More mortgages, lower growth?
Bezemer, Dirk ; Grydaki, Maria ; Zhang, Lu 等
More mortgages, lower growth?
I. INTRODUCTION
A large empirical literature had established the positive effects
of the growth in bank credit on output growth, in data from the 1960s
until the mid 2000s. (1) Recent research however shows that above a
threshold level, a high credit-to-gross domestic product (GDP) ratio may
slow down rather than boost growth (Arcand, Berkes, and Panizza 2012;
Beck etal. 2012; Cecchetti and Kharroubi 2012; Manganelli and Popov
2013; Rousseau and Wachtel 2011; Shen and Lee 2006; Valickova, Havranek,
and Horvath 2013). To illustrate, Figure 1 shows the unconditional
growth correlation of bank credit stocks scaled by GDP across 50
economies since the 1970s. The correlation of credit to output growth
was not significantly different from zero in the 1990s and 2000s. This
motivates our paper.
Different explanations have been proposed. Wachtel (2011) questions
the interpretation of credit/GDP ratios as indicating financial
deepening, and notes it may also indicate increasing financial
fragility. Beck, Degryse, and Kneer (2014) and Beck et al. (2012)
identify the growth in nonintermediation activities and in nonenterprise
credit, respectively, as causes of the weakening growth effectiveness of
the financial sector. Rousseau and Wachtel (2011) suggest 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, Berkes, and Panizza (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 Kharroubi (2013)
present evidence that skilled labor is drawn away from R&D intensive
industries into finance during a credit boom, so that the financial
sector may grow at the expense of the real sector. Earlier, Stockhammer
(2004) analyzed a causal relation for selected Organization for Economic
Co-operation and Development (OECD) economies between expanding asset
markets and a slowdown in investment.
In this paper, we analyze newly collected data originally reported
in national central bank statistics on four categories of bank credit,
for 46 economies over 1990-2011, with country coverage and time period
dictated by data availability. As in several other papers (Arcand,
Berkes, and Panizza 2012; Beck etal. 2012; Cecchetti and Kharroubi
2012), we observe high growth of credit relative to GDP in this new
sample. We also see a rapid increase of the share of household mortgage
debt in total debt, again in line with Beck et al. (2012) and also with
recent findings reported in Bezemer (2014) and Jorda et al. (2014). In
Figure 3 in the next section, we observe in a cross section of countries
over 1990-2011 that total domestic bank debt rose from below 80% to over
120% of GDP, with mortgage credit rising from 20% to 50% and credit to
nonfinancial business remaining stable around 40% of GDP. Credit booms
in the 1990s and 2000s caused credit to asset markets to become a large
(in some countries, the largest) part of bank credit. For instance, in
the Netherlands, credit to asset markets (mostly, household mortgage
credit) accounted for 70% of outstanding loans in 2011, up from less
than 50% in 1990. (2)
The reasons for the disproportional growth and then defaults in
mortgage lending since the 1990s have been explored in several recent
studies. They include policy preferences for increased home ownership
and relaxation of mortgage lending rules (Dell' Ariccia et al.
2012; Jorda etal. 2014), credit market deregulation (Favara and Imbs
2015), and financial innovations such as new forms of securitization
(Jimenez et al. 2012) as part of the originate-to-distribute model of
lending (Purnanandam 2011). Softening of loan standards and
securitization were amplified by low policy rates (Maddaloni and Peydro
2011), while house price declines and deteriorating underwriting
standards triggered the exceptional rise in defaults from 2007 (Mayer et
al. 2009). In this paper, we do not research these and other causes of
mortgage credit expansion, but analyze its effect on output growth. We
hypothesize that the use of credit matters to its growth effectiveness.
The distinction is that credit supporting asset transactions may have
weaker or negative growth effects compared to credit supporting
transactions in goods and services, as suggested by Werner (1997, 2012).
Werner (1997) applied this insight to data on Japan for the 1980s and
1990s. In this paper, we extend this disaggregation to a large panel of
countries.
Our paper is closely related to Beck et al. (2012). This paper was
the first, to our knowledge, to decompose bank credit into enterprise
credit and household credit. They examined the different effects of
these two types of credit on real sector outcomes for 45 countries over
1994-2005. They show that enterprise credit is positively associated
with growth, whereas household credit is not. The present paper is
different in several respects. Compared to Beck et al. (2012) who
investigate a number of real sector outcomes, we focus on growth only.
While Beck et al. (2012) analyze one aggregate household credit
category, we separate mortgage and consumption credit. The distinction
appears important: our results are driven by mortgage credit, not
consumer credit. Further, we observe each credit category independently
as reported in central bank statistics. (3) And while Beck et al. (2012)
use a cross-section of data, we build a panel data set for four credit
categories over a longer time-period. While this allows for more
efficient estimates, our findings broadly confirm those of Beck et al.
(2012).
Our paper also relates to Beck, Degryse, and Kneer (2014), where
the activities of banks are differentiated according to whether banks
intermediate or undertake other, nonintermediation activities. They
contrast the credit-to-GPP-ratio (as proxy for intermediation
activities) and the value-added share of the financial sector (as proxy
for size) and find that intermediation activities are positively
associated with growth whereas an increase in size is not. We
corroborate their main finding that more finance is not necessarily
better. Instead of using the total-credit-to-GDP ratio to measure the
financial intermediation, we focus on how the composition of financial
intermediation activities matters for growth. To do so, we use
disaggregated credit-to-GDP ratios for different credit categories. We
now motivate this distinction.
In most of the credit-growth literature to date, "credit"
is tacitly interpreted as credit to the nonfinancial sector, supporting
production of goods and services (for recent exceptions, see Beck et al.
2012; Beck, Degryse, and Kneer 2014; Bezemer 2014; Jorda et al. 2014).
Biggs, Mayer, and Pick (2010) show that with only nonfinancial-sector
credit, the dynamics of credit, debt, and capital are identical, so that
the growth effect of credit can indeed only be positive. The depth of
financial markets can then be viewed simply as a measure of
economies' productive absorption capacity (Masten, Coricelli, and
Masten 2008) and negative growth coefficients present a puzzle.
In contrast, if credit which finances transactions in assets
(rather than in goods and services) is included in the analysis, the
growth coefficient need not be positive. Credit growth may now inflate
asset markets rather than leading to growth in GDP. This by itself
decreases the credit-growth correlation. It also increases growth of the
credit-to-GDP ratio, since credit stocks grow without (or with much
less) growth in GDP.
Why would financial deepening, measured by credit stocks, have a
negative impact on output growth? Credit stock measures capture
agents' ability to use finance to reallocate factors of production,
which may support growth. This is the traditional, positive
"financial development" effect on growth (King and Levine
1993). But credit stocks are also debt stocks, which may depress growth
through more financial fragility and larger uncertainty, through larger
debt servicing out of income, through a debt overhang effect, or through
a negative wealth effect on consumption. Theoretically, the growth
effect of credit stocks is therefore ambiguous; with large credit
stocks, it may well be negative. What matters is "how large a
credit boom [is] relative to the possibilities of productive uses for
loans" (Lorenzoni 2008; Boissay et al. 2015). Credit growth in
support of other outcomes than production by nonfinancial business (such
as investment in existing real estate) will result in smaller growth
coefficients than credit growth to nonfinancial business. At high levels
of debt, the effect may be negative: a rise in overall debt levels has
been widely noted as a growth retarding factor (Barajas, Chami, and
Yousefi 2013; Boissay et al. 2015; Jorda, Schularick, and Taylor 2013;
Lorenzoni 2008; Radelet and Sachs 1998; Reinhart 2010; Rousseau and
Wachtel 2011; Schularick and Taylor 2012; Wachtel 2011).
To the extent that credit for real estate transactions is household
mortgage credit, this argument is reinforced by the literature on
household credit and growth. Jappelli and Pagano (1994) argue that more
household credit leads to lower private savings and so slower economic
growth. Beck et al. (2012) show that credit to households (most of which
is mortgages, in most economies) has negligible growth effects. Earlier,
Xu (2000) had identified business investment, not household spending, as
the channel through which financial development affects growth.
Buyukkarabacak and Valev (2010) and Buyukkarabacak and Krause (2009)
find that countries with more household credit have higher probabilities
of crisis and weaker external balances. Jappelli, Pagano, and di Maggio
(2008), Barba and Pivetti (2009), and Sutherland et al. (2012) find
positive crisis and recession effects of the expansion of household
credit, respectively. Mian and Sufi (2014) show the close relationship
between the severity of a recession and the build-up of household debt
that preceded it. In the present paper we go beyond this: we find a
negative growth coefficient on average, not just during recessions.
We find that the growth coefficient of total bank credit stocks,
traditionally used to measure financial development, was insignificant
or negative in fixed-effect (FE) panel regressions over the sample
period--also before the 2008 crisis and also when controlling for the
level of credit-to-GDP ratios, for institutions, and for financial
crises. In generalized-method-of-moment (GMM) estimations, this appears
to be due primarily to the negative growth coefficient for credit to
asset markets, predominantly household mortgages. These findings hold up
in regressions with Rajan and Zingales's (1998) methodology, and in
a battery of robustness checks. The negative growth coefficient of
mortgage credit--and of bank credit to asset markets generally--helps
clarify why financial development was not good for growth in our
1990-2011 sample.
