The empirical literature started with King and Levine’s (1993) seminal Finance and Growth: Schumpeter Might
Be Right. This literature builds on Schumpeter (1934), Goldsmith (1969), McKinnon (1973) and Shaw (1973).
Levine (2005) and Ang (2008) provide overviews. Beck et al. (2009) show recent empirical results.
2
through financial instability and debt crisis. Thus, the effect of credit stocks on growth is a

priori ambiguous and may be negative, in line with recent studies.

Our second contribution is to distinguish between the uses of credit into ‘nonfinancial

business and consumption’ and ‘financial and real estate’ credit (as also in Werner, 1997,

2012; Beck et al., 2009; Büyükkarabacak and Krause, 2009; Büyükkarabacak and Valev,

2010). Bank credit to nonfinancial business and consumption loans to households both

finance transactions in goods and services, which implies a direct link to growth.

Mortgage credit to households and credit to nonbank financial business finance

transactions in assets (including real estate), which has an indirect impact on output

growth through balance sheet effects. These two differentiations of credit (into stocks and

flow, and into ‘nonfinancial’ and ‘financial and real estate’ credit) suggest a two-by-two

matrix of credit-growth effects, with different channels relevant to each quarter.

Our new data show that over 1990-2011, there was strong growth of credit stocks

relative to GDP, in line with the literature. What is not found in the literature so far (which

uses data until 2004 or 2005 at the latest) is how exceptionally large the growth in

credit/GDP was after 2004: on average from .73 to .95 in our sample. We also find that the

share of nonfinancial credit in total credit decreased substantially, with a strong rise in the

share of especially mortgage credit. We observe strong positive correlation of growth in

nonfinancial-sector credit with output growth, and substantial negative correlation of all

credit stocks with output growth.

In the empirical analysis, we explore the credit-growth relation in three specifications:

fixed-effect panel data regressions, dynamic panel estimations (GMM models) and Rajan

and Zingales’ (1998) methodology. We find that the growth effect of nonfinancial credit

flows is several times larger than the growth effect of financial-sector and real estate credit

flows. Correlations of credit stocks (of either type) with output growth are negative.

Interaction effects between credit stocks and credit flows are also negative. Each of these

findings is robust across specifications. We also conduct robustness checks with respect to

3
the method of constructing credit aggregates, the level of economic development, the

business cycle, and the impact of nonbank borrowing.

The results suggest that many of the economies in our sample suffer from ‘too much

finance’ (Arcand et al., 2012). Negative debt effects of credit stocks seem to more than

outweigh positive reallocation effects, for all categories of credit. With higher levels of

financial development, positive liquidity effects on growth of credit flows diminish. We

also find that distinguishing between credit categories is helpful in understanding flow

effects.

In the next section we discuss credit’s different effects on output growth and make

links to the literature. In section 3 we present the new data. In section 4 we discuss the

methodology. Section 5 presents the empirical findings and in section 6 we run a series of

robustness checks. Section 7 concludes the paper with a discussion of the results and

suggestions for future research.

2. Re-examining the credit-growth nexus

In this section we introduce the two conceptual contributions in this paper and connect

them to the literature. These are (i) to investigate the effect on output growth of stocks and

flows of credit separately, (ii) for different categories of credit. There are only few papers

on each of these issues; none so far combines them.

2.1. Re-examining the credit-growth nexus: stocks and flows

Both stocks and flows of credit relate to GDP growth, but in different ways (Biggs et al,

2009). Credit flows increase agents’ ability to finance expenditures. This is a direct short-

term ‘liquidity effect’ on output, since “[l]oans cause deposits and those deposits cause an

expansion of transactions” (Caporale and Howells, 2001:555; also Borio and Lowe, 2004).

Credit stocks, on the other hand, reflect agents’ ability to use the outstanding stock of

loans to reallocate factors of production to more productive uses, which may support

growth. But since all credit is debt, credit stocks are also debt stocks, which may depress
4
growth through more financial fragility and larger uncertainty, through larger debt

servicing out of income (which diminishes effective demand), or through a negative

wealth effect on consumption. Another way in which a large financial sector may harm

growth is by competing for resources with the nonfinancial sector (Cecchetti and

Kharroubi, 2012).

At low levels of the credit-to-GDP ratio these may be minor side effects, outweighed

by the benefits of more credit available to re-allocate resources to more productive uses. At

high credit-to-GDP levels, a net negative effect of credit stocks on growth may result. Since

the stock of credit is commonly used as a measure for financial development, we call its

effect on growth the ‘financial development’ effect. Given separate stock and flow effects,

controlling for flow effects is important in assessing the (traditionally studied) stock

effects.

With the exception of Biggs et al. (2010), this distinction is new to the empirical credit-

growth literature, but it is far from novel. In the fiscal macro literature, it is well known

that flows of government deficit spending may boost growth in the short term, but by

simultaneously raising stocks of public debt, growth may be harmed. The impact of

deficits and of debt levels differ. What goes for public debt, goes for private debt. Our

distinction between stocks and flows is not more profound than this.2

To illustrate, consider an entrepreneur who takes out a loan and spends this on fixed

capital formation. The seller may also spend all or some of the money on yet other goods

and services, and so on. Thus, the immediate effect is an increase in transactions in capital

goods and other goods and services, which translates into an increase in recorded GDP.

This is the liquidity effect of new credit, measured by annual credit flows. After this effect

has been accounted for, the loan has longer-term effects, measured by credit stocks (the

accumulation of past flows). The enlarged liquidity in the economy may help re-allocate

production factors to more productive uses, and so help growth. And the loan has created

an equivalent debt for the entrepreneur, which may hinder or even harm growth in several

2 We thank Jan Libich for this point.

5
ways, as just explained. The net effect of credit stocks (which are simultaneously debt

stocks) depends on the balance between these helpful and harmful potentialities.

In the recent literature there is some evidence of net negative or absent financial

development effects, alongside the positive liquidity effects (which remain mostly

unexplored). Beck et al. (2012) distinguish between the financial sector’s ‘intermediation

activities’ akin to the liquidity effect (which they find increases growth) and its size (which

has no effect on growth). Arcand et al. (2012) use different empirical approaches to show

that there can indeed be ‘too much finance’ – finance starts having a negative effect on

output growth when credit to the private sector reaches 110% of GDP. Importantly, their

measure includes all credit, not just bank credit. Cecchetti et al. (2011) also observe that

‘beyond a certain level, debt is a drag on growth’. They assess that for government and

household debt, the threshold is around 85% of GDP while for corporate debt, it is around

90% of GDP.

Different explanations have been proposed. Rousseau and Wachtel (2011) consider

that since the 1990s many countries liberalized their financial markets before the

associated legal and regulatory institutions were sufficiently well developed, undermining

the positive impact of financial deepening on growth. Arcand et al. (2012) develop a model

in which the expectation of a bailout may lead to a financial sector which is too large with

respect to the social optimum. Cecchetti and Karroubi (2012) note that the financial

industry competes for resources with the rest of the economy, particularly with skill-

intensive parts. They present evidence that more skill and R&D intensive industries suffer

more productivity losses during a financial boom. Earlier Stockhammer (2004) analysed a

causal relation for selected OECD economies between expanding asset markets and a

slowdown in fixed capital formation. Easterly et al. (2000) point to a volatility channel.

They show that the volatility of growth tends to decrease and then increase with

increasing financial depth.

We add to this the observation that credit is also debt, which may explain negative

effects of financial development at high credit/GDP levels which implies high private
6
debt/GDP levels. We are able to test this interpretation by separately studying the effects

of debt stocks and credit flows. None of the studies just quoted differentiates between

stock and flows effects. One exception is Biggs et al. (2009, 2010), who show that GDP is a

function of both the stock and the flow of credit. In a theoretical model they suggest that

“for reasonable values of depreciation and interest rates, the coefficient on the flow of

credit is substantially larger than the coefficient on the stock of credit (Biggs et al., 2009: 3).

In this paper we find that this is the case in our data; indeed, the coefficient on the stock of

credit is negative.

2.2. Re-examining the credit-growth nexus: the uses of credit

Our second contribution is to “distinguish between different categories of credit, which

perform different economic functions”, as the LSE The Future of Finance report urges

(Turner et al., 2010:16). Cecchetti et al. (2011) distinguish credit by sector, as government,

household and corporate debt. Büyükkarabacak and Valev (2010) distinguish between

(nonfinancial) enterprise and household credit, as do Beck et al. (2012). We distinguish

debt by its function. In particular, we recognize that the economy is composed of a real

sector where goods and services are produced and traded, and a financial and real estate

sector whose primary function is to originate and trade assets, including real estate.3

Consequently, two broad functions of bank credit can be distinguished as: financing

production and consumption, and financing asset transactions (with existing real estate

being the most important asset category in most economies). These uses of credit can be

(imperfectly) mapped onto the credit categories of: credit to nonfinancial business plus

household consumption loans on one hand, and mortgages plus credit to financial

business on the other hand. Because bank credit to nonfinancial business and consumption

loans to households both finance transactions in goods and services, there is a direct link

3We distinguish between credit to the construction industry (which is part of nonfinancial business) and
mortgages.
7
to growth in GDP, which is the total value of final transaction in goods and services.4 This

direct link does not exists for mortgages and credit to financial business, which finance

asset transaction. Asset transactions are not included in GDP, but may have indirect effects

on GDP.

Therefore credit facilitating investment in financial capital may provide income

through capital gains, interest and dividends to individuals, but not necessarily national

income growth through investment and output growth. Whereas the depth of financial

markets is usually taken as “a measure of economies’ absorption capacity” (Brezigar

Masten et al., 2008), here we recognize that growth in asset markets can outpace the

economy’s capacity to productivity invest financial resources, as also in the Boissay et al.

(2013) theoretical model. This may happen already at low levels of credit-to-GDP ratios,

but with high credit growth rates. What matters is “how large a credit boom [is] relative to

the possibilities of productive uses for loans” (Boissay et al., 2013:4; also Lorenzoni, 2008).

The same distinction is relevant to the credit-instability nexus. If credit-financed asset

market expansion leads to credit growth beyond GDP growth, credit-to-GDP ratios are

rising, possibly to leverage levels which lead to financial fragility and instability, or which

imply large debt servicing burden or negative wealth effects. Thus, it is no coincidence to

see a fall in the share of nonfinancial sector credit in total credit, simultaneous with rising

credit-to-GDP ratio’s and a falling growth effect from credit. Borio and Lowe (2004) show

that the combination of credit booms and asset price booms (rather than just rapid credit

growth) is a good predictor of financial instability. That is, specifically credit that inflates

asset prices is a cause for rising debt-to-GDP ratios and financial instability.

Indeed, it is on the markets for mortgages, stocks, bonds and derivative products that

financial fragility typically develops, often leading to crisis and lower growth. Much of

4 There is a theoretical argument that consumer credit does not lead to higher growth since it merely smooths
consumption. Several studies include mortgages into an overall category of ‘household credit’ and find that
it has no growth effect (e.g. Beck et al., 2012, Büyükkarabacak and Krause, 2009). This is compatible with our
observation that mortgage credit does not finance consumption, hence has no direct growth effect. Note that
consumption is here consumption of goods and services (included in GDP), not of assets (not included in
GDP).
8
this is linked to household credit: mortgages and its derivatives are a large part of such

asset markets, and large share of asset market investment goes through pension funds

which allocate household savings. In two recent studies, Beck et al. (2012) and

Büyükkarabacak and Valev (2010) find that credit to households (most of which is

mortgages, in most economies) has negligible growth effects, while credit to nonfinancial

business has strong growth effects. Similarly, Jappelli and Pagano (1994) argue that more

household credit leads to lower private savings and so slower economic growth. Jappelli

et al. (2008) and Barba and Pivetti (2009) find positive crisis/instability effects of the

expansion of household credit. Sutherland et al. (2013:2) find that "when private sector

debt levels, particularly for households, rise above trend, the likelihood of recession

increases." Note that household credit expansion resulting in real estate and asset market

investments by definition leads to rising credit-to-GDP ratios and thus to private debt

levels, while enterprise credit expansion does not.5 This observation links household debt

to the rise in overall debt levels, which has been widely noted as a factor increasing the

risk of crisis (Lorenzoni, 2008; Barajas et al., 2013; Reinhart, 2010; Schularick and Taylor,

2012; Jorda et al., 2012; Boissay et al., 2013).

3. Data

We collected data from the consolidated balance sheet of Monetary Financial Institutions

observed in central bank sources, for 50 countries over 1990-2011. On the asset side of the

balance sheet, loans to nonbanks are reported separately as mortgages to households,

household consumption credit, credit to nonfinancial business, and credit to financial

business (insurance, pension funds, and other nonbank financial firms).6 To the best of our

5 Loans spent on assets do not increase recorded GDP but they do increase credit and debt levels, so that the
credit-to-GDP ratio rises. Loans spent on goods and services increase both GDP and credit stocks, at stable
credit-to-GDP ratios. We show this also in the regressions below.
6 A fifth category is bank lending to government, which is however often not reported and in any case

mostly small.
9
knowledge, no data with similar detail has been collected and reported before.7 In the

Appendix we report sources and compare our data to other data sets. In this section we

introduce definitions for the key variables in the analysis: stocks and flows of different

credit categories. We discuss their development over time.