While this is our main contribution, we augment the analysis with
the distinction between stocks and flows of credit (Biggs, Mayer, and
Pick 2010). The rationale for this distinction is that credit flows are
a stimulus to growth due to more spending, different from the
traditional reallocation effect of more financial development. 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" (Borio and Lowe 2004, 555; Caporale and Howells
2001). This "expansion of transactions" will be GDP growth
insofar as transactions of goods and services (not of assets) are
involved. Without this distinction, we might overestimate financial
development effects, which are credit stock effects.
This stock-flow distinction is new to the empirical credit-growth
literature, but there is a clear parallel in the fiscal macro
literature. Flows of government deficit spending may boost growth in the
short term, but by simultaneously raising stocks of public debt they may
decrease longer-term growth. The (positive) impact of deficits differs
from the (negative) impact of debt. What goes for public debt, goes for
private debt. We therefore analyze credit stocks and credit flows
separately. We find more negative growth effects of financial
development (measured by credit stocks scaled by GDP) when controlling
for the positive effect of credit flows. (4) Note that our main argument
does not depend on the stock-flow distinction.
Our paper provides evidence for the argument made in Bezemer (2014)
that the empirical credit-growth literature inspired by Schumpeter
(starting with King and Levine 1993) needs to take differentiation of
credit into account. A "[d]istinction between debts according to
purpose, however difficult to carry out," as Schumpeter (1939, 148)
wrote, may help understand changes in the growth-effectiveness of
credit. In the next section we present the new data. Sections III and IV
present the methodology and empirical findings. Section V concludes the
paper with a summary, discussion, and conclusion.
II. DATA
We collected data from the consolidated balance sheets of monetary
financial institutions in central bank sources, for 46 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).
(5) To the best of our knowledge, no data with similar detail has been
collected and reported before. (6) 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
credit categories. We discuss their development over time and across
countries.
A. Definitions and Trends
We define credit stocks as the credit-to-GDP ratio:
(1) [s.sub.i,t] = [C.sub.i,t]/[GDP.sub.i,t],
where i denotes country, t denotes time, and C is a credit measure.
We measure credit flows by the annual change of credit stocks relative
to lagged GDP. as follows (Biggs, Mayer, and Pick 2010):
(2) [f.sub.i,t] = ([C.sub.i,t] - [C.sub.i,t-1])/[GDP.sub.i,t-1].
We aggregate the four types of credit into two broader categories:
"nonfinancial" credit (credit to nonfinancial business plus
household consumption loans) and "asset market" credit
(mortgages plus credit to financial business). The latter follows the
"finance, insurance and real estate" sectors classification of
the U.S. National Income and Product Accounts. (7)
Three features stand out in the 1990-2011 data: the expansion of
credit relative to GDP over time (Figure 2), the changing composition of
credit stocks (Figure 3), and the correlation of stocks and flows of
credit categories with economic growth (Figure 4). Figure (2A) shows
that for a balanced panel of 14 countries in our data--selected on data
availability--on average the total-credit-to-GDP ratio increased from
75% to 120% over 1990-2011. Although the finance-growth relation differs
between developed and emerging economies (Rioja and Valev 2004), we
observe that the increase is pronounced in both country groupings.
Figure 2B shows the trends for five selected developed economies. In
Spain, the credit-to-GDP ratio rose over 1992-2011 from 118% in 1992 to
389% at the time of the 2008 financial crisis. The increases are also
pronounced in the Netherlands, from 77% to 210%; in Greece, from 33% to
115%; and in the United Kingdom, from 39% to 90% over the 1990-2011
period. Figure 2C for emerging economies shows that here much of the
increase in the credit-to-GDP ratio occurred in the 2000s. In Croatia
for instance, the credit-to-GDP ratio increased from 55% in 2001 to 150%
in 2011. Declines were rare and often associated with episodes of
financial crisis. The average of country credit ratios peaked and then
declined slightly after the 2008 financial crisis.
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.
Figure 3A shows that most of the growth in the credit-GDP ratio is due
to growth of credit to asset markets, especially mortgage credit (bank
credit to nonbank financials is small in these data). The ratio of
nonfinancial credit to GDP is roughly stable over time around 40%. We
study the shifting credit composition in more detail. Figure 3B first
illustrates that the share of nonfinancial credit in total credit varies
considerably across countries. It appears to be negatively correlated to
income levels. Figure 3C shows the shift in credit composition over
time. The vertical distance to the diagonal measures a country's
shift in the share of nonfinancial credit in total credit between its
first and last observation. The share was nondeclining in 10 countries,
positive in one and falling in all others.
A third observation is on the credit-growth relation, for stocks
and flows of credit. Table 1 presents the growth correlations of stocks
and flows of the two credit aggregates and the four categories of
credit. There appears to be a robustly negative cross-section relation
over 1990-2011 of credit stocks relative to GDP with real per capita GDP
growth, though with significant scatter and possible nonlinearity around
the trend line (Figure 4A). There also appears to be a positive
correlation over time of per capita output growth with total-credit
flows (Figure 4B). Panel A in Table 1 shows that the negative
correlation of credit stocks with growth is mainly driven by mortgages
and (to a lesser extent) financial-sector credit. The correlation of
growth with credit to all asset markets is less negative and less
significant. Panel B further shows that flows of nonfinancial credit
have the highest correlation with growth, closely followed by its two
components, nonfinancial business credit and household consumption
loans. Growth correlations of credit flows to financial business and
household mortgage credit flows are much smaller. We also note the large
correlations of total credit stocks with mortgage credit stocks.
III. EMPIRICAL STRATEGY
We regress real GDP per capita growth on annual stocks and flows of
total credit and of the two credit aggregates, controlling for other
determinants of growth. Given the short time span of our sample, we use
3-year averages of the underlying annual data to iron out business cycle
fluctuations. (8) We start with a baseline FE panel data baseline model
over 1990-2011 for 46 countries. Then we estimate system-GMM and
difference-in-difference models to account for endogeneity. The baseline
specification is:
(3) [g.sub.i,t] = [alpha] + [[beta].sub.1] [s.sub.i,t] +
[[beta].sub.2] [f.sub.i,t] + [gamma] [X.sub.i,t] + [[phi].sub.i] +
[[phi].sub.t] + [[epsilon].sub.i,t],
where [g.sub.i,t] is the growth rate of real GDP per capita (2000
constant US dollar) of country i in 3-year period t; Coefficients
[[beta].sub.1] and [[beta].sub.2] capture the relations of credit stocks
([s.sub.i,t]), and credit flows ([f.sub.i,t]) with growth, respectively,
where we will estimate a total-credit measure, "nonfinancial"
credit and "asset market" credit separately. [X.sub.it] is a
vector of control variables, including the level of real GDP per capita
at the beginning of t, trade openness (imports plus exports as a
percentage of GDP), government expenditure as a share of GDP, inflation,
education (average years of schooling of the adult population), and a
composite country risk indicator as a proxy for institutional quality,
ranging from 50 (low institutional quality) to 100 (high institutional
quality). We include unobserved country-specific time-invariant effects
in [[phi].sub.i], time dummies [[phi].sub.t] and a white-noise error
term with mean zero [[epsilon].sub.i,t]. In robustness checks we will
also include an interaction term of credit flows with credit stocks and
a systematic banking crises indicator (Laeven and Valencia 2013). (9)
Table 2 summarizes definitions, sources, and descriptive statistics.
Since financial development may be endogenous to growth, we also
estimate a GMM dynamic panel model. (10) We difference (3) to obtain:
(4) [DELTA][g.sub.i,t] = [[beta].sub.1] [DELTA][s.sub.i,t] +
[[beta].sub.2] [DELTA] [f.sub.i,t] + [gamma] [DELTA] [X.sub.i,t] +
[DELTA] [[phi].sub.t] + [DELTA][[epsilon].sub.i,t],
and then estimate Equations (3) and (4) using system-GMM
estimation. The endogenous credit variables are now instrumented by
their lags in Equation (4). We use lagged differences as instruments for
the levels Equation (3) and lagged variables in levels as instruments
for the differenced Equation (4). (11) The consistency of the GMM
estimator depends on the validity of instruments and on the validity of
the assumption that the error term, [[epsilon].sub.i,t], does not
exhibit serial correlation. We apply the 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.
Third, we will also use the Rajan and Zingales (1998)
industry-level methodology to account for the endogeneity of credit to
growth. (12) In contrast to past studies based on cross sectional data
(including Rajan and Zingales 1998), we use panel data. Our approach has
two distinctive features compared to similar analyses. First, we are
able to control for a wider range of industry-time and industry-country
fixed effects. This alleviates omitted variables bias. Second, by
including the credit variable itself, in addition to its interaction
with financial dependence, our specification allows for an assessment of
the direct effect of credit on industry-level growth. Our specification
is:
(5) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],
where j denotes industry, i denotes country, and t denotes time
(i.e., a 3-year period). This specification is closely related to Braun
and Larrain (2005); growth is measured as the annual percentage change
of industry real value added. (13)
Share is defined as the size of each industry as a percentage of
manufacturing value added at the beginning of each 3-year period.