3.1. Measuring stocks and flows of credit categories

We measure credit flows by the annual change of credit stocks relative to lagged GDP (as

in Biggs et al., 2009):

Ci ,t − Ci ,t −1
FLOW =
GDPi ,t −1

where i denotes country, t denotes year and C is a credit measure. We define credit stocks

conventionally as

Ct
STOCK =
GDPt

We aggregate the four categories of credit which were observed in the raw data into two

broader categories: credit to the nonfinancial sector and credit to the financial and real

estate sectors. The former consists of credit to non-financial business plus household

consumption loans. The latter includes mortgages plus credit to financial business, as in

the ‘finance, insurance and real estate’ sectors classification of the U.S. National Income

and Product Accounts. The motivation for this classification, and for our credit aggregates,

is that nonfinancial-sector credit directly finances transactions in goods and services, while

financial-sector credit (including real estate) finances the creation and trading of financial

and real estate assets. In this respect, the present paper differs from other studies which

split credit into ‘enterprise’ and ‘household’ credit. As we will see below, in practice the

difference is not a large one on average, but it is in countries where most household credit

is non-secured consumption credit.

7 Related data sets are in Beck et al. (2012) and BIS (2013). The Beck et al. (2012) data has shorter time
coverage than our data. The BIS (2013) include both equity and bank credit and does not differentiate bank
credit. We refer to the Data Appendix for a comparative discussion.

10
 While this delineation is useful with respect to credit’s growth effects (as just

explained), we emphasize that its measurement is necessarily imprecise, for a number of

reasons. For instance, mortgage credit often also serves as consumer credit through home

equity withdrawals, while business credit often includes business mortgage credit.

Financial businesses do not only trade assets but also deliver services. Conversely,

nonfinancial businesses realize part of their returns in trading financial assets (see e.g.

Krippner, 2005 on the U.S.). But as a broad distinction this categorization will help us trace

the different effect of different types of credit on GDP growth. For robustness purposes,

we will also analyze growth effects of all four types of credit, to see whether our broad

characterisation of credit is justified.

3.2. Trends in the data

The 1990-2011 data show three trends which are of interest: the expansion of credit relative

to GDP over time, the changing composition of credit stocks, and the relation of stocks and

flows of credit categories with economic growth. Figure 1 shows that on average, the total-

credit-to-GDP ratio increased from 73% to 95% over 1990-2011. The increase was

particularly pronounced in Spain, where the credit-to-GDP ratio rose from 118% to 369%;

in Greece, from 33% to 115%; in the UK, from 39% to 90%. Declines in the credit-to-GDP

ratio were rare and often associated with episodes of financial crisis (for instance in

Pakistan, the Philippines, Hungary and Argentina). Figure 1 further shows that on

average, much of the increase in credit relative to GDP occurred after 2002. It peaked and

then declined slightly after the 2008 financial crisis.

<Figure 1 Total bank credit stocks over 1990-2011>

A second trend is the changing composition of credit. Table A1 in the Appendix shows

that on average lending to nonfinancial business and household mortgage lending are the

two principal credit categories. Especially in OECD economies, the share of credit to

11
financial business and real estate in total credit increased strongly. For instance, in the

Netherlands this combined share was less than 30% in 1990 but 48% in 2011. Conversely,

the share of total bank credit flowing to nonfinancial business dropped from 55% to 30% in

the Netherlands. On average over all countries, the 1990-2011 decline in the nonfinancial

business credit share was from 47% to 37%. This was balanced by increases in the shares of

financial business (from 8% to 11%), non-secured consumer credit (from 12 to 13%) and,

especially, mortgages (from 32% to 38%). Figure 2 shows the evolution of bank credit

stocks over time.8

< Figure 2 Average bank credit stock composition (top) and ratio to GDP (bottom), 1990-2011>

Figure 2 shows that the rate of mortgage credit growth was higher than any of the other

categories from the mid-1990s to the mid-2000s. Stocks of mortgage credit surpassed

nonfinancial business credit stocks from 2002, mostly. Nonfinancial business credit

experienced the steepest growth after 2002 and the steepest fall after the 2007 crisis, which

reflect the real-sector impact of the credit boom and bust (Borio, 2012). These dynamics

prompt a rethink of banks’ role, traditionally conceived as financing investment and

production in the real sector. In 2012, more than half of outstanding bank loans stocks

finance real estate and financial assets and instruments.

A third set of observations is on the credit-growth relation, for stocks and flows of

credit. There appears to be a robustly negative cross-section relation over 1990-2011

between private credit stocks relative to GDP and real per capita GDP growth, though with

significant scatter and possible nonlinearity around the trend line (Figure 3a). There also

8This is a lower bound measure, since the data do not include securitized credit, which tends to be larger for
real estate and financial-sector credit than for nonfinancial-sector credit . If banks sold loans to nonbanks we
no longer observe them, since our data are taken from MFI balance sheets. The difference can be substantial.
For the Netherlands, total credit including securitized loans in special purpose vehicles (SPV) was 30%
higher in 2008 (rising from zero until 1998) than total credit without securitization, and none of this was in
nonfinancial business credit or nonsecured consumer loans. Thus the credit data used in this paper overstate
the share of credit to the nonfinancial sectors in the Netherlands, and plausibly in a number of other
countries. To our knowledge, data on loan securitization are not available in consistent cross-country format.
12
appears to be positive correlation over time of per capita output growth with total-credit

flows (Figure (3b), also Biggs et al., 2010).

<Figure 3 Credit stocks, credit flows and economic growth over 1990-2011>

Finally, Table 1 explores correlations between the stocks and flows of the two credit

aggregates and four categories of credit and growth. Panel A shows that total credit stocks

have a significant negative correlation with growth. This is mainly driven by mortgages

and (to a lesser extent) financial-sector credit. The correlation between real-sector credit

stocks and growth is less negative and less significant. Credit flows to the nonfinancial

sectors have the highest correlation with growth, closely followed by its two components,

nonfinancial business credit and household consumer loans. In contrast, the growth

correlations of credit flows to financial sectors and real estate are less than half of those for

the nonfinancial sectors. We note also the very high correlations of total credit stocks with

mortgage credit stocks.

<Table 1 Do credit stocks and flows correlate with economic growth?>

4. Methodology

We regress real GDP per capita growth on annual stocks and flows of total credit and

the two credit categories (all as GDP ratios) controlling for initial income levels,

government spending, trade openness, inflation, years of schooling and institutional

quality (control variables are selected based on the literature). We start with a fixed-

effect panel data baseline model over 1990-2011 for 50 countries. Then we estimate

GMM models to take into account endogeneity of credit variables. As in many finance-

growth studies, we use 3-year averages of the underlying annual data. Descriptive

statistics are reported in Table 2. The baseline specification is:

git = α + β1cit + β 2 LEit + γ Xit + δ yi ,t 0 + ϕi + ψ t + ε it (1)

13
where git is the growth rate of real GDP per capita of country i in three-year period t, yi,t0

the level of real GDP per capita at the beginning of t and β1 and β2, coefficients capturing

the growth effects of credit stocks (ci), and credit flows (LEit), separately for stocks and

flows of nonfinancial-sector credit and financial-sector plus real estate credit.

Coefficient δ captures the convergence effect, Xit is the vector of control variables and εit

is the white noise error term with mean zero. We also include unobserved country-

specific time-invariant effects in φi and simultaneous shocks across countries in time

dummy ψt,. In the robustness analysis we will also include the square of total-credit

stocks, capturing any nonlinear financial development effects on growth (Deidda and

Fattouh, 2002) and an interaction term of credit flows with credit stocks.

<Table 2 Descriptive Statistics (3-year averaged data) >

Since it is always possible that higher growth causes acceleration of lending (rather than

the other way round), or an unobserved third factor causing both, the baseline

specification (1) may suffer from endogeneity. One way to account for this is to estimate

a generalized-method-of-moments (GMM) dynamic panel model (Arellano and Bover,

1995; Blundell and Bond, 1998). This combines regressions in levels and in differences,

yielding unbiased estimators for the coefficients of interest. We difference (1) to obtain:

14
 Δgit = β1∆cit + β 2 ∆LEit + γ∆Xit + δ∆yi ,t 0 + ∆ψ t + ∆ε it (2)

and then estimate the system GMM of equations (1)-(2) on 3-year averages, where the

endogenous credit variables are instrumented by their lags in equation (2). The

consistency of the GMM estimator depends both on the validity of the assumption that

the error term, ε, does not exhibit serial correlation and on the validity of instruments.

We apply Hansen test for over-identifying restrictions, testing for the overall validity of

the instruments, along with a test for second order serial correlation of the residuals.

An alternative approach to account for endogeneity of credit to growth is to apply

the Rajan and Zingales (1998) methodology to interact credit stocks with an industry-

specific measure for dependence on external finance. In this methodology, a positive

coefficient of this interaction term in a growth regression is interpreted as evidence that

industries that are more dependent on external finance grow faster at higher levels of

financial development. Rajan and Zingales (1998) and a number of subsequent

applications present evidence that financial development (or financial crisis) causes

growth (or contraction).9 We use the following specification:10

Growthi ,c ,t = α 1sharei ,c ,t 0 + α 2 crediti ,c ,t + α 3 ( crediti ,c ,t × external financial depedencec ) +

(3)
α 4 dummy var iables +

9 Using European micro-level data for 1996–2005, Bena and Ondko (2012) show that firms in industries
with growth opportunities use more external finance in financially more developed countries. This result
is particularly strong for firms that are more likely to be financially constrained and dependent on
domestic financial markets, such as small and young firms. Kroszner et al. (2007) use a similar approach
to show that sectors highly dependent on external finance experience a greater contraction during a
banking crisis in countries with deeper financial systems. Raddatz (2006) shows that sectors with larger
liquidity needs are more volatile and experience deeper crises in financially underdeveloped countries.
10 It is worth noting that in contrast to the past studies based on cross sectional data, we use panel

data. Therefore, our approach has one distinctive advantage compared to Rajan and Zingales (1998). By
including the credit variable itself, in addition to its interaction with financial dependence, our
specification allows for a more direct assessment of the effect of credit on growth. In this spirit, our
specification is closed related to Braun and Larrain (2005).

15
where i denotes industry, c denotes country and t denotes time (i.e., a 3-year period).

The variable credit includes a variety of measures on (different types of) credit stocks

and flows. The set of dummy variables include country, year, industry, industry-year

and industry-country fixed effects.

5. Results

Table 3 presents the results for total-credit variables of the fixed-effect panel baseline

model (columns 1-3) and the system GMM model (columns 4-6). Results for stocks are

in columns (1) and (4), results for flows in column (2) and (5) and both are included in

columns (3) and (6). Table 3 shows there are robust results on growth of credit/GDP

stocks, in line with recent papers which argue that the credit-growth relation has

weakened in recent decades (Rousseau and Wachtel, 2011; Ketteni et al. (2007) if credit

flows are not included: the weakly significant negative coefficient in the FE specification

still remains in the GMM model. This is the same even when we include credit flows:

credit/GDP stocks are still negatively correlated to growth in both specifications. This

motivates looking at credit aggregates separately.

<Table 3 Credit and Growth: Stock and Flow Effects>

We proceed to look ‘under the surface’ of the total-credit variable by distinguishing

credit to nonfinancial business and consumer credit from credit to financial business

and real estate. Table 4 reports the results, with baseline model results in columns (1)-

(6) and the corresponding system GMM results in columns (7)-(12), with identical

coefficient signs as in the baseline panel results. In all specifications the validity of the

instruments and the absence of second-order autocorrelation were ensured.

We find that both types of credit stocks correlate negatively to growth. A one

standard deviation increase in the stocks of either real-sector or financial-sector credit

16
corresponds to 0.7 standard deviation decrease in growth, which is equal to a 2.24%

income loss in this sample.11 In contrast, credit flows to the nonfinancial sectors are

positively and significant correlated to growth. It appears that the non-significance of

some of the result on total-credit flows in Table 3 were due to lumping together credit

supporting income growth and credit supporting asset market growth. The result in

column (9) implies that a one standard deviation increase in the nonfinancial-sector

credit flows is associated with a 0.23 standard deviation increase in growth, which is

equal to an additional .74% income growth in this sample.12 No positive growth effect

from credit flows to financial business and real estate can be detected.