Similar to our country-level specifications above, s and f denote the
stocks and flows of credit categories. ED is the external financial
dependence indicator, taken from Rajan and Zingales (1998). We include a
series of dummy variables to control for industry ([[mu].sub.j]),
country (([phi].sub.i]), time (([[phi].sub.t]), industry-time
([[delta].sub.j,t]), and industry-country ([[eta].sub.j,i]); fixed
effects. We include the same vector of control variables [X.sub.i,t] as
in Equations (9) and (10), which vary at the country-time dimension.
Finally, [[epsilon].sub.j,i,t], is an error term.
Our industry-level analysis covers an unbalanced panel of 36 ISIC
three- and four-digit manufacturing industries for 41 countries during
1990-2011 from the United Nations Industrial Development Organization
Industrial Statistics Database (INDSTAT4). We ensure that the number of
industries available through time is constant across each individual
country, while the number of industries across countries may vary. Table
A4 in the Appendix lists the 41 countries and the availability of
industry coverage. Table A5 lists 36 industries, ISIC code, and the
value of external financial dependence per industry.
It is worth noting that our industry-level methodology presupposes
that the dependence on external finance is sufficiently constant over
time that the industry ranking does not change. In defense of this
assumption, Rajan and Zingales (1998) argue (but do not test) that
external dependence on finance reflects technological factors of an
industry, such as project scale, gestation and cash harvest period, and
the need for continual investment. These industry-inherent
characteristics are likely to be persistent over time. In support,
Kroszner, Laeven, and Klingebiel (2007) report a correlation of 0.82
between the external dependence index for the period 1980-1999 measure
and the original Rajan-Zingales measure for 1980-1989. Haltenhof et al.
(2014) also find remarkable stability of the external dependence
measures over the 1980s, 1980-1997, and 1990-2011. The latter period
supports the use of 1980s values in our 1990-2011 sample.
A more problematic assumption may be that the U.S.-based measure of
external dependence is a valid proxy for the same industries across
countries. Furstenberg and Kalckreuth (2006) show that even in the
United States, the measure for financing conditions in manufacturing
industries changes when using an alternative source of industry data.
Furthermore, they find that the presumed correlations between external
dependence and a number of structural/technological industry
characteristics, are in fact weak. Nevertheless, Kroszner, Laeven, and
Klingebiel (2007) provide some empirical validation of this assumption.
They find that their results are robust when using other countries
(Canada and noncrisis countries) as the benchmark, instead of the United
States. In view of these caveats, we present the Rajan-Zingales results
as additional rather than conclusive evidence.
IV. ESTIMATION RESULTS
In this section we present estimation results for stocks and flows
of a total-credit measure and of credit aggregates. We then proceed with
a variety of robustness checks and a discussion of our findings.
A. Credit Stocks, Credit Flows, and Their Growth Effects
Table 3 presents the results of the FE panel baseline model
(columns 1-3) and the system-GMM model (columns 4-6). Results for credit
stocks are in columns (1) and (4), results for credit flows in columns
(2) and (5). Credit stocks, the common measure for financial
development, have no significant positive correlation to growth, in line
with other studies (Rousseau and Wachtel 2011; Stengos et al. 2007;
Valickova, Havranek, and Horvath 2013). We go beyond this observation in
columns (3) and (6), where both stocks and flows are included. We
observe negative (but weakly significant) growth effects of credit/GDP
stocks. That is, controlling for the positive effect of credit flows,
financial development appears bad for growth. Credit-growth studies
which do not control for the positive effect of credit flows will tend
to overestimate the stock effect, which represents financial
development. But even without controlling for flows (i.e., adopting the
common methodology in the credit-growth literature), the growth effect
of financial deepening was insignificantly different from zero over
1990-2011.
We proceed to distinguish "nonfinancial" from "asset
market" credit. Table 4 reports 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 is not rejected.
We find that stocks of both credit aggregates correlate negatively
to growth. It is striking that the coefficient for nonfinancial credit
stocks is significantly negative in the FE specifications but no longer
significant in the system-GMM models. Conversely, the coefficient for
asset market credit is negative and significant in the system-GMM
specification only. While it is hazardous to attach firm conclusions to
the change in statistical significance, this pattern would be consistent
with a negative effect of asset market credit after controlling for
endogeneity, but not for credit to nonfinancial business after
controlling for endogeneity. This begs the question what endogeneity
problems would be consistent with this pattern? (14) An important
endogeneity highlighted in the credit-growth literature is reverse
causality from growth to credit. We tentatively suggest the following
interpretation. Stocks of asset market credit such as mortgages
accumulate in response to income growth (a positive correlation due to
reverse causation). But they are also a burden to growth due to a
combination of repayment, consumption and debt overhang effects (a
negative correlation). This combination of positive and negative effects
may produce the insignificant estimates in the FE specification, but
correcting for the reverse causality underlying the positive effect, we
obtain significantly negative GMM coefficients. Further, it is plausible
that nonfinancial business borrowing responds less strongly to GDP
growth than does asset market investment, and that nonfinancial business
investments supported by bank credit produce growth effects which are
larger and more persistent than the asset price increases that credit to
asset markets causes. If we accept these two assumptions, then the
positive reverse causality effect is weaker for credit to nonfinancial
business (leading to on balance negative FE estimates) while the
positive "true" credit-growth effect is stronger (leading to
on balance insignificant GMM effects). We emphasize that this
interpretation is tentative and subject to further research.
In column (12), a one standard deviation increase in the stock of
asset-market credit corresponds to 0.74 standard deviation decrease in
the growth rate, which is equal to a 1.83 percentage points decrease in
growth in this sample. (15) Considering that the average growth rate in
our sample is 2.3 percentage points, the effect is large. The result in
column (9) implies that a one standard deviation increase in
nonfinancial credit flows is associated with a 0.32 standard deviation
increase in growth, which is equal to an additional 0.79 percentage
point increase in growth in this sample. (16) Overall, the results
suggest that controlling for endogeneity, the growth effect of financial
deepening of asset markets, as measured by credit stocks, was negative.
B. Industry-level Evidence
Estimation results applying the Rajan and Zingales (1998)
methodology are shown in Table 5. The "external dependence on
finance" variable is defined as the annual excess of investment
over profit, that is, the annual flow of bank credit and other borrowing
to finance investment. Thus, it captures ability to access external
finance. Columns (1)-(3) show the results for total credit, columns
(4)-(6) and columns (7)-(9) report results for nonfinancial credit and
credit to asset markets, respectively. The results are in line with the
panel data estimations. We find that the coefficient for credit stocks
is again consistently negative, with more significant coefficients for
credit to asset markets. The positive coefficients for the interaction
of credit stocks and financial dependence suggest that firms in
industries which are better able to access external finance, experience
smaller growth-retarding effects from debt stocks. In line with this,
the coefficient for nonfinancial credit flows is positive. Coefficients
for flows of (mostly) household mortgage credit are insignificantly
different from zero--which is unsurprising in this industry-level
analysis. The bottom panel of Table 5 reports marginal effects. The
implied growth difference between high external dependence (ED) and low
ED industries is 3-4 percentage points growth.
C. Interacting Credit Stocks and Flows
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 (Cecchetti and Kharroubi 2012;
Stockhammer 2004). Not accounting for these effects might partly drive
our results through omitted-variable bias. We therefore introduce an
interaction term of credit flows with stocks (i.e., with financial
development). Table 6 reports the results. We find a negative
interaction effect between credit stocks and credit flows for
nonfinancial credit and for total credit, with weak significance. At
higher levels of financial development (credit stocks), the growth
effect of credit flows is indeed smaller. Possible interpretations
include diminishing returns to credit and a balance sheet effect of
debt.
D. Robustness Tests
We run a number of robustness checks. Table 7 summarizes the
findings. Due to space limitations, we do not include full regression
tables, which are available on request. 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 relations of their underlying
components. We report FE results for each of the four underlying credit
categories in columns (1a)-(4a) and system-GMM results in columns
(1b)-(4b). We find that the negative relations between credit stocks and
growth holds overall but is particularly strong for nonfinancial
business credit (column (1a)) and mortgage credit (column (3b)). This is
unsurprising since they constitute the bulk of their respective
aggregates. None of the four components have coefficients with an
opposite sign to their aggregate. This suggests that the stock
aggregates do not hide significant heterogeneity in the underlying
credit-growth relations. Credit flows to nonfinancial business are
positively related to growth. Coefficients for flows of mortgage and
consumer credit are both insignificant in the system-GMM results.