<Table 4 Differentiating between categories of credit>

The Rajan-Zingales (1998) results are shown in Table 5. Columns (1)-(3) show the results

for total credit, columns (4)-(6) and (7)-(9) reports results for nonfinancial-sector credit

and financial-sector credit, respectively. The results are broadly in line with the panel

data estimations; if anything they strengthen the results. We find that the coefficient for

credit stocks is again consistently negative, with no difference between credit

categories. The coefficient for credit flows to the nonfinancial sectors is clearly positive,

but not for credit flows to financial business and real estate. The coefficient for credit

flows to the financial and real estate sectors (financing non-produced assets and real

estate) is insignificantly different from zero. The positive coefficients for the interaction

of all credit stocks and financial dependence suggest that industries which are more

dependent on external finance experience smaller growth-retarding effects from credit

11The calculation is based on the coefficients in column (9) and (12). The magnitude is (-0.09*25)/3.209=0.7
and (-0.074*32)/3.209=0.7 for nonfinancial-sector credit and financial-sector credit, respectively. 25 and 32
are one standard deviation of nonfinancial-sector and financial-sector credit stocks, respectively. 3.209 is
one standard deviation of output growth rate.
12 The calculation is 0.178*4.24/3.209=0.23, where 4.24 and 3.209 are one standard deviation of

nonfinancial-sector credit flows and output growth rate, respectively.

17
stocks. The implied differences are approximately 3 percentage points growth for both

real- and financial-sector credit stocks.13 Coefficients for the interaction terms between

credit flows and external financial dependence are insignificant.14 Overall, the results

from our industry-level analysis are in line with panel and GMM findings.

<Table 5 Credit stocks and flows effect on growth: industry-level evidence>

6. Robustness Analyses

In this section we probe the robustness of our results with respect to the method of

constructing credit aggregates, the level of economic development, financial crisis, the

impact of nonbank borrowing and interactions between stocks and flows of credit,

nonlinear credit-growth effects, country-specific banking crisis, and country outliers.

We estimate successive panel fixed-effects and GMM specifications including each of

these factors.

6.1. Credit aggregates and their components

We first explore how the results change when we replace the two credit aggregates with

their components. This is motivated by the concern that the aggregates might be hiding

heterogeneity in the credit-growth relation of their underlying components. This would

be apparent from growth coefficient signs of credit components which are opposite to

13 One way to understand this is to note that debt is most harmful to growth when its repayment
substitutes for investment and wages. In industries with larger external dependence on finance, where
new finance can more easily be attracted, debt can then more easily be refinanced rather than paid off.
14 This is unsurprising as ‘external dependence on finance’ is defined as the annual excess of investment

over profit, i.e. the annual flow of bank credit and other borrowing to finance investment. Interacting this
with credit flows is unlikely to produce significant coefficients.

18
those of their aggregates. We therefore run the analysis for each of the four underlying

credit categories. In Table 6 we show the baseline model panel fixed-effect result in

columns (1)-(4), and the corresponding system GMM results in columns (5)-(8).

<Table 6 The Four Categories of Credit and Their Relation to Growth >

First, we find that the negative relationship between credit stocks and growth holds

overall but is particularly strong for non-financial business credit and mortgage credit.

This is unsurprising since they constitute the bulk of their respective categories,

nonfinancial-sectors credit and financial-business plus mortgages. None of the four

components have coefficients with an opposite sign to their stock aggregate. This

suggests that the stock aggregates do not hide heterogeneity in the underlying credit-

growth relation.

Second, credit flows to non-financial business are positively related to growth in

columns (1) and (5). The financial business credit coefficient is positive only in column

(4). Coefficients for flows of mortgages and consumer are both insignificant in the GMM

results.

6.2. The credit-growth relation over levels of economic development

There is considerable heterogeneity of the credit-growth relationship across levels of

economic development. For example, Beck et al. (2012) show that the positive growth

effect of enterprise credit is only significant for countries between the 25th and 50th

percentile of GDP per capita, and insignificant beyond the 75th percentile. It is thus

important to understand how our measures of stocks and flows of credit categories

relate to income growth across development levels. Based on the distribution of the

sample-average GDP per capita, we construct two sub samples, one excluding countries

19
in the lowest quantile (a ‘high-income’ subsample), and the other one excluding the

highest quantile (a ‘low-income’ subsample).

The results are reported in Table 7. All credit stocks have negative growth

coefficients across development levels, in both panel fixed-effects and GMM

specifications. Credit flows to nonfinancial business have positive growth coefficients in

both panel fixed-effects and GMM specification in high-income countries, while for

low-income countries the coefficient is positive but insignificant in the GMM

specification, consistent with Beck et al. (2012).

<Table 7 The credit-growth relation across development levels>

6.3. The 2008 Financial Crisis and What Followed

Economic growth peaked in many economies in our sample before the 2008 crisis, and

was significantly lower in the aftermath of that crisis during 2008-2011. The link

between credit and growth varies over the business cycle (Braun and Larrain, 2005;

Borio, 2012; Jorda et al., 2012). In particular, the link between nonfinancial sector credit

and growth is much stronger in a simultaneous downturn as happened post 2007, while

at the same time the link between mortgages (which are fixed) and growth weakens

considerably during a crisis (see Biggs et al. 2010, and Figure 2). The concern may then

be that our results are driven by the extraordinary 2008-2011 years. To explore this we

construct a new sample by excluding the post-2007 observations and re-estimate both

our specifications. The results in Table 8 are entirely consistent with our longer sample:

credit stocks have significantly negative growth coefficients, nonfinancial-sectors credit

flows and have significantly positive growth coefficients, and credit flows to financial

business and mortgages have insignificant coefficients. Our main results are not driven

by the crises and recessions which followed the 2008 crisis.

<Table 8 Results excluding the crisis years 2008-2011>

20
6.4. Stock market development

Another important issue is that bank credit is one source of finance, with equity

markets being the other major source. For many countries in our sample stock markets

are absent or very small; for some, stock market borrowing is far more important than

bank credit as a source of external finance. This may potentially bias our results. To

explore this, we introduce stock market capitalization as a share of GDP as an

additional control variable. In Table 9 we report the results. Stock market capitalization

by itself is not a driver of growth. Its inclusion in the model does not affect the results

on credit stocks. Apparently, access to alternative finance does not alleviate the negative

effects of financial overdevelopment and high debt burdens. Plausibly, in a ‘too much

finance’ scenario both banks and stock markets are beyond their optimum size. This

merits further exploration in future work. But including stock market capitalization in

the model strengthens the growth coefficients of nonfinancial-and financial sectors

credit flows, and they are significant in the GMM specification.

<Table 9 Results including stock market capitalization>

6.5. Interactions between stocks and flows of credit

So far, we treated the growth effects of credit stocks and flows independently, as if the

effect of obtaining new loans is independent of debt levels. One can think of a number

of plausible mechanisms linking both, in most cases weakening the positive growth

effect of credit flows at higher levels of credit stocks. In particular, with rising debt

levels, a larger share of new loans will go towards refinancing existing debt rather than

investment and innovation in nonfinancial business (Stockhammer, 2004; Checchetti

and Kharroubi, 2012). Again, not accounting for these effects might partly drive our

results through omitted-variable bias. We therefore introduce an interaction term of

21
credit flows with stocks (financial development). Table 10 reports the results. We find a

robustly negative interaction effect between credit stocks and credit flows. At higher

levels of financial development (credit stocks), the liquidity effect of credit flows is

lower. This can be interpreted in terms of a conventional diminishing-return effect on

credit flows. Alternatively, it is possible that with higher debt/GDP ratios, more of new

credit is used to service debts rather than to invest and consume. While these

interpretations should be the subject of future research, here our primary aim is to note

that the stock and flow effects of credit aggregates in this specification are consistent

with our earlier results.

<Table 10 Does the effect of credit flows depend on credit stocks?>

6.6. Non-linear credit-growth effects

Recent studies postulate a non-linear effect of credit stocks (financial development) on

output growth (Arcand et al., 2012; Shen and Lee, 2006). In line with this literature, we

introduce a squared term of credit stocks in our specifications. Table 11 presents the

results. There is some evidence of nonlinearities in the credit-growth relation in the

panel regressions, but this is not robust to the GMM specification. We conclude that in

our data there is no strong evidence of nonlinerarity, and that the core results stand also

in this specification including a nonlinear term.

<Table 11 Accounting for nonlinearities>

6.7. Banking Crises

We examined whether the occurrence of crisis constitute an explanation of the negative

growth effect of credit stocks. The results are reported in Tables 12 and 13. Different

from Rousseau and Wachtel (2009), we show that our main results are not driven by

22
country-specific banking crisis. Too much credit is not just bad because it causes crisis,

but also before the crisis.

<Table 12 banking crises- all credit >

<Table 12 banking crises- real vs. financial credit >

6.8. Country outliers

We checked whether our results are driven by specific countries with high credit

stocks/low growth (Denmark and Switzerland) and low credit stocks/high growth

(Latvia and Uruguay). We also dropped Spain due to its large credit stocks. The results

are reported in Tables 14 and 15. We show that our main results are not driven by the

country outliers. The significance of effects of both credit stocks and flows improved.

<Table 14 Results after removing country outliers (All credit)>

<Table 15 Results after removing country outliers (Real vs. Financial credit)>

7. Summary, Discussion and Conclusions

In this paper we used newly collected data to re-examine the credit-growth nexus over

1990-2011 in 46 economies. We make two contributions. We distinguish between the

stock and flow effects of credit on growth and between the uses of credit into

‘nonfinancial business and consumption’ and ‘financial business and real estate’ credit.

Our new data show that over 1990-2011, there was strong growth of credit stocks

relative to GDP while the share of nonfinancial credit in total credit decreased

substantially. In the analysis, we find positive growth effects of credit flows to the

nonfinancial sectors, especially nonfinancial business, which suggests a positive

‘liquidity effect’. This diminishes at higher level of financial development, as a negative

23
interaction effect between credit stocks and credit flows suggests. There is no significant

liquidity effect of credit flows for mortgage and financial credit flows. We find

significantly negative effects on output growth of credit stocks of all categories,

suggesting that many of the economies in our sample suffer from ‘too much finance’

(Arcand et al., 2012). These results hold also when accounting for endogeneity in GMM

and Rajan and Zingales (1998) type models. They are also robust across a large number

of specifications and robustness checks. We conclude that both innovations in the

credit-growth study contribute to a better understanding.

Going beyond total-credit-to-GDP ratios or a ‘banks vs. markets’ approach, there is a

need to study the different effect of different kinds of credit, and to think hard about

what ’financial development’ is, and how it relates to debt build-up. We do this by

studying stock and flow effects of credit aggregates. If ‘credit is as credit does’, then we

attempt to map ‘what credit does’ as a stock and as a flow, and on asset markets as

distinct from goods-and-services markets. Overall, our findings are in line with

Rousseau and Wachtel’s (2011) observation that “the widespread liberalization of

financial markets that occurred in the late 1980s and early 1990s made financial

deepening less effective.” We show that in particular, it was the growth in credit to

asset markets rather than to nonfinancial business which made financial deepening less

effective. This is part of a new research agenda on credit, debt, growth and crisis.

This research agenda can be extended in several directions. First, the impact of

bond, equity and other asset market is equally subject to stock and flow effects, which

need to be included in the analysis of financial development. For some countries (such

as the US and UK), a focus on bank credit implies missing a large part of the picture.

Second, restricting the analysis to domestic credit is not appropriate in economies

where international capital flows and/or foreign banks are important for household and

business credit (e.g. Ireland, Hungary and Latvia). Third, much remains to be done in

the measurement of credit flows. Some central banks provide detailed public data while
24
others do not. Harmonization of data provision would greatly assist further research.

Finally, some papers have started modelling stock versus flow effects (Biggs et al., 2010)

and enterprise versus household credit (Japelli and Pagano, 1994) but most work in this

area so far is empirical, as is this paper. We are also in need of a better theoretical

understanding of the functional differentiation of credit. We trust this paper is an

inspiration.

25
Data Appendix

The aim of the data base is to provide a detailed description of monetary financial

institutions’ (banks and credit unions) loan assets where the counterparty is a domestic

non-government nonbank. We collected data from the consolidated balance sheet of

monetary financial institutions’ from central bank sources of 50 countries over 1990-

2011. On the asset side of the balance sheet, loans to nonbanks are reported. We

included a country in the data set if loans were reported separately for mortgages to

households, household consumption credit, credit to nonfinancial business, and credit

to financial business (insurance firms, pension funds, and other nonbank financial

firms).15

Domestic bank credit includes loans by both domestic and foreign banks, in

domestic and foreign currency. For reasons of consistency, it excludes non-bank lending

and securitized bank loans. There are some countries which have large nonbank debt

markets or much securitization, so that loan assets on banks’ balance sheets paint only a

small part of the picture. For one extreme example, this is why ‘total bank credit’ values

for the US are comparatively low: most credit in the US is nonbank credit (bonds and

short term paper) and a large part of loans (especially, mortgages) is securitized so that

it cannot be observed on banks’ balance sheets. The total stock of credit market

instruments relative to GDP in the US was 386% in 2011 (BEA flow of fund data), of

which only 34% was bank credit (this data). However, the US is exceptional in this

respect.

For each country, the source was always the country’s central bank. There is large

diversity in reporting formats. Only few central banks distinguish deposit taking

15An alternative would be to collect data from the liabilities side of the counterparty, in a country’s flow
of fund data. However, not all countries provide sufficiently detailed flow of funds data on bank loans by
sector. What is often reported is total borrowing, including equity market borrowing while we focus on
the analysis of bank credit. Also, to the extent that equity is held in the private nonfinancial sector, this is
a debt from the private nonfinancial sector to the private nonfinancial sector.

26
institutions within the broader category of Monetary Financial Institutions. Most do not

differentiate between lending to public sector firms and private sector firms, or between

domestic currency loans and foreign currency loans. Some central banks (e.g.

Switzerland’s) report credit to ten or fifteen business sectors of the economy separately,

which we collapsed into ‘financial’ and ‘nonfinancial’. Some report bank lending to

nonbanks as well as interbank lending (which we excluded from the data). In a few

cases (Israel), bank loans were only reported from the borrower’s viewpoint, i.e. as

liabilities. Some report only ‘household’ and ‘business’ lending. In these cases, we

assigned household lending to mortgages, unless we had evidence that it was

unsecured consumer lending. Some data go back much before 1990; Switzerland’s goes

back to 1906, the US to 1952. But on average, data before 1990 were rare.

We conclude this brief data description with two notes on lending to government

and on mortgages. In principle, to the extent that government spending financed by

bank loans causes GDP, this is one of the channels through which bank loans affect

GDP. However, most government spending is financed by bonds rather than private

loans, and total government lending by banks is usually small. We choose not to

include this in our data. Mortgages in our data are household mortgages, which is only

part of total mortgages. Some countries also report business mortgage lending

separately from other lending to business, and in these cases it is clear that a substantial

part of lending top business is lending secured by real estate. But the use of secured

lending to business will be more linked to production and trade, and thus GDP, while

the use of mortgages to household is almost exclusively to purchase real estate assets.

Thus, the impact on GDP will be different, which suggest that separating out

households mortgages is functional, but separating out business mortgages is less so.

Apart from that, it was not practicable to do this. Since only few countries report

business mortgages, we cannot consistently include total mortgages.

27
Table A1 Credit stocks across countries (% of GDP)

Non-
Financial
financial Consumer Mortgage Total
business
Country Start End business
ALB 1998 2011 14.555 . 3.912 32.775 36.116
ARG 1993 2011 10.520 2.987 2.013 . 15.519
ARM 2005 2011 11.164 4.070 1.938 . 17.172
AUS 1994 2011 34.709 7.473 48.527 6.784 97.493
AUT 1995 2011 47.674 . 38.202 6.609 92.485
BEL 1999 2011 30.167 . 33.217 1.369 64.753
BGR 1998 2011 25.696 6.409 5.261 . 37.365
BRA 1994 2011 19.671 . 7.822 3.887 31.380
CAN 1990 2011 18.373 25.991 46.601 0.213 91.179
CHE 1990 2011 43.152 . 92.904 1.516 137.571
CHL 1990 2011 43.381 36.427 11.262 . 91.070
CZE 1999 2011 17.040 5.253 9.828 3.768 35.890
DEU 1990 2011 52.984 10.877 28.536 3.203 95.601
DNK 2000 2011 44.121 25.340 78.656 7.189 155.306
EGY 1991 2011 36.225 . 7.714 . 43.939
ESP 1992 2011 59.677 62.062 106.913 . 228.651
EST 2001 2011 26.257 2.548 24.144 7.209 60.158
FIN 2009 2011 31.617 15.151 42.604 1.090 90.462
FRA 1993 2011 36.181 4.744 26.560 4.112 71.598
GBR 1990 2011 14.545 8.795 37.018 18.184 78.542
GRC 1998 2011 40.108 11.698 21.462 2.384 75.652
HKG 1990 2011 68.830 11.604 39.323 15.192 134.950
HRV 2001 2011 61.326 . 32.582 9.445 103.353
HUN 1990 2011 15.393 3.325 5.278 3.317 27.314
IDN 2002 2011 15.269 5.275 2.049 2.440 25.032
IND 2001 2011 25.292 3.259 3.248 2.364 34.162
ISL 2003 2011 14.132 . 41.737 . 55.869
ISR 1999 2011 58.939 10.813 20.979 . 90.730
ITA 1998 2011 22.691 2.776 23.921 11.673 61.062
JPN 1990 2011 56.596 3.209 28.040 8.535 96.380
KOR 2007 2011 60.225 20.535 30.291 4.793 115.843
LTU 1993 2011 18.373 2.981 7.350 1.732 30.435
LUX 1999 2011 30.183 . 33.755 50.920 123.176
LVA 1993 2011 14.806 2.234 5.511 1.388 23.940
MAR 2003 2011 26.348 5.493 21.330 0.510 53.681

28
Table A1 (continued)

Non-
Financial
financial Consumer Mortgage Total
business
Country Start End business
MEX 2000 2011 7.955 3.109 8.690 . 19.754
MKD 2003 2011 11.970 . 7.094 . 19.063
NLD 1990 2011 49.114 8.104 60.016 19.990 137.225
NOR 1990 2011 33.570 . 65.403 . 98.972
NZL 1990 2011 32.034 4.846 55.136 25.089 117.105
POL 1996 2011 11.113 8.923 11.766 1.201 33.004
PRT 1990 2011 42.992 11.402 43.458 5.997 103.849
SGP 1990 2011 63.951 . 25.767 14.404 104.122
SVK 2004 2011 21.342 . 16.886 2.712 40.940
SVN 2004 2011 47.830 . 19.934 . 67.764
SWE 1996 2011 46.764 . 59.248 48.100 154.112
TWN 2003 2011 69.007 19.559 42.992 4.188 135.745
UKR 2005 2011 39.580 . 19.441 5.243 64.263
URY 2005 2011 13.232 . 6.589 . 19.821
USA 1990 2011 9.547 6.032 18.823 . 34.401

Comparison to Similar Data

Beck et al. (2012) and Büyükkarabacak and Valev (2010) were the first to study similar

data, using a data set for 73 countries over the years 1994 to 2005. These papers are

ground breaking in that they are the first studies to look at growth effects of different

credit aggregates across countries. Our data is not an update of this, but is newly

collected. There are two principal reasons. We aimed to separate out mortgage and

other household credit and to observe each credit category at source. The Beck et al.

(2012) data combines mortgage and other household credit into one household credit

category. The data is based on the financial development and structure (FDS) data base

described in Beck, Demirguc-Kunt and Levine (2000) and updated in Beck et al. (2013).

Here "private credit" captures the financial intermediation with the private nonfinancial

29
sector, including mortgages, as explained in note 5 in Beck, Demirguc-Kunt and Levine

(2000) (“claims on real estate (=mortgage credit) is included for nonbanks lending”). The

definition in the data of the variable “private credit by deposit money banks to GDP (%)

is “private credit by deposit money banks to GDP”. Also, the Beck et al. (2012) credit

data are deflated by the cpi deflator and then divided by real (deflated) GDP. Our data

is nominal credit divided by nominal GDP.

In observing the different credit aggregates, Beck et al. (2012) start with a 'total

credit' (TC) measure taken from the FDS data base, which is credit to nonfinancial

business (BC) plus credit to households. Credit to households includes consumer credit

(CC) plus mortgages (MC), but as noted these are not distinguished. The ‘household

credit’ measure in Beck et al. (2010) and in Büyükkarabacak and Valev (2010) is defined

as (TC-BC), i.e. all non-business credit, including both MC and CC.

Table A2 is a comparison of our data to the Beck et al (2010) data. We find that the

data are mostly in agreement, except for a few countries. In the Czech Republic, our

credit/GDP ratio is about half of those in the two other data sets. Personal

communications with the Czech National Bank suggest that part of the reason is

widespread credit write-downs and therefore data revisions since 2005, a large

reduction in the number of banks, and the inclusion of foreign banks. The same applies

to Slovakia, Iceland and Uruqay. For Sweden, our data yield a credit/GDP ratio which

is much higher than in the Beck et al (2012) data, which is about double again than

Büyükkarabacak and Valev (2010). Reclassifications of what counts as a bank may be

behind this. There is also some disparity with Great Britain which we are investigating.

30
Table A2 Comparison to other data sets

Private Credit Non-financial Business Credit Household Credit

Ours BECK2012 BUY2010 Ours BECK2012 BUY2010 Ours BECK2012 BUY2010

1994-2005 1994-2005 1990-2006 1994-2005 1994-2005 1990-2006 1994-2005 1994-2005 1990-2006

ARG 0.166 0.212 0.201 0.117 0.087 0.086 0.049 0.125 0.115
AUS 0.777 0.823 0.806 0.311 0.279 0.285 0.466 0.544 0.52
AUT 0.840 1.005 1.035 0.486 0.653 0.683 0.354 0.352 0.352
BEL 0.651 0.744 0.748 0.315 0.314 0.319 0.336 0.43 0.439
BGR 0.198 0.219 0.245 0.148 0.145 0.157 0.050 0.075 0.088
CAN 0.869 0.962 1.012 0.184 0.188 0.128 0.685 0.773 0.892
CHE 1.369 1.603 1.60 0.445 0.604 0.62 0.924 1 0.98
CZE 0.250 0.484 0.481 0.170 0.314 0.309 0.080 0.171 0.172
DEU 0.970 1.053 1.053 0.558 0.653 0.605 0.412 0.4 0.375
DNK 1.277 0.894 0.338 0.379 0.133 0.095 0.898 0.761 0.247
EGY 0.495 0.446 0.432 0.411 0.372 0.355 0.083 0.075 0.073
EST 0.256 0.286 0.336 0.142 0.176 0.208 0.113 0.111 0.127
FRA 0.613 0.85 0.86 0.344 0.339 0.337 0.269 0.511 0.513
GBR 0.578 1.269 1.337 0.140 0.557 0.293 0.438 0.712 1.04
GRC 0.518 0.663 0.691 0.327 0.379 0.389 0.191 0.283 0.3
HUN 0.212 0.231 0.302 0.143 0.189 0.218 0.069 0.042 0.085
IDN 0.205 0.252 0.249 0.142 0.17 0.169 0.063 0.082 0.08
IND 0.249 0.219 0.227 0.203 0.156 0.159 0.046 0.063 0.068
ISL 0.551 0.918 0.916 0.118 0.492 0.39 0.434 0.426 0.526
JPN 0.903 1.549 1.105 0.596 1.07 0.747 0.307 0.479 0.357
LTU 0.159 0.149 0.177 0.126 0.104 0.13 0.032 0.045 0.064
LVA 0.145 0.199 0.236 0.117 0.16 0.181 0.029 0.039 0.055
MAR 0.409 0.187 . 0.211 0.14 . 0.198 0.046 .
MEX 0.179 0.186 0.194 0.078 0.087 0.122 0.101 0.099 0.072
NLD 1.093 1.639 1.152 0.465 0.63 0.478 0.628 1.01 0.98
NZL 0.861 1.118 1.152 0.301 0.703 0.444 0.560 0.415 0.718
POL 0.254 0.244 0.229 0.106 0.135 0.162 0.147 0.11 0.07
PRT 0.905 1.103 0.961 0.399 0.507 0.124 0.507 0.596 0.51
SVK 0.281 0.415 0.409 0.180 0.265 0.262 0.101 0.15 0.151
SVN 0.434 0.34 0.362 0.309 0.24 0.252 0.125 0.099 0.11
SWE 0.956 0.636 0.374 0.425 0.233 0.228 0.531 1 0.149
URY 0.174 0.392 0.329 0.128 0.194 0.174 0.046 0.198 0.155
USA 0.324 0.498 0.503 0.095 0.118 0.095 0.229 0.38 0.408
Note: BECK2012 and BUY2010 refer to Beck, et. al (2012) and Büyükkarabacak and Valev (2010), respectively. This table shows the
comparison of private credit (excluding financial business credit), non-financial business credit and household credit (the sum of
consumer credit and mortgage credit) for 33 countries (that exist in all three dataset except MAR) between our dataset, Beck, et. al.
(2012) and Büyükkarabacak and Valev (2010). We takes the average of our credit data during the period 1994-2005, which is in line
with Beck, et. al. (2012). As a result, FIN and KOR dropped out due to limited time span in our dataset. 15 countries in our dataset,

31
namely BRA, CHL, ARM, HKG, HRV, ISR, ITA, ESP, MKD, NOR, TWN, UKR, LUX, SGP do not exist in either Beck, et. al. (2012)
nor Büyükkarabacak and Valev (2010).

A more recent and somewhat comparable data set is the March 2013 Bank of

International Settlement ‘Long series on credit to private non-financial sectors’ (BIS,

2013). A description of the data is in Dembiermont et al. (2013), including a link to data

documentation. In the BIS data, only ‘lending by all sectors’ (i.e. bank and securities

markets) is disaggregated to households and enterprises (except for Brazil, Portugal,

Saudi Arabia and Russia). Bank debt is not disaggregated. This implies on one hand

that the BIS data provide a more complete picture of all loans to the private sector,

while on the other hand they do not include lending to the nonbank financial sector

(which is substantial in some countries). Another limitation of the BIS data is that by

including in one credit measure also nonbank lending (which mostly is lending through

securities markets), it is not possible to study the unique role of bank loans. And since

bank debt is not disaggregated, we cannot directly compare the BIS data to our data.