Further, a potential bias may arise from the equal treatment of
countries with high and low levels of credit stocks, if the relation
between credit and growth is nonlinear over credit stocks. First, we
check whether our results are driven by countries with high-credit
stocks but low growth (Denmark, Spain, and Switzerland) or low credit
levels but high growth (Armenia, India, and Uruguay). We drop these six
countries and report results for "nonfinancial credit" and
"asset market credit" in columns (5a)-(6a) and columns
(5b)-(6b). Second, we test whether the results are similar in countries
with high and low levels of credit stocks. We construct two subsamples
based on the distribution of the average credit stocks per country, one
excluding countries in the lowest quantile (a
"high-credit-stocks" subsample), and the other excluding the
highest quantile (a "low-credit-stocks" subsample). Results
are shown in columns (7a)-(10a) and (7b)-(10b). In both analyses, our
results do not qualitatively change.
Moreover, Rousseau and Wachtel (2011) find that the positive
relationship that was estimated using the data from the 1960s to the
1980s disappeared over the subsequent 15 years as a result of the
increased incidence of crises. Other papers show that the link between
credit and growth varies over the business cycle (Borio 2014; Braun and
Larrain 2005; Jorda, Schularick, and Taylor 2013). 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 in columns
(11a)-(12a) and (11b)-(12b). The results are consistent with our longer
sample.
We also address Rousseau and Wachtel's (2011) argument by
including the Laeven and Valencia (2013) systematic banking crises
variable. We characterize a 3-year country observation as a crisis
episode if the country was in crisis for at least 1 year during this
period. (17) Of the 46 countries in our sample, 20 experienced at least
one crisis episode. We introduce an interaction term between credit
stocks and crisis episodes, controlling for any independent effect of
crises on growth. The results in columns (13a)-(14a) and (13b)-(14b)
show that the coefficient for nonfinancial credit stocks is significant
and negative in the FE estimation, the coefficient for asset market
credit is significant and negative in the GMM estimation, just as in
Table 4. Our results are not driven by country-specific banking crisis.
V. SUMMARY AND CONCLUSION
Financial deepening is a double-edged sword. It supports
investments and increases the economy's capacity to reallocate
factors of production. But a large credit-to-GDP ratio may be a drag on
growth. It may imply high levels of private debt, reduce investment and
innovation, and induce volatility, financial fragility, and crisis. We
show that credit to real estate and other asset markets tends to
increase the credit-to-GDP ratio while stocks of credit to nonfinancial
business rise roughly in line with GDP. In recent decades, a shift in
the composition of credit toward real estate and other asset markets has
therefore coincided with rising credit-to-GDP ratios. It may have
diminished the growth effectiveness of credit.
To test this conjecture, we present and analyze new, hand-collected
data for 46 economies over 1990-2011. We document and explore trends in
credit categories. We find that the growth coefficient of different
credit stocks scaled by GDP is insignificant or negative, especially
credit stocks supporting asset markets. We observe insignificant or
negative correlations of credit stocks with output growth. This holds up
in FE panel data regressions, dynamic panel estimations (system-GMM
models), in regressions with the Rajan and Zingales (1998) methodology,
and in robustness checks. These results are confirmed in an
industry-level difference-in-difference analysis. The positive effect of
credit flows diminishes at higher levels of financial development.
These results are in line with declined growth effectiveness of
financial development as a result of a change in the use of bank credit.
Bank credit has shifted away from nonfinancial business toward asset
markets, where it has no or small growth effects. This shift toward more
credit to asset markets also implies faster growth of credit stocks
relative to GDP, which may be harmful in itself.
A clear limitation of our study is the short time sample, which is
dictated by data availability. We would also like to see a better
disaggregation of credit, for instance between business mortgages and
other business credit. Given the recent crises in commercial real estate
markets in several economies, this would seem a relevant distinction.
Researchers depend here on the quality and detail of the data provided
by central banks. Another limitation is the use of country-level data.
In future research, the effect of bank loan portfolios on output growth
could be examined in matched bank-firm data. Recent studies in this vein
show promising results (e.g., Jimenez et al. 2012). The use of
microlevel data also opens up new avenues for dealing with endogeneity
problems.
In summary, our new data and analysis suggest that what was true in
the 1960s, 1970s, and 1980s when the field of empirical credit-growth
studies blossomed, is no longer true in the 1990s and 2000s. Banks do
not primarily lend to nonfinancial business and financial development
may no longer be good for growth. These trends predate the 2008 crisis.
They prompt a rethink of the role of banks in the process of economic
growth.
Our findings are consistent with broader concerns with a world
which has too much rather than too little financial development.
Cecchetti and Kharroubi (2012) 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."
Piketty (2014) suggests that a large ratio of capital to income may
depress growth, where capital is the sum of financial and fixed capital.
His empirical work shows that most of the increase in the capital-income
ratio is due to the increase in the value of financial assets. In our
data, we observe large increases in bank lending supporting asset
markets and insignificant or negative growth correlations of these
credit stocks. Mian and Sufi (2014) emphasize the role of household
leverage in a consumption slowdown in high-debt economies. Summers
(2013) suggests that equilibrium real interest rates may have been
declining over the last decades, possibly to negative values. In these
views, more financial development leading to more savings, more
financial capital, lower interest rates, and more debt may not stimulate
growth. Our estimates show that even though credit flows may constitute
a stimulus to growth, credit stocks--the traditional measure for
financial development--have negative or insignificant growth
coefficients.
The common theme between these analyses and our paper appears to be
that there are costs to having an economy and a financial system
increasingly geared toward growing markets for real estate and financial
assets. This opens up a wide array of research questions. It is not
clear that these trends arise because of growing inequality, as Piketty
suggests. It is unclear which of the many reasons suggested by Summers
are relevant to negative real returns. We do not know whether the
finance-growth relation we document for the last two decades is a
temporary or secular trend. These are subjects for future research.
doi: 10.1111/ecin.12254
ABBREVIATIONS
CPI: Consumer Price Index
ED: External Dependence
FDS: Financial Development and Structure
FE: Fixed-Effect
GDP: Gross Domestic Product
GMM: Generalized-Method-of-Moment
ICRG: International Country Risk Guide
INDSTAT4: United Nations Industrial Development Organization
Industrial
Statistics Database
OECD: Organization for Economic Co-operation and Development
TC: Total Credit
APPENDIX: DATA
The aim of the database is to provide a detailed description of
monetary financial institutions' (banks and credit unions) loan
assets where the counterparty is a domestic nongovernment nonbank. We
collected data from the consolidated balance sheet of monetary financial
institutions from central bank sources of 46 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). (18)
Lending to government by banks is usually a very small part of
total bank lending. 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 to 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.
Domestic bank credit includes loans by both domestic and foreign
banks, in domestic and foreign currency. For reasons of consistency, it
excludes nonbank lending and securitized bank loans. Some countries 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 United States are comparatively low: most credit in the
United States 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 United States was 386% in 2011
(BEA flow of fund data), of which only 34% was bank credit (this data).
However, the United States 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 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 10 or 15 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). 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 1977, the United States to 1960. But on average, data
before 1990 were rare.
COMPARISON TO SIMILAR DATA
Beck et al. (2012) and Biiyukkarabacak 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 are not an update of this, but are
newly collected. We aimed to separate out mortgage and other household
credit and to observe each credit category at source. The Beck et al.
(2012) data combine mortgage and other household credit into one
household credit category. The data are based on the financial
development and structure (FDS) database described in Beck,
Demirguc-Kunt, and Levine (2000) and updated in Beck, Demirguc-Kunt, and
Levine (2010). Here "private credit" captures the financial
intermediation with the private nonfinancial 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"). In observing the different credit aggregates,
Beck et al. (2012) start with a "total credit" (TC) measure
taken from the FDS database, which is credit to nonfinancial business
(BC) plus credit to households. The "household credit" measure
in Beck, Demirguc-Kunt, and Levine (2010) and in Biiyukkarabacak and
Valev (2010) is defined as (TC-BC), that is, all nonbusiness credit,
including both consumer credit and mortgage credit. These are not
distinguished. The Beck et al. (2012) credit data are deflated by the
consumer price index (CPI) deflator and then divided by real (deflated)
GDP. Our data is nominal credit divided by nominal GDP.
Table A2 is a comparison of our data to the Beck et al. (2012)
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
writedowns 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 Uruguay. 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 the Biiyukkarabacak and Valev (2010)
measure. Reclassifications of what counts as a bank may be behind this.
There is also some disparity on the United Kingdom.
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, Drehmann, and Muksakunratana (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.
Since bank debt is not disaggregated, we cannot directly compare the BIS
data to our data.