32
References

Ang, J. (2008), “A Survey of Recent Developments in the Literature of Finance and

Growth,” Journal of Economic Surveys, 22: 536-76.

Arellano, M. and O. Bover (1995), “Another Look at the Instrumental Variable

Estimationof Error-Component Models,” Journal of Econometrics 68: 29-52.

Arcand, J-L., E. Berkes and U. Panizza (2012), “Too Much Finance?,” IMF Working Paper

No. 161.

Barajas, A., R. Chami and S. Yousefi (2013), “The Finance and Growth Nexus Re-

Examined: Do All Countries Benefit Equally?,” IMF Working Paper No. 13/130.

Barba, A. and M. Pivetti (2009), “Rising Household Debt: Its Causes and

Macroeconomic Implications-A Long-Period Analysis,” Cambridge Journal of

Economics 33: 113-37.

Beck, T., A. Demirgüç-Kunt and R. Levine (2000), “A New Database on Financial

Development and Structure,” The World Bank Economic Review 14: 597-605.

Beck, T., A. Demirgüç-Kunt and R. Levine (2009), “Financial Institutions and Markets

Across Countries and over Time: Data and Analysis,” World Bank Policy Research

Working Paper No. 4943.

Beck, T., A. Demirgüç-Kunt, R. Levine, M. Čihák and E. Feyen (2013), “Financial

Development and Structure Dataset (updated April 2013),”

http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/0,,conte

ntMDK:20696167~pagePK:64214825~piPK:64214943~theSitePK:469382,00.html

Beck, T., B. Büyükkarabacak, F. Rioja and N. Valev (2012), “Who Gets the Credit? And

Does It Matter? Household vs. Firm Lending across Countries,” The B.E. Journal of

Macroeconomics Contributions 12, article 2.

33
Beck, T., H. Degryse and C. Kneer (2012), “Is More Finance Better? Disentangling

Intermediation and Size Effects of Financial Systems,” European Banking Center

Discussion Paper No. 2012-01.

Beck, T., R. Levine, and A. Levkov (2010), “Big Bad Banks? The Impact of U.S. Branch

Deregulation on Income Distribution,” Journal of Finance 65: 1637-67.

Bena, J. and P. Ondko (2012), “Financial Development and the Allocation of External

Finance,” Journal of Empirical Finance 19: 1-25.

Biggs, M., T Mayer and A Pick (2009), “Credit and Economic Recovery,” DNB Working

Paper No. 218.

Biggs, M., T. Mayer and A. Pick (2010), “Credit and Economic Recovery: Demystifying

Phoenix Miracles”, SSRN Working Paper.

BIS, (2013), “Long Series on Credit to Private Non-Financial Sectors,” BIS Quarterly

Review (March).

Blundell, R. and S. Bond (1998), “Initial Conditions and Moment Restrictions in

Dynamic Panel Data Models,” Journal of Econometrics 87: 115-43.

Boissay, F., F. Collard and F. Smets (2013), “Booms and Systemic Banking Crises,” ECB

Working Paper No. 1514.

Borio, C. (2012), “The Financial Cycle and Macroeconomics: What Have We Learnt?,”

BIS Working Paper No. 395.

Borio, C. and A. Lowe (2004), “Securing Sustainable Price Stability: Should Credit Come

back from the Wilderness?,” BIS Working Paper No. 157.

Braun, M. and B. Larrain (2005), “Finance and the Business Cycle: International, Inter-

Industry Evidence,” The Journal of Finance 60: 1097-128.

Brezigar Masten, A., F. Coricelli and I. Masten (2008), “Non-linear Growth Effects of

Financial Development: Does Financial Integration Matter?,” Journal of International

Money and Finance 27: 295–313.

34
Büyükkarabacak, B. and N. Valev (2010), “The Role of Household and Business Credit

in Banking Crises,” Journal of Banking and Finance 34: 1247–56.

Büyükkarabacak, B. and S. Krause (2009), “Studying the Effects of Household and Firm

Credit on the Trade Balance: The Composition of Funds Matters,” Economic Enquiry

47: 653–66.

Caporale, G.M. and P. Howells (2001), “Money, Credit and Spending: Drawing Causal

Inferences,” Scottish Journal of Political Economy 48: 547-57.

Cecchetti, S. and E. Kharroubi (2012), “Reassessing the Impact of Finance on Growth,”

BIS Working Papers No. 381.

Cecchetti, S., M. Mohanty and F. Zampolli (2011), “The Real Effects of Debt,” BIS

Working Papers No. 352.

Deidda, L. and B. Fattouh (2002), “Non-linearity between Finance and Growth,”

Economics Letters 74: 339-45.

Dembiermont, C., M. Drehmann and S. Muksakunratana (2013), “How much does the

private sector really borrow - a new database for total credit to the private non-

financial sector,” BIS Quarterly Review (March).

Easterly, W., R. Islam and J. Stiglitz (2000), “Shaken and Stirred: Explaining Growth

Volatility,” Paper at the 2000 Annual World Bank Conference on Development

Economics.

Goldsmith, R. (1969), Financial Structure and Development, Yale University Press.

Jappelli, T. and M. Pagano (1994), “Saving, growth and Liquidity Constraints,” The

Quarterly Journal of Economics 109, 83–109.

Jappelli, T., M. Pagano and M. Di Maggio (2008), “Households’ Indebtedness and

Financial Fragility,” Paper presented at the IMF 9th Jacques Polak Annual Research

Conference.

35
Jorda, O., M. Schularick, and A. Taylor (2012), “When Credit Bites Back: Leverage,

Business Cycles, and Crises,” Federal Reserve Bank of San Francisco Working Paper No.

2011-27.

Ketteni, E., T. Mamuneas, A. Savvides and T. Stengos (2007), "Is the Financial

Development and Economic Growth Relationship Nonlinear?," Economics Bulletin

15: 1-12.

King, G.R., and R. Levine (1993), “Finance and Growth: Schumpeter Might be Right,”

The Quarterly Journal of Economics, 108: 717-37.

Krippner, G. (2005), “The Financialization of the American Economy,” Socio-Economic

Review 3: 173-208.

Kroszner, R., L. Laeven and D. Klingebiel (2007), “Banking Crises, Financial

Dependence, and Growth,” Journal of Financial Economics 84: 187-228.

Levine, R. (2005), Finance and Growth: Theory and Evidence, Handbook of Economic

Growth, Volume 1A, Edited by Philippe Aghion and Steven N. Durlauf, Elsevier.

Lorenzoni, G. (2008), “Inefficient Credit Booms,” The Review of Economic Studies 75: 809–

33.

McKinnon, R. (1973), Money and Capital in Economic Development, Brookings

Institution.

Raddatz, C. (2006), “Liquidity Needs and Vulnerability to Financial

Underdevelopment,” Journal of Financial Economics 80: 677–722.

Rajan, R. and L. Zingales (1998), “Financial Dependence and Growth”, The American

Economic Review 88: 559–86.

Reinhart, C. (2010), “This Time is Different Chartbook: Country Histories on Debt,

Default, and Financial Crises,” NBER Working Paper No. 15815.

Rousseau, P. and P. Wachtel (2011), "What is Happening to the Impact of Financial

Deepening on Economic Growth?," Economic Inquiry 49: 276-88.

36
Schularick, M. and A. M. Taylor (2012), “Credit Booms Gone Bust: Monetary Policy,

Leverage Cycles, and Financial Crises, 1870-2008,” American Economic Review, 102:

1029-61.

Schumpeter, J.A. (1934), The Theory of Economic Development, Harvard University

Press.

Shaw, E. (1973), Financial Deepening and Economic Development, Oxford University

Press, New York.

Shen, C-H. and C-C Lee (2006), “Same Financial Development Yet Different Economic

Growth—Why?,” Journal of Money, Credit and Banking, 38: 1907-44.

Stockhammer, E. (2004), “Financialisation and the Slowdown of Accumulation,”

Cambridge Journal of Economics 28: 719-41.

Sutherland, D., P. Hoeller, R. Merola and V. Ziemann (2013), “Debt and Macroeconomic

Stability,” OECD Working Paper No. 1003.

Tobin, J. (1984), “On the Efficiency of the Financial System,” Lloyds Bank Review No. 153.

Turner, A. and others (2010), The Future of Finance: The LSE Report, London School of

Economics and Political Science.

Valickova, P., T. Havranek and R. Horvath (2013), “Financial Development and

Economic Growth: A Meta-Analysis,” IOS Working Paper No. 331.

Werner, R. (1997), “Towards a New Monetary Paradigm: A Quantity Theorem of

Disaggregated Credit, with Evidence from Japan,” Kredit und Kapital 30: 276-309.

Werner, R. A. (2012), “Towards a New Research Programme on ‘Banking and the

Economy’ – Implications of the Quantity Theory of Credit for the Prevention and

Resolution of Banking and Debt Crises,” International Review of Financial Analysis 25:

1-17.

37
 TABLES AND FIGURES

Figure 1 Total bank credit stocks, 1990-2011

95 90
Total bank credit stocks/GDP
75 80 70 85

1990 1995 2000 2005 2010

year
Average over 50 countries, 1990-2011

38
Figure 2 Average bank credit stock composition (top) and ratio to GDP (bottom), 1990-2011

39
Figure 3 Credit stocks, credit flows and economic growth over 1990-2011

(a) credit stocks and growth

6

URY IND
ALB
ARM
LVA BGR
SVK
POL TWN
IDN EST
4
GDP per capita growth

MAR CHL SGP

MKDLTU
CZE KOR HKG
EGY
ARG UKR HRV SWE
AUS
2

BRA LUX
SVN AUT
NOR NLD
GBR ISR
DEU
PRT ESP
USA GRC CAN
HUN
MEX ISLBEL
FRA NZL
JPN CHE
DNK
ITA
0

FIN
-2

0 50 100 150 200 250

Total credit stocks
Period average (1990-2011)

(b) credit flows and growth

15
10
5
0
-5

1990 1995 2000 2005 2010

year

GDP per capita growth total credit flows

40
TABLES

Table 1 Do credit stocks and flows correlate with economic growth?

GDP p.c. Total Nonfinancial Financial Non-financial Consumer Mortgage Fin. Bus.
Panel A - Stocks
growth credit sector sector business credit credit credit

GDP p.c. growth 1

Total credit -0.326*** 1

Nonfinancial sector (a+b) -0.223** 0.824*** 1

Financial sector (c+d) -0.328*** 0.907*** 0.508*** 1

a. Non-financial business -0.200* 0.757*** 0.952*** 0.443*** 1

b. Consumer credit -0.166 0.557*** 0.591*** 0.406*** 0.315*** 1

c. Mortgage credit -0.351*** 0.910*** 0.599*** 0.938*** 0.491*** 0.561*** 1

d. Fin. business credit -0.157 0.548*** 0.132 0.734*** 0.175* -0.0531 0.452*** 1
Panel B - Flows

GDP p.c. growth

1

Total credit
0.287*** 1

Nonfinancial sector (a+b)

0.361*** 0.817*** 1

Financial sector (c+d)

0.137 0.864*** 0.415*** 1

a. Non-financial business
0.330*** 0.801*** 0.971*** 0.416*** 1

b. Consumer
0.294*** 0.493*** 0.635*** 0.224* 0.431*** 1

c. Mortgage
0.111 0.784*** 0.473*** 0.824*** 0.448*** 0.338*** 1

d. Financial business
0.105 0.561*** 0.157 0.748*** 0.187* -0.0128 0.240** 1

Note: The tables report pairwise correlation coefficients between growth and different types of credit flows,* p < 0.05, ** p < 0.01, *** p < 0.001.

41
Table 2 Descriptive Statistics (3-year averaged data)
Source Unit No.of obs. mean sd min max
Credit Stocks
Total credit Own Calculation % of GDP 250 77.490 51.519 2.045 381.584
Real-sector Own Calculation % of GDP 250 40.978 25.889 1.728 187.026
Financial-sector Own Calculation % of GDP 250 36.512 32.568 0.246 194.559
Non-financial business Own Calculation % of GDP 250 33.197 18.964 1.728 92.696
Consumer Own Calculation % of GDP 176 11.052 13.525 0.333 94.330
Mortgage Own Calculation % of GDP 250 29.831 27.753 0.246 194.559
Financial business Own Calculation % of GDP 180 33.197 12.415 0.049 76.323

Credit Flows
Total credit Own Calculation % of lagged GDP 238 7.121 7.520 -4.335 70.304
Real-sector Own Calculation % of lagged GDP 238 3.499 4.249 -5.396 32.055
Financial-sector Own Calculation % of lagged GDP 238 3.622 4.325 -2.931 38.249
Non-financial business Own Calculation % of lagged GDP 238 2.719 3.250 -5.624 16.767
Consumer Own Calculation % of lagged GDP 169 1.099 1.914 -0.808 15.288
Mortgage Own Calculation % of lagged GDP 238 2.926 3.563 -2.823 38.249
Financial business Own Calculation % of lagged GDP 172 0.947 2.360 -2.620 21.979

Other Variables
GDP per capita growth WDI Percentage points 250 2.500 3.209 -19.706 16.995
Initial GDP per capita WDI In log 250 9.379 0.936 6.142 10.611
Trade openness WDI % of GDP 250 93.461 75.570 15.546 424.013
Government size WDI % of GDP 250 17.664 4.830 7.197 28.413
Inflation WDI Percentage points 250 9.175 68.478 -3.123 1070.948
Stock market capitalization Beck (2012) % of GDP 234 67.801 67.820 0.428 480.512

42
Table 3 Credit and Growth: Stock and Flow Effects

FE System GMM
(1) (2) (3) (4) (5) (6)
Total Credit
Credit stocks -0.017* -0.021* -0.048* -0.053*
(0.009) (0.012) (0.027) (0.027)
Credit flows 0.050 0.075 0.136** 0.036
(0.040) (0.047) (0.061) (0.046)

Initial GDPPC -0.241 -0.134 -0.197 -0.143 -2.201** 0.450

(0.358) (0.368) (0.365) (0.822) -1.091 (0.706)

Trade 0.006 0.009 0.005 0.007** 0.005* 0.006*

(0.015) (0.015) (0.015) (0.003) (0.003) (0.004)

Government 0.132 0.114 0.204 -0.040 -0.023 -0.028

(0.242) (0.216) (0.229) (0.073) (0.049) (0.074)

Inflation 0.117 0.022 0.006 -0.023 -0.007 -0.089

(0.098) (0.087) (0.080) (0.097) (0.082) (0.081)
Education 0.406 0.082 0.146 -0.064 0.209 -0.209
(0.540) (0.539) (0.529) (0.217) (0.162) (0.209)
Institution 0.422* 0.454* 0.421* 0.128 0.097 0.116
(0.231) (0.245) (0.237) (0.135) (0.109) (0.148)

Time FE Yes Yes Yes Yes Yes Yes

Observations 236 224 224 236 224 224
No. of id 46 46 46 46 46 46

R-squared 0.351 0.372 0.392

AR(2) 0.129 0.177 0.163

Overidentification 0.113 0.492 0.19

Note: This table presents the results based on equation (1) and (2). Columns (1)-(3) present the results using the FE estimation
technique, whereas columns (4)-(6) show results using the system GMM estimation procedure. The dependent variable is the
average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as the
credit-to-GDP ratio; Flow of credit captures the change in credit as explained in the text; Initial GDPPC is the real GDP per capita at
the beginning of each 3-year period; Trade measures trade openness, defined as total imports plus exports divided by GDP;
Government is government consumption as a share of GDP; Inflation is the change in CPI; Education is the average year of
schooling; Institution is the composite country risk measure, taken from international country risk guide (ICRG). AR(2) is the
Arellano-Bond serial correlation test (p-value is reported); Over-identification is the Hansen J statistic (p-value is reported). All
specifications include time dummies (coefficients not reported). Constant is not reported.
Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

43
Table 4 Differentiating between categories of credit

FE System GMM
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Real-sector Credit Financial-sector Credit Real-sector Credit Financial-sector Credit
Credit stocks -0.042** -0.056** -0.019 -0.019 -0.068 -0.097** -0.069** -0.074**
(0.020) (0.025) (0.013) (0.015) (0.054) (0.044) (0.031) (0.032)
Credit flows 0.157* 0.201** 0.017 0.043 0.349*** 0.178* -0.013 0.008
(0.085) (0.086) (0.042) (0.050) (0.104) (0.096) (0.069) (0.047)

Initial GDPPC -0.196 -0.185 -0.208 -0.253 -0.085 -0.131 -0.466 -2.062* 0.416 -0.399 -2.077** -0.233
(0.355) (0.388) (0.387) (0.366) (0.363) (0.369) -1.007 -1.146 (0.892) (0.863) -1.017 (0.879)

Trade 0.004 0.007 0.001 0.008 0.009 0.008 0.009** 0.004 0.008* 0.005* 0.006*** 0.005
(0.014) (0.015) (0.014) (0.016) (0.015) (0.015) (0.004) (0.003) (0.004) (0.003) (0.002) (0.003)

Government 0.091 0.155 0.189 0.139 0.057 0.124 -0.078 0.000 -0.064 -0.013 -0.034 0.003
(0.238) (0.225) (0.225) (0.250) (0.209) (0.233) (0.068) (0.050) (0.074) (0.078) (0.046) (0.074)

Inflation 0.100 0.003 -0.035 0.127 0.037 0.037 -0.026 -0.052 -0.139* 0.028 0.027 -0.027
(0.096) (0.085) (0.074) (0.102) (0.088) (0.086) (0.107) (0.077) (0.082) (0.093) (0.098) (0.085)
Education 0.524 0.071 0.303 0.310 0.091 0.070 -0.027 0.228 -0.202 0.017 0.170 -0.085
(0.552) (0.540) (0.493) (0.515) (0.548) (0.550) (0.235) (0.177) (0.187) (0.203) (0.157) (0.194)
Institution 0.423* 0.434* 0.401* 0.427* 0.472* 0.454* 0.044 0.104 0.012 0.168 0.115 0.185
(0.233) (0.237) (0.231) (0.231) (0.248) (0.242) (0.111) (0.109) (0.130) (0.130) (0.095) (0.133)

Time FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 236 224 224 236 224 224 236 224 224 236 224 224
No. of id 46 46 46 46 46 46 46 46 46 46 46 46

R-squared 0.357 0.386 0.417 0.344 0.365 0.371

AR(2) 0.123 0.260 0.200 0.117 0.137 0.142

Overidentification 0.320 0.451 0.308 0.359 0.368 0.275

Note: This table presents the results based on equation (1) and (2). Columns (1)-(6) present the results using the FE estimation technique, where columns (1)-(3) and (4)-(6) examines the growth effect of
real-sector credit and financial-sector credit, respectively. Columns (7)-(12) show results using the system GMM estimation procedure, similarly distinguishing between real-sector credit in columns (7)-(9)
and financial-sector credit in columns (10-(12). The dependent variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as
the credit-to-GDP ratio; Flow of credit captures the change in credit as explained in the text; Initial GDPPC is the real GDP per capita at the beginning of each 3-year period; Trade measures trade
openness, defined as total imports plus exports divided by GDP; Government is government consumption as a share of GDP; Inflation is the change in CPI; Education is the average year of schooling;
Institution is the composite country risk measure, taken from international country risk guide (ICRG).AR(2) is the Arellano-Bond serial correlation test (p-value is reported); Over-identification is the
Hansen J statistic (p-value is reported). All specifications include time dummies (coefficients not reported). Constant is not reported. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

44
Table 5 Credit stocks and flows effect on growth: industry-level evidence

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Total Credit Nonfinancial-sector Credit Financial-sector Credit
Stock of credit -0.046*** -0.059*** -0.091*** -0.149*** -0.072*** -0.070***
(0.012) (0.013) (0.027) (0.029) (0.022) (0.024)
ED*stock of credit 0.050** 0.049** 0.094* 0.113** 0.084** 0.072
(0.022) (0.024) (0.048) (0.051) (0.041) (0.046)
Flow of credit 0.091* 0.162*** 0.346*** 0.490*** 0.009 0.097
(0.052) (0.057) (0.124) (0.131) (0.073) (0.082)
ED*flow of credit 0.132 0.074 0.158 0.052 0.212 0.119
(0.110) (0.121) (0.246) (0.260) (0.142) (0.158)
Initial share 0.469** 0.302 0.296 0.468** 0.291 0.280 0.470** 0.305 0.302
(0.207) (0.192) (0.192) (0.207) (0.192) (0.190) (0.208) (0.193) (0.193)
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Time FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry-country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry-time FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 5,774 5,424 5,424 5,774 5,424 5,424 5,774 5,424 5,424
Number of countries 45 45 45 45 45 45 45 45 45
R-squared 0.424 0.453 0.455 0.424 0.455 0.459 0.424 0.451 0.453

Marginal effect
for high dependence
-0.023 -0.048 -0.033
industry
Marginal effect
for low dependence
-0.042 -0.083 -0.065
industry
Implied differential effect 0.019 0.036 0.032
Note: This table presents the industry-level evidence based on equation (3). Columns (1)-(3) presents the results for total credit, columns (4)-(6) presents them for real-sector credit,whereas columns (7)-(9)
present the results for financial-sector credit. The dependent variable is the average growth rate of real value added over each 3-year period. Stock of credit is measured as the credit-to-GDP ratio, whereas
flow of credit is the change in credit as explained in the text. ED is external finance dependence, taken from Rajan and Zingales (1998). Initial share is the share of each industry in a country’s total
manufacturing value added at the beginning of each 3-year period. The set of dummies includes country, year, industry, industry-year and industry-country fix effects (coefficients not reported). Constant
is not reported. The least three rows show the marginal growth effect of credit (stock and/or flow) for an industry in the 75th percentile and an industry in the 25th percentile in the external finance
dependence index. The difference between these two is the implied differential effect. All standard errors in parentheses are adjusted for industry-country level heteroskedasticity and autocorrelation, ***
p<0.01, ** p<0.05, * p<0.1

45
Table 6 The Four Categories of Credit and Their Relation to Growth
FE System GMM
(1) (2) (3) (4) (5) (6) (7) (8)
NonFin Consumer Mortgage FinBus NonFin Consumer Mortgage FinBus
Credit stocks -0.093*** -0.066 -0.018 -0.003 -0.139*** -0.152 -0.080** -0.070
(0.026) (0.050) (0.018) (0.027) (0.050) (0.118) (0.039) (0.058)
Credit flows 0.256*** 0.243 0.033 0.109** 0.317*** 0.104 0.008 -0.083
(0.068) (0.329) (0.065) (0.047) (0.080) (0.144) (0.049) (0.180)

Initial GDPPC -0.139 -0.339 -0.106 0.382 0.687 0.359 -0.171 -1.054
(0.376) (0.745) (0.367) (0.565) (0.858) -1.245 (0.864) -1.425

Trade 0.000 -0.009 0.008 0.013*** 0.011* 0.002 0.001 0.011***

(0.014) (0.027) (0.015) (0.005) (0.005) (0.006) (0.004) (0.003)

Government 0.157 0.126 0.108 -0.163 -0.058 -0.100 -0.023 -0.030

(0.217) (0.242) (0.236) (0.146) (0.069) (0.108) (0.062) (0.068)

Inflation -0.033 0.070 0.038 -0.044 -0.182*** 0.054 -0.012 -0.065

(0.071) (0.099) (0.085) (0.063) (0.067) (0.106) (0.080) (0.095)
Education 0.161 0.669 0.116 0.057 -0.280 -0.297 -0.073 0.165
(0.484) (0.666) (0.547) (0.547) (0.200) (0.241) (0.175) (0.247)
Institution 0.403* 0.667* 0.457* 0.293** 0.009 0.074 0.172 -0.010
(0.234) (0.383) (0.242) (0.111) (0.128) (0.192) (0.136) (0.108)

Time FE Yes Yes Yes Yes Yes Yes Yes Yes

Observations 224 159 224 163 224 159 224 163
No. of id 46 31 46 34 46 31 46 34
R-squared 0.428 0.424 0.369 0.568
ar2 0.210 0.219 0.165 0.038
Overidentification 0.374 0.968 0.365 0.243
Note: This table presents the results for four disaggregated categories of credit, namely non-financial business credit (NonFin), consumer credit (Consumer), mortgage credit (Mortgage) and financial
business credit (Finbus). Columns (1)-(4) apply the FE estimation technique, whereas columns (5)-(8) employ the system GMM estimation procedure. The dependent variable is the average growth rate of
real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as the (different types of) credit-to-GDP ratio; Flow of credit captures the change in (different types of)
credit as explained in the text; Initial GDPPC is the real GDP per capita at the beginning of each 3-year period; Trade measures trade openness, defined as total imports plus exports divided by GDP;
Government is government consumption as a share of GDP; Inflation is the change in CPI.; Education is the average year of schooling; Institution is the composite country risk measure, taken from
international country risk guide (ICRG). AR(2) is the Arellano-Bond serial correlation test (p-value is reported); Over-identification is the Hansen J statistic (p-value is reported). All specifications include
time dummies (coefficients not reported). Constant is not reported. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