TABLE A1
Credit Stocks Across Countries (% of GDP)
Non
financial Con- Mort- Financial
Country Start End Business sumer gage Business Total
ALB 2005 2011 23.468 6.541 32.775 62.784
ARM 2005 2011 11.163 4.069 1.938 0.39 17.17
AUS 1994 2011 34.748 7.462 48.507 8.001 98.718
AUT 1995 2011 47.674 38.202 6.609 92.485
BEL 1999 2011 30.167 5.185 28.032 9.669 73.053
BGR 1998 2011 25.722 6.408 5.261 37.391
BRA 1994 2011 19.671 5.755 2.067 3.887 31.38
CAN 1990 2011 40.139 25.225 43.519 11.652 120.535
CHE 1990 2011 43.161 92.931 1.516 137.608
CHL 1990 2011 43.381 36.427 11.262 91.07
CZE 1997 2011 24.613 5.287 8.591 3.089 41.58
DEU 1990 2011 52.973 10.874 28.532 3.204 95.583
DNK 2000 2011 44.121 25.34 78.656 7.189 155.306
EGY 1991 2011 36.225 7.714 43.939
ESP 1992 2011 59.497 61.904 106.719 228.12
EST 1999 2011 22.55 2.161 20.497 6.496 51.704
FIN 2002 2011 27.272 13.409 33.675 0.685 74.356
FRA 1993 2011 36.177 12.514 26.556 4.111 79.358
GBR 1990 2011 21.11 8.693 36.584 26.032 92.419
GRC 1990 2011 34.011 7.142 15.43 0.984 56.583
HKG 1990 2011 68.83 11.604 39.323 15.192 134.949
HRV 2001 2011 61.143 32.591 9.446 103.18
HUN 1990 2011 15.393 3.325 5.279 3.316 27.313
IDN 2002 2011 15.269 5.33 1.994 2.44 25.033
IND 2001 2011 25.292 3.259 3.248 2.364 34.163
ISL 2003 2011 14.132 41.737 55.869
ISR 1999 2011 57.681 10.545 20.44 88.666
ITA 1998 2011 22.691 2.776 23.921 11.673 61.061
JPN 1990 2011 56.62 3.211 28.048 8.538 96.417
LTU 1993 2011 18.373 2.981 7.35 1.732 30.436
LUX 1999 2011 30.183 8.318 33.755 50.92 123.176
MAR 2001 2011 14.651 2.906 11.232 0.293 28.789
MEX 2000 2011 7.955 3.109 8.69 19.754
NLD 1990 2011 49.114 8.104 60.016 19.99 137.224
NOR 1995 2011 33.606 11.069 49.059 2.875 96.609
NZL 1990 2011 32.034 4.846 55.136 25.089 117.105
POL 1996 2011 14.785 9.476 7.048 1.201 32.51
PRT 1990 2011 41.924 11.202 37.904 10.876 101.906
SGP 1990 2011 59.365 23.2 13.397 95.962
SVK 2004 2011 19.773 4.337 11.163 2.403 37.676
SVN 2004 2011 48.487 11.476 8.729 4.933 73.625
SWE 1996 2011 55.302 11.221 40.686 48.119 155.328
TWN 1997 2011 70.616 19.951 38.949 3.426 132.942
UKR 2005 2011 39.58 19.441 5.243 64.264
URY 2005 2011 13.883 8.908 22.791
USA 1990 2011 9.547 6.032 18.823 34.402
TABLE A2
Comparison to Other Datasets
Private Credit
Ours BECK2012 BUY2010
1994-2005 1994-2005 1990-2006
AUS 0.777 0.823 0.806
AUT 0.84 1.005 1.035
BEL 0.651 0.744 0.748
BGR 0.198 0.219 0.245
CAN 1.029 0.962 1.012
CHE 1.369 1.603 1.6
CZE 0.347 0.484 0.481
DEU 0.97 1.053 1.053
DNK 1.277 0.894 0.338
EGY 0.495 0.446 0.432
EST 0.194 0.286 0.336
FRA 0.69 0.85 0.86
GBR 0.62 1.269 1.337
GRC 0.444 0.663 0.691
HUN 0.212 0.231 0.302
IDN 0.205 0.252 0.249
IND 0.249 0.219 0.227
ISL 0.551 0.918 0.916
JPN 0.903 1.549 1.105
LTU 0.159 0.149 0.177
MAR 0.224 0.187 --
MEX 0.179 0.186 0.194
NLD 1.093 1.639 1.152
NZL 0.861 1.118 1.152
POL 0.238 0.244 0.229
PRT 0.85 1.103 0.961
SVK 0.219 0.415 0.409
SVN 0.476 0.34 0.362
SWE 0.963 0.636 0.374
URY 0.219 0.392 0.329
USA 0.324 0.498 0.503
Nonfinancial Business Credit
Ours BECK2012 BUY2010
1994-2005 1994-2005 1990-2006
AUS 0.311 0.279 0.285
AUT 0.486 0.653 0.683
BEL 0.315 0.314 0.319
BGR 0.148 0.145 0.157
CAN 0.396 0.188 0.128
CHE 0.445 0.604 0.62
CZE 0.272 0.314 0.309
DEU 0.558 0.653 0.605
DNK 0.379 0.133 0.095
EGY 0.411 0.372 0.355
EST 0.11 0.176 0.208
FRA 0.344 0.339 0.337
GBR 0.187 0.557 0.293
GRC 0.294 0.379 0.389
HUN 0.143 0.189 0.218
IDN 0.142 0.17 0.169
IND 0.203 0.156 0.159
ISL 0.118 0.492 0.39
JPN 0.596 1.07 0.747
LTU 0.126 0.104 0.13
MAR 0.122 0.14 --
MEX 0.078 0.087 0.122
NLD 0.465 0.63 0.478
NZL 0.301 0.703 0.444
POL 0.143 0.135 0.162
PRT 0.391 0.507 0.124
SVK 0.142 0.265 0.262
SVN 0.339 0.24 0.252
SWE 0.534 0.233 0.228
URY 0.143 0.194 0.174
USA 0.095 0.118 0.095
Household Credit
Ours BECK2012 BUY2010
1994-2005 1994-2005 1990-2006
AUS 0.466 0.544 0.52
AUT 0.354 0.352 0.352
BEL 0.336 0.43 0.439
BGR 0.05 0.075 0.088
CAN 0.633 0.773 0.892
CHE 0.924 1 0.98
CZE 0.075 0.171 0.172
DEU 0.412 0.4 0.375
DNK 0.898 0.761 0.247
EGY 0.083 0.075 0.073
EST 0.084 0.111 0.127
FRA 0.346 0.511 0.513
GBR 0.433 0.712 1.04
GRC 0.15 0.283 0.3
HUN 0.069 0.042 0.085
IDN 0.063 0.082 0.08
IND 0.046 0.063 0.068
ISL 0.434 0.426 0.526
JPN 0.307 0.479 0.357
LTU 0.032 0.045 0.064
MAR 0.102 0.046 --
MEX 0.101 0.099 0.072
NLD 0.628 1.01 0.98
NZL 0.56 0.415 0.718
POL 0.095 0.11 0.07
PRT 0.458 0.596 0.51
SVK 0.077 0.15 0.151
SVN 0.138 0.099 0.11
SWE 0.429 1 0.149
URY 0.077 0.198 0.155
USA 0.229 0.38 0.408
Notes: BECK2012 and BUY2010 refer to Beck et al. (2012) and
Buyukkarabacak and Valev (2010), respectively. This table
shows the comparison of private credit (excluding financial
business credit), nonfinancial business credit and household
credit (the sum of consumer credit and mortgage credit) for
31 countries (that exist in all three dataset except MAR)
between our dataset, Beck et al. (2012), and Buyukkarabacak
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, LVA, FIN, and KOR dropped out due
to limited time span in our dataset. Thirteen countries in
our dataset, namely BRA, CHL, ARM, HKG, HRV, ISR, ITA, ESP,
NOR, TWN, UKR, LUX, SGP do not exist in either Beck et al.
(2012) or Biiyukkarabacak and Valev (2010).
TABLE A3
Country Coverage
AUS, BDI. BFA, BHS. BOL, CIV, CMR. CRI, DOM, ECU,
EGY, FIN, FJI, GAB, GHA, GMB, GRC, GTM, HND, IND,
IRL, ISR, ITA, JPN, KEN, KOR, LKA. MDG, MEX, MLT,
MYS, NER, NGA, NPL, PAK, PAN, PER, PHL, PRT, PRY,
SEN, SGP, SLV, SWZ, TGO, THA, TTO, TUR, URY, USA
TABLE A4
Industry Coverage Across Countries
ALB(11), AUS(36), AUT(36), BEL(36), BGR(36),
BRA(12), CAN(35), CHE (18), CHL(30),
CZE(31), DEU(36), DNK(32), EGY(36), ESP(36),
EST(36), FIN(34), FRA(36), GBR(36),
GRC(36), HKG(9), HUN(36), IDN(36), IND(32), ISL(29),
ISR(24), ITA(36), JPN(36),
LTU(36), LUX(29), MAR(36), MEX(36), NLD(36),
NOR(36), NZL(14), POL(36), PRT(36),
SGP(36), SVK(34), SVN(36), SWE(36), TWN(28),
URY(33), USA(34)
Note: The number in the parenthesis indicates the number
of industries available.
TABLE A5
Industry Classification and External Financial Dependence
External
ISIC code Sector Dependence (ED)
311 Food products 0.14
313 Beverages 0.08
314 Tobacco -0.45
321 Textiles 0.4
322 Apparel 0.03
323 Leather -0.14
324 Footwear -0.08
331 Wood products 0.28
332 Furniture 0.24
341 Paper products 0.18
342 Printing and publishing 0.2
352 Other chemical products 0.22
353 Refineries 0.04
354 Petroleum and coal 0.33
355 Rubber products 0.23
356 Plastic products 1.14
361 Pottery -0.15
362 Glass and products 0.53
369 Nonmetal products 0.06
371 Iron and steel 0.09
372 Nonferrous metal 0.01
381 Metal products 0.24
382 Machinery 0.45
383 Electrical machinery 0.77
384 Transport equipment 0.31
385 Professional equipment 0.96
390 Other manufacturing 0.47
3211 Spinning -0.09
3411 Pulp and paper 0.15
3511 Basic chemicals 0.25
3513 Synthetic resins 0.16
3522 Drugs 1.49
3825 Office and computing 1.06
3832 Radio 1.04
3841 Ship building 0.46
3843 Motor vehicles 0.39
Note: The external dependence on finance is taken
from Rajan and Zingales (1998).
REFERENCES
Ang, J. B. "A Survey of Recent Developments in the Literature
of Finance and Growth." Journal of Economic Surveys, 22(3), 2008,
536-76.
Arcand, J.-L., E. Berkes, and U. Panizza. "Too Much
Finance?" IMF Working Papers 12/161, International Monetary Fund,
Washington, DC, 2012.
Arellano, M., and O. Bover. "Another Look at the Instrumental
Variable Estimation of Error-components Models." Journal of
Econometrics, 68(1), 1995, 29-51.
Barajas, A., R. Chami, and R. Yousefi. "The Finance and Growth
Nexus Re-examined: Do All Countries Benefit Equally?" IMF Working
Papers 13/130, International Monetary Fund, Washington, DC, 2013.
Barba, A., and M. Pivetti. "Rising Household Debt: Its Causes
and Macroeconomic Implications--A Long-period Analysis." Cambridge
Journal of Economics, 33(1), 2009, 113-37.
Barro, R. J., and J. W. Lee. "A New Data Set of Educational
Attainment in the World, 1950-2010." Journal of Development
Economics, 104, 2013, 184-98.
Beck, T., A. Demirguc-Kunt, and R. Levine. "A New Database on
the Structure and Development of the Financial Sector." World Bank
Economic Review, 14(3), 2000, 597-605.
--. "Financial Institutions and Markets Across Countries and
Over Time: The Updated Financial Development and Structure
Database." World Bank Economic Review, 24(1), 2010, 77-92.
Beck, T., B. Biiyukkarabacak, F. K. Rioja, and N. T. Valev.
"Who Gets the Credit? and Does It Matter? Household vs. Firm
Lending Across Countries." The B.E. Journal of Macroeconomics,
12(1), 2012, 1-46.
Beck, T., H. Degryse, and C. Kneer. "Is More Finance Better?
Disentangling Intermediation and Size Effects of Financial
Systems." Journal of Financial Stability, 10, 2014, 50-64.
Bena, J., and P. Ondko. "Financial Development and the
Allocation of External Finance." Journal of Empirical Finance,
19(1), 2012, 1-25.
Bezemer, D. "Schumpeter Might Be Right Again: The Functional
Differentiation of Credit." Journal of Evolutionary Economics, 24,
2014, 935-50.
Biggs, M., T. Mayer, and A. Pick. "Credit and Economic
Recovery: Demystifying Phoenix Miracles." DNB Working Papers 218,
Amsterdam, 2010.
BIS. "Long Series on Credit to Private Non-financial
Sectors." BIS Quarterly Review, 17(1), 2013, 68-85.
Blundell, R., and S. Bond. "Initial Conditions and Moment
Restrictions in Dynamic Panel Data Models." Journal of
Econometrics, 87(1), 1998, 115-43.
Boissay, F., F. Collard, and F. Smets (forthcoming). "Booms
and Systemic Banking Crises." Journal of Political Economy, 2015.
Borio, C. E. V. "The Financial Cycle and Macroeconomics: What
Have We Learnt?" Journal of Banking & Finance, 45, 2014,
182-98.
Borio, C. E. V., and P. Lowe. "Securing Sustainable Price
Stability: Should Credit Come Back from the Wilderness?" BIS
Working Papers 157, Bank for International Settlements, Basel,
Switzerland, 2004.
Braun, M., and B. Larrain. "Finance and the Business Cycle:
International, Inter-industry Evidence." Journal of Finance, 60(3),
2005, 1097-128.
Buyukkarabacak, B., and S. Krause. "Studying the Effects of
Household and Firm Credit on the Trade Balance: The Composition of Funds
Matters." Economic Inquiry, 47(4), 2009, 653-66.
Buyukkarabacak, B., and N. T. Valev. "The Role of Household
and Business Credit in Banking Crises." Journal of Banking &
Finance, 34(6), 2010, 1247-56.
Caporale, G. M., and P. Howells. "Money, Credit and Spending:
Drawing Causal Inferences." Scottish Journal of Political Economy,
48(5), 2001, 547-57.
Cecchetti, S., and E. Kharroubi. "Reassessing the Impact of
Finance on Growth." BIS Working Papers 381, Bank for International
Settlements, Basel, Switzerland, 2012.
--. "Why Does Financial Sector Growth Crowd Out Real Economic
Growth." Tech, rep., Bank for International Settlements, Basel,
Switzerland, 2013.
Dell'Ariccia, G., D. Igan, and L. Laeven. "Credit Booms
and Lending Standards: Evidence from the Subprime Mortgage Market,"
Journal of Money, Credit and Banking, 44(3), 2012,367-84.
Dembiermont, C., M. Drehmann, and S. Muksakunratana. "How Much
Does the Private Sector Really Borrow--A New Database for Total Credit
to the Private Non-financial Sector." BIS Quarterly Review, 17(1),
2013, 68-85.
Favara, G., and J. Imbs. "Credit Supply and the Price of
Housing." American Economic Review, 105(3), 2015, 958-92.
Furstenberg, G. M., and U. Kalckreuth. "Dependence on External
Finance: An Inherent Industry Characteristic?", Open Economies
Review, 17, 2006, 541-59.
Goldsmith, R. Financial Structure and Development. New Haven, CT:
Yale University Press, 1969.
Haltenhof, S., S. Lee, and V. Stebunovs. "The Credit Crunch
and Fall in Employment During the Great Recession". FEDS Working
Paper No. 201406. Ann Arbor, Michigan, 2014.
Jappelli, T., and M. Pagano. "Saving, Growth, and Liquidity
Constraints." Quarterly Journal of Economics, 109(1), 1994, 83-109.
Jappelli, T., M. Pagano, and M. di Maggio. "Households'
Indebtedness and Financial Fragility." CSEF Working Papers 208,
Centre for Studies in Economics and Finance (CSEF), University of
Naples, Italy, 2008.
Jimenez, G., S. Ongena, J.-L. Peydro, and J. Saurina. "Credit
Supply and Monetary Policy: Identifying the Bank Balance-Sheet Channel
with Loan Applications." American Economic Review, 102(5), 2012,
2301-26.
Jorda, O., M. Schularick, and A. M. Taylor. "When Credit Bites
Back." Journal of Money, Credit, and Banking, 45(2), 2013, 3-28.
--. "The Great Mortgaging: Housing Finance, Crises, and
Business Cycles." NBER Working Paper 20501, National Bureau of
Economic Research, Cambridge, MA, 2014.
King, R. G., and R. Levine. "Finance and Growth: Schumpeter
Might Be Right." Quarterly Journal of Economics, 108(3),
1993,717-37.
Krippner, G. R. "The Financialization of the American
Economy." Socio-Economic Review, 3(2), 2005, 173-208.
Kroszner, R. S., L. Laeven, and D. Klingebiel. "Banking
Crises, Financial Dependence, and Growth." Journal of Financial
Economics, 84(1), 2007, 187-228.
Laeven, L., and F. Valencia. "Systemic Banking Crises
Database." IMF Economic Review, 61(2), 2013, 225-70.
Levine, R. "Finance and Growth: Theory and Evidence," in
Handbook of Economic Growth, Vol. 1, Chapter 12, edited by P. Aghion and
S. Durlauf. Amsterdam: Elsevier, 2005, 865-934.
Lorenzoni, G. "Inefficient Credit Booms." Review of
Economic Studies, 75(3), 2008, 809-33.
Maddaloni, A., and J. L. Peydro. "Bank Risk-taking,
Securitization, Supervision, and Low Interest Rates: Evidence from the
Euro-area and the U.S. Lending Standards." Review of Financial
Studies, 24(6), 2011, 2121 -65.
Manganelli, S., and A. Popov. "Financial Dependence, Global
Growth Opportunities, and Growth Revisited." Economics Letters,
120(1), 2013, 123-25.
Masten, A. B., F. Coricelli, and I. Masten. "Non-linear Growth
Effects of Financial Development: Does Financial Integration
Matter?" Journal of International Money and Finance, 27(2), 2008,
295-313.