46
Table 7 The credit-growth relation across development levels
FE System GMM
(1) (2) (3) (4) (5) (6) (7) (8)
VARIABLES Real_HI Real_LI Fin_HI Fin_LI Real_HI Real_LI Fin_HI Fin_LI
Credit stocks -0.048* -0.012 -0.037 -0.002 -0.087 -0.071* -0.092** -0.060**
(0.026) (0.016) (0.027) (0.015) (0.052) (0.040) (0.042) (0.028)
Credit flows 0.182* 0.012 0.180* -0.021 0.180* -0.063 0.201 0.066
(0.102) (0.048) (0.106) (0.072) (0.093) (0.074) (0.132) (0.115)

Initial GDPPC -0.301 0.045 -0.491 -0.423 -0.527 0.246 0.288 1.344
(0.428) (0.373) (0.506) (0.466) (0.611) (0.831) (0.865) (0.943)

Trade -0.007 -0.001 0.020 0.031 0.008* 0.006 0.003 0.000

(0.022) (0.022) (0.024) (0.022) (0.004) (0.004) (0.004) (0.005)

Government 0.217 0.114 0.256 0.122 -0.030 0.065 -0.091 -0.124

(0.446) (0.468) (0.265) (0.282) (0.076) (0.090) (0.100) (0.083)

Inflation 0.007 0.121 0.045 0.104 -0.089 0.141 -0.079 -0.006

(0.123) (0.133) (0.089) (0.093) (0.167) (0.092) (0.076) (0.070)
Education -0.109 -0.431 0.304 0.147 -0.102 -0.089 -0.141 -0.107
(0.517) (0.547) (0.579) (0.620) (0.235) (0.238) (0.220) (0.234)
Institution 0.465 0.503 0.595* 0.704** 0.069 0.222 0.115 0.066
(0.357) (0.366) (0.330) (0.335) (0.119) (0.160) (0.163) (0.148)

Time FE Yes Yes Yes Yes Yes Yes Yes Yes

Observations 179 179 165 165 179 179 165 165
R-squared 0.375 0.339 0.454 0.429
No. of id 36 36 35 35 36 36 35 35

AR(2) 0.217 0.170 0.197 0.180

Overidentification 0.714 0.895 0.701 0.529

Note: This table presents the results of the growth effect of real-sector (Real) and financial-sector (Fin) credit in high-income(HI) and low-income (LI) countries. Columns (1)-(4) apply the FE estimation
technique, whereas columns (5)-(8) employ the system GMM estimation procedure. The dependent variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year
period. Stock of credit is measured as the credit-to-GDP ratio; Flow of credit captures the change in (different types of) credit as explained in the text; Initial GDPPC is the real GDP per capita at the
beginning of each 3-year period; Trade measures trade openness, defined as total imports plus exports divided by GDP; Government is government consumption as a share of GDP; Inflation is the change
in CPI; Education is the average year of schooling; Institution is the composite country risk measure, taken from international country risk guide (ICRG). AR(2) is the Arellano-Bond serial correlation test
(p-value is reported); Over-identification is the Hansen J statistic (p-value is reported). All specifications include time dummies (coefficients not reported). Constant is not reported.
Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

47
Table 8 Result excluding the crisis years 2008-2011

FE GMM
(1) (2) (3) (4)
No-Crisis
Real Fin Real Fin
Credit stocks -0.089*** 0.000 -0.140*** -0.072**
(0.024) (0.024) (0.041) (0.032)
Credit flows 0.287*** -0.003 0.302*** -0.069
(0.078) (0.082) (0.079) (0.249)

Initial GDPPC 0.080 0.087 1.348 0.645

(0.465) (0.448) (0.858) (0.867)

Trade 0.011 0.024 0.012** 0.008*

(0.018) (0.017) (0.005) (0.004)

Government 0.027 -0.213 -0.037 0.023

(0.255) (0.255) (0.093) (0.079)

Inflation -0.119* -0.046 -0.222** -0.058

(0.060) (0.070) (0.090) (0.093)
Education -0.103 -0.076 -0.287 -0.081
(0.565) (0.567) (0.241) (0.230)
Institution 0.348 0.377 -0.022 0.105
(0.302) (0.318) (0.158) (0.126)

Time FE Yes Yes Yes Yes

Observations 178 178 178 178
No. of id 44 44 44 44
R-squared 0.322 0.255

AR(2) 0.228 0.238

Overidentification 0.153 0.538

Note: Columns (1) and (2) apply the FE estimation technique; columns (3)-(4) employ the system GMM estimation procedure. The dependent variable is
the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as the credit-to-GDP ratio;
Flow of credit captures the change in credit as explained in the text; Initial GDPPC is the real GDP per capita at the beginning of each 3-year period;
Trade measures trade openness, defined as total imports plus exports divided by GDP; Government is government consumption as a share of GDP;
Inflation is the change in CPI; Education is the average year of schooling; Institution is the composite country risk measure, taken from international
country risk guide (ICRG). AR(2) is the Arellano-Bond serial correlation test (p-value is reported); Over-identification is the Hansen J statistic (p-value is
reported). All specifications include time dummies (coefficients not reported). Constant is not reported. Robust standard errors in parentheses, ***
p<0.01, ** p<0.05, * p<0.1.

48
Table 9 Results including stock market capitalization
FE GMM
(1) (2) (3) (4)
Real Fin Real Fin
Credit stocks -0.039 -0.017 -0.078* -0.072**
(0.026) (0.018) (0.045) (0.031)
Credit flows 0.164* 0.027 0.207* 0.011
(0.088) (0.055) (0.118) (0.053)

Initial GDPPC 0.039 0.086 -0.056 0.003

(0.362) (0.335) (0.966) -1.131

Trade 0.014 0.022 0.009* 0.005

(0.015) (0.013) (0.005) (0.004)

Government 0.115 0.054 -0.109 0.039

(0.166) (0.179) (0.106) (0.132)

Inflation -0.035 0.011 -0.086 -0.071

(0.076) (0.069) (0.118) (0.106)
Education -0.242 -0.517 0.113 -0.136
(0.705) (0.698) (0.865) (0.758)
Institution 0.493 0.551* 0.093 0.136
(0.298) (0.291) (0.117) (0.129)
Stock market -0.017 -0.021 -0.008 0.001

capitalization (0.021) (0.021) (0.006) (0.008)

Time FE Yes Yes Yes Yes

Observations 214 214 214 214
No. of id 45 45 45 45
R-squared 0.450 0.427

AR(2) 0.236 0.186

Overidentification 0.245 0.214

Note: Stock market capitalization (as a share of GDP) is the additional control variable. Columns (1)-(2) apply the FE estimation technique; columns (3)-
(4) employ the system GMM estimation procedure. The dependent variable is the average growth rate of real GDP per capita (constant 2005 US dollar)
over each 3-year period. Stock of credit is measured as the credit-to-GDP ratio; Flow of credit captures the change in credit as explained in the text;
Initial GDPPC is the real GDP per capita at the beginning of each 3-year period; Trade measures trade openness, defined as total imports plus exports
divided by GDP; Government is government consumption as a share of GDP; Inflation is the change in CPI; Education is the average year of schooling;
Institution is the composite country risk measure, taken from international country risk guide (ICRG). AR(2) is the Arellano-Bond serial correlation test
(p-value is reported); Over-identification is the Hansen J statistic (p-value is reported). All specifications include time dummies (coefficients not
reported). Constant is not reported. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

49
Table 10 Does the effect of credit flows depend on credit stocks?

FE GMM
(1) (2) (3) (4) (5) (6)
Total Real Fin Total Real Fin

Credit stocks -0.014 -0.040 -0.012 -0.025 -0.059* -0.046*

(0.012) (0.024) (0.016) (0.019) (0.034) (0.024)

Credit flows 0.202*** 0.458*** 0.162* 0.254*** 0.478*** 0.272**

(0.055) (0.125) (0.093) (0.049) (0.107) (0.123)

Stocks * flows -0.001*** -0.003*** -0.001* -0.001*** -0.003*** -0.002**

(0.000) (0.001) (0.001) (0.000) (0.001) (0.001)

Initial GDPPC -0.143 -0.117 -0.110 -1.126 0.096 -0.766

(0.378) (0.401) (0.381) -1.022 (0.830) (0.719)

Trade 0.006 0.003 0.009 0.005* 0.005* 0.004

(0.014) (0.014) (0.015) (0.003) (0.003) (0.003)

Government 0.290 0.260 0.159 -0.009 -0.050 -0.012

(0.235) (0.241) (0.229) (0.047) (0.048) (0.054)

Inflation -0.008 -0.050 0.038 -0.078 -0.135** -0.006

(0.077) (0.074) (0.085) (0.065) (0.060) (0.076)
Education 0.228 0.369 0.079 0.037 -0.088 -0.011
(0.516) (0.487) (0.547) (0.137) (0.137) (0.135)
Institution 0.387* 0.358 0.445* 0.113 0.009 0.138
(0.230) (0.215) (0.243) (0.133) (0.107) (0.115)

Time FE Yes Yes Yes Yes Yes Yes

Observations 224 224 224 224 224 224
No. of id 46 46 46 46 46 46
R-squared 0.414 0.445 0.377

AR(2) 0.284 0.263 0.183

Overidentification 0.875 0.923 0.832

Note: Columns (1)-(3) apply the FE estimation technique, whereas columns (4)-(6) employ the system GMM estimation procedure. The dependent
variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as the credit-to-
GDP ratio; Flow of credit captures the change in credit as explained in the text; Initial GDPPC is the real GDP per capita at the beginning of each 3-year
period; Trade measures trade openness, defined as total imports plus exports divided by GDP; Government is government consumption as a share of
GDP; Inflation is the change in CPI; Education is the average year of schooling; Institution is the composite country risk measure, taken from
international country risk guide (ICRG).AR(2) is the Arellano-Bond serial correlation test (p-value is reported); Over-identification is theHansen J
statistic (p-value is reported). All specifications include time dummies (coefficients not reported). Constant is not reported. Robust standard errors in
parentheses, *** p<0.01, ** p<0.05, * p<0.1.

50
Table 11 Accounting for nonlinearities

FE System GMM
(1) (2) (3) (4) (5) (6)
Total Real Fin Total Real Fin

Credit stocks -0.079*** -0.101* -0.100** -0.116*** -0.187** -0.181***

(0.022) (0.054) (0.042) (0.020) (0.080) (0.043)
Credit stocks 0.000* 0.000 0.000* 0.000 0.001 0.001**
squared (0.000) (0.001) (0.000) (0.000) (0.001) (0.000)
Credit flows 0.159*** 0.301*** 0.132** 0.126* 0.339*** 0.065
(0.035) (0.065) (0.059) (0.067) (0.079) (0.079)
Initial GDPPC -0.038 -0.062 -0.057 -1.053 -0.656 -0.204
(0.366) (0.382) (0.392) (1.337) (1.563) (0.850)
Trade 0.001 -0.002 0.007 0.009 0.009* 0.006
(0.014) (0.014) (0.015) (0.006) (0.005) (0.005)
Government 0.198 0.157 0.130 0.019 -0.019 0.040
(0.241) (0.232) (0.233) (0.103) (0.083) (0.097)
Inflation -0.039 -0.069 0.019 -0.148 -0.183** -0.070
(0.069) (0.064) (0.085) (0.094) (0.081) (0.095)
Education -0.206 0.008 -0.066 -0.076 -0.118 -0.090
(0.558) (0.493) (0.622) (0.206) (0.197) (0.245)
Institution 0.390* 0.392* 0.434* 0.310 0.126 0.324**
(0.232) (0.225) (0.240) (0.207) (0.155) (0.142)

Time FE Yes Yes Yes Yes Yes Yes

Observations 218 218 218 218 218 218
Number of id 45 45 45 45 45 45
R-squared 0.427 0.450 0.384
AR(2) 0.132 0.207 0.109
Overidentification 0.930 0.838 0.797
Note: Columns (1)-(3) apply the FE estimation technique, whereas columns (4)-(6) employ the system GMM estimation procedure. The dependent
variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as the credit-to-
GDP ratio; Initial GDPPC is the real GDP per capita at the beginning of each 3-year period; Trade measures trade openness, defined as total imports plus
exports divided by GDP; Government is government consumption as a share of GDP; Inflation is the change in CPI; Education is the average year of
schooling; Institution is the composite country risk measure, taken from international country risk guide (ICRG). AR(2) is the Arellano-Bond serial
correlation test (p-value is reported); Over-identification is the Hansen J statistic (p-value is reported). All specifications include time dummies
(coefficients not reported). Constant is not reported. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