Mayer, C., K. Pence, and S. M. Sherlund. "The Rise in Mortgage
Defaults." Journal of Economic Perspectives, 23(1), 2009, 27-50.
McKinnon, R. Money and Capital in Economic Development. Washington,
DC: Brookings Institution, 1973.
Mian, A., and A. Sufi. House of Debt. Chicago: University of
Chicago Press, 2014.
Piketty, T. Capital in the 21st Century. Cambridge, MA: Harvard
University Press, 2014.
Pumanandam, A. "Originate-to-Distribute Model and the
Sub-prime Mortgage Crisis." Review of Financial Studies, 24, 2011,
1881-915.
Raddatz, C. "Liquidity Needs and Vulnerability to Financial
Underdevelopment." Journal of Financial Economics, 80(3), 2006,
677-722.
Radelet, S., and J. D. Sachs. "The East Asian Financial
Crisis: Diagnosis, Remedies, Prospects." Brookings Papers on
Economic Activity, 29(1), 1998, 1-90.
Rajan, R. G., and L. Zingales. "Financial Dependence and
Growth." American Economic Review, 88(3), 1998, 559-86.
Reinhart, C. M. "This Time Is Different Chartbook: Country
Histories on Debt, Default, and Financial Crises." NBER Working
Papers 15815, National Bureau of Economic Research, Inc., Cambridge, MA,
2010.
Rioja, F., and N. Valev. "Finance and the Sources of Growth at
Various Stages of Economic Development." Economic Inquiry, 42(1),
2004, 127-40.
Rousseau, P. L., and P. Wachtel. "What Is Happening to the
Impact of Financial Deepening on Economic Growth?" Economic
Inquiry, 49(1), 2011, 276-88.
Schularick, M., and A. M. Taylor. "Credit Booms Gone Bust:
Monetary Policy, Leverage Cycles, and Financial Crises, 1870-2008."
American Economic Review, 102(2), 2012, 1029-61.
Schumpeter, J. The Theory of Economic Development. Cambridge, MA:
Harvard University Press, 1934.
--. Business Cycles. Cambridge, MA: Harvard University Press, 1939.
Shaw, E. Financial Deepening and Economic Development. Oxford:
Oxford University Press, 1973.
Shen, C.-H., and C.-C. Lee. "Same Financial Development Yet
Different Economic Growth: Why?" Journal of Money, Credit, and
Banking, 38(7), 2006, 1907-44.
Stengos, T., A. Savvides, T. Mamuneas, and E. Ketteni. "Is the
Financial Development and Economic Growth Relationship Nonlinear?"
Economics Bulletin, 15(14), 2007, 1-12.
Stockhammer, E. "Financialisation and the Slowdown of
Accumulation." Cambridge Journal of Economics, 28(5), 2004,719-41.
Summers, L. "Speech at the 14th Annual IMF Research
Conference: Crises Yesterday and Today," November 8, 2013.
Sutherland, D., P. Hoeller, R. Merola, and V. Ziemann. "Debt
and Macroeconomic Stability." OECD Economics Department Working
Papers 1003, OECD Publishing, Paris, 2012.
Valickova, P, T. Havranek, and R. Horvath. "Financial
Development and Economic Growth: A Meta analysis." Working Papers
IES 2013/04, Charles University Prague, Faculty of Social Sciences,
Institute of Economic Studies, Prague, 2013.
Wachtel, P. "The Evolution of the Finance Growth Nexus."
Comparative Economic Studies, 53, 2011, 475-88.
Werner, R. A. "Towards a New Monetary Paradigm: A Quantity
Theorem of Disaggregated Credit, with Evidence from Japan." Kredit
und Kapital, 30(2), 1997, 276-309.
--. "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, 2012, 1-17.
Xu, Z. "Financial Development, Investment, and Economic
Growth." Economic Inquiry, 38(2), 2000, 331-44.
(1.) This literature builds on Schumpeter (1934, 1939), Goldsmith
(1969), McKinnon (1973), and Shaw (1973). Levine (2005) and Ang (2008)
provide overviews. The latest year analyzed in this literature is 2005.
(2.) In our dataset, there are seven countries for which we have
data going back substantially before 1990: Switzerland, Chile, Germany,
Hong Kong, Japan, Portugal, and the United States. We plot the
development of credit composition in these countries in Figures A1 and
A2 in the Appendix. In each of them, the stock of mortgage credit rose
faster than the stock of nonfinancial-business credit especially since
the 1990s and except for Hong Kong, this decline occurred mostly in the
2000s.
(3.) We provide a more detailed discussion in the Appendix.
(4.) We follow the traditional measure for credit stocks, which is
the summation of current and past credit flows (Equation (1) below).
Therefore the growth coefficient for credit stocks includes the growth
effect of current credit flows. In the analysis we will account for this
effect by including credit flows as a separate variable.
(5.) A fifth category is bank lending to government, which is
however often not reported and in any case mostly small.
(6.) Related data sets are in Beck et al. (2012) (which ends in
2005 and does not have 15 countries included in our data) and BIS (2013)
which includes both bonds and bank credit and does not differentiate
bank credit. We refer to the Appendix for a comparative discussion.
(7.) Three notes are in order. First, the present paper differs
from other studies which distinguish between credit into
"enterprise" and "household" credit (Beck et al.
2012; Buyukkarabacak and Krause 2009; Buyukkarabacak and Valev 2010). In
practice the difference is not a large one on average as credit stocks
to financial business and household consumption credit are both
relatively small. Second, we aggregated into two categories for reasons
of parsimony in presentation; alternative aggregations are possible but
do not qualitatively affect our results. For robustness purposes, we
also analyze growth effects of all four types of credit below. We will
find that the decisive distinction is between household mortgages and
nonfinancial business credit. Third, while this delineation is useful,
its measurement is necessarily imprecise. For instance, mortgage credit
often also serves as consumer credit through home equity withdrawals,
while business credit includes business mortgage credit. Conversely,
nonfinancial businesses realize part of their returns in trading
financial assets (see, e.g., Krippner 2005 on the United States).
(8.) We also computed results with the more common 5-year
intervals, which are available on request. We estimated FE and system
GMM models and for the latter we use two lags of the endogenous
variables as internal instruments. The loss of observations compared to
using 3-year periods is large, e.g. from 237 to 143 observations in the
first specification (column (1) in Table 3). Still, we are able to
replicate the qualitative findings. In some cases the results have
stronger significance, in other cases they weaken; but the signs of
coefficients are never the opposite of what we report here.
(9.) All country-level variables are taken from the World Bank
Development Indicators, except education (which is retrieved from the
Barro and Lee 2013 database) and institutional quality, extracted from
the International Country Risk Guide (ICRG) database.
(10.) See Arellano and Bover (1995) and Blundell and Bond (1998).
The GMM specification combines regressions in levels and in differences,
yielding unbiased estimators for the coefficients of interest.
(11.) In all specifications, two or three lags are used as
instruments. In all cases, the number of instruments is smaller than the
number of countries.
(12.) Rajan and Zingales (1998) utilize an industry-specific index
of external financial dependence, defined as capital expenditures minus
cash flow from operations divided by capital expenditures. They rank
industries by the median (U.S. Compustat) firm's external
dependence on finance and observe that industries that are more
dependent on external finance grow faster in countries with more
developed financial systems, measured as the credit-to-GDP ratio. By
exploiting cross-industry variations while controlling for a range of
country-specific and industry-specific factors, this widely used
methodology alleviates endogeneity concerns. Other studies support this
approach. 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 more financially developed countries. This result is
particularly significant for firms that are more likely to be
financially constrained and dependent on domestic financial markets,
such as small and young firms. Kroszner, Laeven, and Klingebiel (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.
(13.) As the industry-specific deflators are not available across a
large number of countries, we choose to deflate industry nominal value
added by the country-specific consumer price index (CPI), as in Braun
and Larrain (2005). Albeit imperfect, this provides a good approximation
for a wide range of countries in our sample.
(14.) We thank an anonymous referee for asking the question.
(15.) The calculation is: (-0.057*32.2)/2.475 = 0.74, where 32.2
and 2.475 are one standard deviation of asset-market credit stocks and
one standard deviation of the output growth rate, respectively.
(16.) The calculation is 0.189*4.24/2.475 = 0.32, where 4.24 and
2.475 are one standard deviation of nonfinancial credit flows and output
growth rate, respectively.
(17.) Alternatively, we characterize a 3-year episode as crisis if
the country was in crisis for at least 2 years during a 3-year period.
Our results are quantitatively similar.
(18.) An 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.
DIRK BEZEMER, MARIA GRYDAKI and LU ZHANG *
* For helpful comments on earlier drafts of this paper we thank the
editor Ola Olsson, two anonymous referees and seminar participants in
the June 2013 EWEPA conference in Helsinki, the September 2013 2nd IES
seminar in Prague, the October 2013 IMK conference in Berlin, the
October 2013 NED conference in Amsterdam, November 2013 research
seminars in Groningen and Stirling, and the April 2014 Royal Economic
Society conference in Manchester. Thanks especially to Joeri Schasfoort
and Anna Samarina for help in organizing the data. Any remaining errors
or omissions are ours. We acknowledge financial support from the
Equilibrio Foundation and the Institute for New Economic Thinking.