51
Table 12 Does the effect of credit stock depend on the occurrence of banking crisis? (All credit)

FE System GMM
(1) (2) (3) (4) (5) (6)
Total Credit

Credit stocks -0.029** -0.050*** -0.022 -0.044***

(0.012) (0.013) (0.018) (0.009)
Credit flows 0.057* 0.149*** 0.143 0.187***
(0.031) (0.032) (0.097) (0.042)
Crisis -4.210** -3.303 -4.432** -3.251 -1.466 -4.394***
(2.033) (1.968) (1.980) (2.396) (1.573) (1.497)
Crisis*Credit stocks 0.030** 0.021* 0.044*** 0.021 0.002 0.039***
(0.014) (0.011) (0.015) (0.021) (0.007) (0.010)
Initial GDPPC 0.105 0.102 0.047 -3.276 -2.125* 0.246
(0.391) (0.357) (0.421) (6.535) (1.227) (0.837)
Trade 0.006 0.012 0.003 0.008* 0.005* 0.005*
(0.012) (0.012) (0.011) (0.005) (0.003) (0.003)
Government 0.162 0.157 0.289 0.010 -0.006 -0.030
(0.255) (0.248) (0.260) (0.136) (0.056) (0.044)
Inflation 0.155* 0.066 0.041 0.044 0.009 -0.035
(0.082) (0.086) (0.069) (0.115) (0.079) (0.053)
Education 0.012 -0.246 -0.324 0.326 0.208 -0.055
(0.494) (0.663) (0.551) (0.621) (0.163) (0.140)
Institution 0.334** 0.375** 0.312** 0.245 0.085 0.042
(0.161) (0.176) (0.154) (0.458) (0.101) (0.069)

Time FE Yes Yes Yes Yes Yes Yes

Observations 236 224 224 236 224 224
Number of id 46 46 46 46 46 46
R-squared 0.427 0.421 0.487
AR(2) 0.160 0.122 0.156
Overidentification 0.557 0.545 0.304
Note: Columns (1)-(3) apply the FE estimation technique, whereas columns (4)-(6) employ the system GMM estimation procedure. The dependent
variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as the credit-to-
GDP ratio; Crisis is a dummy variable indicating a country is in a banking crisis for at least 1 year in a 3-year period. Initial GDPPC is the real GDP per
capita at the beginning of each 3-year period; Trade measures trade openness, defined as total imports plus exports divided by GDP; Government is
government consumption as a share of GDP; Inflation is the change in CPI; Education is the average year of schooling; Institution is the composite
country risk measure, taken from international country risk guide (ICRG). AR(2) is the Arellano-Bond serial correlation test (p-value is reported); Over-
identification is the Hansen J statistic (p-value is reported). All specifications include time dummies (coefficients not reported). Constant is not reported.
Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

52
Table 13 Does the effect of credit stock depend on the occurrence of banking crisis? (Real vs. Financial credit)

FE System GMM
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
VARIABLES Real Real Real Fin Fin Fin Real Real Real Fin Fin Fin

Credit stocks -0.064*** -0.101*** -0.037** -0.053** -0.073 -0.051 -0.022 -0.052**
(0.024) (0.024) (0.015) (0.020) (0.052) (0.039) (0.022) (0.023)
Credit flows 0.160** 0.284*** 0.027 0.128** 0.345** 0.275** -0.113 0.171*
(0.068) (0.061) (0.035) (0.050) (0.169) (0.104) (0.112) (0.093)
Crisis -3.911* -2.865 -4.008* -3.799** -3.140* -3.675** -4.046 -0.633 -2.813** -2.679** -2.435* -3.671***
(2.127) (1.973) (2.039) (1.766) (1.760) (1.769) (2.704) (1.630) (1.103) (1.174) (1.369) (1.313)
Crisis*Credit stocks 0.057* 0.037 0.081** 0.046** 0.032* 0.055** 0.064 -0.006 0.043 0.025 0.010 0.050***
(0.031) (0.024) (0.030) (0.019) (0.016) (0.021) (0.059) (0.016) (0.029) (0.021) (0.014) (0.016)
Initial GDPPC 0.087 -0.005 0.013 0.077 0.159 0.104 -4.227 -2.139** -2.192 -3.112 -3.326 -0.526
(0.360) (0.356) (0.407) (0.383) (0.364) (0.408) (4.296) (1.044) (1.897) (5.092) (3.763) (1.048)
Trade 0.007 0.011 0.002 0.007 0.011 0.007 0.011* 0.004 0.007** 0.007** 0.008** 0.004*
(0.011) (0.013) (0.012) (0.012) (0.012) (0.012) (0.006) (0.003) (0.003) (0.003) (0.003) (0.002)
Government 0.134 0.208 0.220 0.182 0.104 0.226 0.012 0.014 0.006 0.008 0.017 -0.020
(0.261) (0.261) (0.257) (0.263) (0.245) (0.266) (0.138) (0.056) (0.072) (0.112) (0.087) (0.048)
Inflation 0.127 0.040 -0.019 0.168* 0.081 0.092 -0.008 -0.045 -0.060 0.067 0.062 0.045
(0.085) (0.088) (0.069) (0.084) (0.082) (0.077) (0.132) (0.083) (0.086) (0.103) (0.113) (0.064)
Education 0.020 -0.346 -0.299 0.084 -0.083 -0.174 0.341 0.234 0.171 0.344 0.313 0.049
(0.559) (0.736) (0.585) (0.443) (0.573) (0.519) (0.440) (0.174) (0.216) (0.456) (0.369) (0.129)
Institution 0.344** 0.368** 0.321** 0.333** 0.384** 0.348** 0.295 0.106 0.138 0.218 0.198 0.091
(0.168) (0.177) (0.159) (0.163) (0.177) (0.164) (0.325) (0.095) (0.141) (0.376) (0.269) (0.084)

Time FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 236 224 224 236 224 224 236 224 224 236 224 224
Number of id 46 46 46 46 46 46 46 46 46 46 46 46
R-squared 0.421 0.422 0.494 0.417 0.415 0.441
AR(2) 0.378 0.230 0.489 0.0547 0.186 0.0775
Overidentification 0.862 0.508 0.243 0.662 0.474 0.305
Note: Columns (1)-(6) present the results using the FE estimation technique, where columns (1)-(3) and (4)-(6) examines the growth effect of real-sector credit and financial-sector credit, respectively.
Columns (7)-(12) show results using the system GMM estimation procedure, similarly distinguishing between real-sector credit in columns (7)-(9) and financial-sector credit in columns (10-(12). The

53
dependent variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is measured as the credit-to-GDP ratio; Flow of credit captures the
change in credit as explained in the text; Crisis is a dummy variable indicating a country is in a banking crisis for at least 1 year in a 3-year period; Initial GDPPC is the real GDP per capita at the
beginning of each 3-year period; Trade measures trade openness, defined as total imports plus exports divided by GDP; Government is government consumption as a share of GDP; Inflation is the change
in CPI; Education is the average year of schooling; Institution is the composite country risk measure, taken from international country risk guide (ICRG).AR(2) is the Arellano-Bond serial correlation test
(p-value is reported); Over-identification is the Hansen J statistic (p-value is reported). All specifications include time dummies (coefficients not reported). Constant is not reported. Robust standard errors
in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

54
Table 14 Does the effect of credit stock depend on country outliers? (All credit)
FE System GMM
(1) (2) (3) (4) (5) (6)
Total Credit

Credit stocks -0.036*** -0.043*** -0.077** -0.080**

(0.013) (0.015) (0.033) (0.036)
Credit flows 0.116*** 0.146*** 0.303*** 0.136**
(0.037) (0.037) (0.061) (0.060)
Initial GDPPC -0.221 0.080 -0.012 0.283 -1.186 0.815
(0.477) (0.482) (0.462) (1.198) (1.495) (0.812)
Trade 0.005 0.007 0.003 0.010** 0.000 0.008
(0.015) (0.014) (0.015) (0.005) (0.003) (0.005)
Government 0.117 0.169 0.242 -0.048 -0.035 -0.018
(0.295) (0.266) (0.283) (0.116) (0.052) (0.105)
Inflation 0.101 -0.008 -0.029 -0.053 -0.067 -0.147*
(0.097) (0.092) (0.075) (0.110) (0.053) (0.075)
Education -0.017 0.367 -0.200 -0.096 0.086 -0.238
(0.581) (0.619) (0.626) (0.294) (0.167) (0.231)
Institution 0.410 0.430 0.398 0.181 0.033 0.158
(0.253) (0.271) (0.257) (0.150) (0.144) (0.195)
Time FE Yes Yes Yes Yes Yes Yes
Observations 211 200 200 211 200 200
No. of id 41 41 41 41 41 41
R-squared 0.337 0.362 0.399
AR(2) 0.137 0.375 0.201
Overidentification 0.324 0.443 0.451
Note: Columns (1)-(3) apply the FE estimation technique, whereas columns (4)-(6) employ the system GMM estimation procedure.
The dependent variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock
of credit is measured as the credit-to-GDP ratio; Crisis is a dummy variable indicating a country is in a banking crisis for at least 1
year in a 3-year period. Initial GDPPC is the real GDP per capita at the beginning of each 3-year period; Trade measures trade
openness, defined as total imports plus exports divided by GDP; Government is government consumption as a share of GDP;
Inflation is the change in CPI; Education is the average year of schooling; Institution is the composite country risk measure, taken
from international country risk guide (ICRG). AR(2) is the Arellano-Bond serial correlation test (p-value is reported); Over-
identification is the Hansen J statistic (p-value is reported). All specifications include time dummies (coefficients not reported).
Constant is not reported. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

55
Table 15 Does the effect of credit stock depend on country outliers? (Real vs. Financial credit)
FE System GMM
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Real Real Real Fin Fin Fin Real Real Real Fin Fin Fin

Credit stocks -0.081*** -0.099*** -0.032 -0.035 -0.094*** -0.134*** -0.091* -0.098*
(0.021) (0.026) (0.021) (0.023) (0.028) (0.038) (0.047) (0.053)
Credit flows 0.261*** 0.291*** 0.065 0.105* 0.492*** 0.316*** -0.134 -0.240
(0.077) (0.070) (0.057) (0.053) (0.108) (0.078) (0.145) (0.217)
Initial GDPPC -0.108 0.057 0.067 -0.248 0.069 -0.006 0.160 -0.272 0.946 0.171 -2.735* -1.405
(0.473) (0.482) (0.454) (0.499) (0.484) (0.493) (1.052) (1.130) (1.276) (1.019) (1.382) (1.493)
Trade 0.002 0.004 -0.003 0.008 0.008 0.008 0.009* 0.000 0.008 0.007* 0.006** 0.009
(0.013) (0.016) (0.014) (0.016) (0.014) (0.015) (0.004) (0.003) (0.006) (0.004) (0.003) (0.006)
Government 0.022 0.186 0.123 0.143 0.067 0.150 -0.155 -0.016 -0.121 0.007 -0.038 0.074
(0.291) (0.278) (0.277) (0.306) (0.257) (0.287) (0.101) (0.062) (0.099) (0.116) (0.063) (0.140)
Inflation 0.071 -0.028 -0.083 0.122 0.021 0.025 -0.038 -0.101 -0.191** 0.018 0.015 -0.037
(0.094) (0.089) (0.063) (0.106) (0.093) (0.090) (0.098) (0.062) (0.077) (0.092) (0.107) (0.124)
Education 0.193 0.385 0.047 0.160 0.279 -0.060 -0.065 0.024 -0.234 -0.022 0.219 0.038
(0.587) (0.591) (0.534) (0.639) (0.664) (0.701) (0.210) (0.143) (0.209) (0.260) (0.177) (0.257)
Institution 0.417 0.409 0.392 0.417 0.461 0.443 0.048 0.014 0.034 0.162 0.199 0.335
(0.258) (0.259) (0.252) (0.255) (0.278) (0.267) (0.133) (0.124) (0.178) (0.136) (0.147) (0.242)
Time FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 211 200 200 211 200 200 211 200 200 211 200 200
No. of id 41 41 41 41 41 41 41 41 41 41 41 41
R-squared 0.346 0.377 0.428 0.319 0.343 0.353
AR(2) 0.182 0.302 0.249 0.128 0.131 0.101
Overidentification 0.491 0.395 0.381 0.461 0.526 0.673
Note: Columns (1)-(6) present the results using the FE estimation technique, where columns (1)-(3) and (4)-(6) examines the growth effect of real-sector credit and financial-sector
credit, respectively. Columns (7)-(12) show results using the system GMM estimation procedure, similarly distinguishing between real-sector credit in columns (7)-(9) and financial-
sector credit in columns (10-(12). The dependent variable is the average growth rate of real GDP per capita (constant 2005 US dollar) over each 3-year period. Stock of credit is
measured as the credit-to-GDP ratio; Flow of credit captures the change in credit as explained in the text; Crisis is a dummy variable indicating a country is in a banking crisis for at
least 1 year in a 3-year period; Initial GDPPC is the real GDP per capita at the beginning of each 3-year period; Trade measures trade openness, defined as total imports plus exports
divided by GDP; Government is government consumption as a share of GDP; Inflation is the change in CPI; Education is the average year of schooling; Institution is the composite

56
country risk measure, taken from international country risk guide (ICRG).AR(2) is the Arellano-Bond serial correlation test (p-value is reported); Over-identification is the Hansen J
statistic (p-value is reported). All specifications include time dummies (coefficients not reported). Constant is not reported. Robust standard errors in parentheses, *** p<0.01, ** p<0.05,
* p<0.1.

57
58
59