Bezemer: Associate Professor, Faculty of Economics and Business,
University of Groningen, 9747 AE Groningen, The Netherlands. Phone
31503633799, Fax 31503 637337, E-mail d.j.bezemer@rug.nl
Grydaki: Lecturer. Division of Economics, University of Stirling.
Stirling, FK9 4LA, UK. Phone 441786467476, Fax 441786467469, E-mail
maria.grydaki@stir.ac.uk
Zhang: Post-doctoral Researcher, Faculty of Economics and Business,
University of Groningen, 9747 AE Groningen, The Netherlands. Phone
31503637075, Fax 31503637337, E-mail lu.zhang@rug.n1
TABLE 1
Credit Stocks, Credit Flows, and Growth: Correlations
GDP p.c. Total Nonfinancial
Growth Credit Sector
Panel A: Stocks
GDP p.c. growth 1
Total credit -0.324 *** 1
Nonfinancial credit (a+b) -0.282 *** 0.827 *** 1
Asset market credit (c+d) -0.287 *** 0.917 *** 0.535 ***
a. Nonfinancial business -0.275 *** 0.786 *** 0.965 ***
b. Consumption -0.190 * 0.622 *** 0.710 ***
c. Mortgage -0.312 *** 0.903 *** 0.606 ***
d. Financial business -0.147 0.626 *** 0.231 ***
Panel B: Flows
GDP p.c. growth 1
Total credit 0.27 *** 1
Nonfinancial credit (a+b) 0.313 *** 0.802 *** 1
Asset market credit (c+d) 0.147 0.856 *** 0.377 ***
a. Nonfinancial business 0.273 *** 0.782 *** 0.952 ***
b. Consumption 0.282 0.440 *** 0.568 ***
c. Mortgage 0.104 0.748 *** 0.384 ***
d. Financial business 0.131 0.605 *** 0.203 *
Financial Nonfinancial Consumer
Sector Credit Credit
Panel A: Stocks
GDP p.c. growth
Total credit
Nonfinancial credit (a+b)
Asset market credit (c+d) 1
a. Nonfinancial business 0.497 *** 1
b. Consumption 0.432 *** 0.502 *** 1
c. Mortgage 0.928 *** 0.543 *** 0.542 ***
d. Financial business 0.777 *** 0.248" 0.097
Panel B: Flows
GDP p.c. growth
Total credit
Nonfinancial credit (a+b)
Asset market credit (c+d) 1
a. Nonfinancial business 0.373 *** 1
b. Consumption 0.19 * 0.35 *** 1
c. Mortgage 0.826 *** 0.356 *** 0.274 ***
d. Financial business 0.761 *** 0.228 ** 0.01
Financial
Mortgage Business
Credit Credit
Panel A: Stocks
GDP p.c. growth
Total credit
Nonfinancial credit (a+b)
Asset market credit (c+d)
a. Nonfinancial business
b. Consumption
c. Mortgage 1
d. Financial business 0.488 *** 1
Panel B: Flows
GDP p.c. growth
Total credit
Nonfinancial credit (a+b)
Asset market credit (c+d)
a. Nonfinancial business
b. Consumption
c. Mortgage 1
d. Financial business 0.263 *** 1
Note: This table reports pairwise correlation coefficients
between growth and different types of credit stocks and
flows, *** p< 0.01, ** p< 0.05, * p<0.1.
TABLE 2
Descriptive Statistics (3-year averaged data)
Variable Source Unit
Credit stocks
Total credit Own Calculation % of GDP
Credit aggregates
Nonfinancial credit Own Calculation % of GDP
Asset market credit Own Calculation % of GDP
Credit categories
Nonfinancial business Own Calculation % of GDP
Consumption Own Calculation % of GDP
Mortgage Own Calculation % of GDP
Financial business Own Calculation % of GDP
Credit flows
Total credit Own Calculation % of lagged GDP
Credit aggregates
Nonfinancial credit Own Calculation % of lagged GDP
Asset market credit Own Calculation % of lagged GDP
Credit categories
Nonfinancial business Own Calculation % of lagged GDP
Consumption Own Calculation % of lagged GDP
Mortgage Own Calculation % of lagged GDP
Financial business Own Calculation % of lagged GDP
Other variables
GDP per capita growth WDI Percentage points
Initial GDP per capita WDI In log
Trade openness WDI % of GDP
Government size WDI % of GDP
Inflation WDI Percentage points
Education Barro and Lee (2013) Years
Institution ICRG Index
Crisis Laeven and Valencia Dummy variable
(2013)
Variable Obs Mean Std Min Max
Credit stocks
Total credit 237 82.174 50.941 9.82 381.584
Credit aggregates
Nonfinancial credit 237 45.135 25.504 5.944 187.026
Asset market credit 228 38.501 32.206 0.245 194.559
Credit categories
Nonfinancial business 237 35.594 18.226 5.565 92.696
Consumption 206 10.976 12.458 0.221 94.33
Mortgage 228 30.273 27.386 0.245 194.559
Financial business 191 9.822 12.302 0.058 76.323
Credit flows
Total credit 228 7.406 7.517 -4.335 70.305
Credit aggregates
Nonfinancial credit 228 3.74 4.239 -4.612 32.055
Asset market credit 219 3.816 4.44 -2.931 38.249
Credit categories
Nonfinancial business 228 2.801 3.217 -4.771 16.767
Consumption 199 1.075 1.779 -1.803 15.288
Mortgage 219 2.976 3.617 -2.825 38.249
Financial business 183 1.006 2.327 -2.621 21.978
Other variables
GDP per capita growth 237 2.306 2.475 -7.602 12.629
Initial GDP per capita 237 9.323 1.095 6.142 10.913
Trade openness 237 94.554 76.798 15.546 424.013
Government size 237 17.81 4.805 7.197 28.413
Inflation 237 4.431 6.679 -3.123 66.008
Education 237 9.553 2.202 3.472 13.262
Institution 237 78.433 6.992 60.867 92.067
Crisis 237 0.11 0.313 0 1
Note: "Total credit" was computed only for country-year
observations where there was at least one nonzero
observation for nonfinancial credit and one observation for
asset market credit.
TABLE 3
Credit and Economic Growth: Stock and Flow Effects
FE
(1) (2) (3)
Total credit
Credit stocks -0.008 -0.013 *
(0.005) (0.007)
Credit flows 0.055 0.067
(0.040) (0.043)
Initial GDPPC -5.632 * -7.132 ** -6.210 **
(2.954) (2.750) (2.974)
Trade 0.012 0.014 * 0.011
(0.009) (0.008) (0.008)
Government -0.374 ** -0.361 ** -0.295
(0.163) (0.171) (0.182)
Inflation -0.102 -0.112 -0.105
(0.097) (0.099) (0.098)
Education 0.54 0.42 0.455
(0.515) (0.477) (0.485)
Institutions 0.182 *** 0.189 *** 0.167 ***
(0.064) (0.059) (0.059)
Time FE Yes Yes Yes
Observations 237 228 228
Number of id 46 46 46
[R.sup.2] 0.484 0.505 0.517
AR(2)
Overidentification
System GMM
(4) (5) (6)
Total credit
Credit stocks -0.02 -0.016 **
(0.014) (0.007)
Credit flows 0.085 0.071
(0.057) (0.054)
Initial GDPPC -2.618 ** -3.071 *** -2.271 ***
(1.122) (1.055) (0.808)
Trade 0.007 ** 0.006 * 0.006 **
(0.003) (0.003) (0.003)
Government 0.008 0.012 -0.004
(0.050) (0.060) (0.046)
Inflation -0.112 -0.114 -0.113
(0.094) (0.088) (0.087)
Education 0.249 0.293 * 0.199
(0.178) (0.167) (0.131)
Institutions 0.259 ** 0.225 ** 0.200 ***
(0.108) (0.104) (0.073)
Time FE Yes Yes Yes
Observations 237 228 228
Number of id 46 46 46
[R.sup.2]
AR(2) 0.485 0.651 0.617
Overidentification 0.403 0.383 0.346
Notes: This table presents the results using total credit
based on Equations (3) and (4). Columns (l)-(3) present the
FE results, columns (4)-(6) show the system-GMM results. The
dependent variable is the average growth rate of real GDP
per capita (constant 2005 US dollar) over each 3-year
period. Credit stocks and flows are defined as in Equations
(1) and (2). Initial GDPPC is real GDP per capita at the
beginning of each 3-year period. Trade is imports plus
exports, divided by GDP. Government is government
consumption divided by GDP. Inflation is the change in CPI.
Education is average years of schooling. Institutions is the
ICRG composite country risk measure. AR(2) is the Arellano-
Bond serial correlation test (we report the p value). Over-
identification is the Hansen J statistic (we report the p
value). All specifications include time dummies
(coefficients not reported). Coefficients for the constant
are not reported. Robust standard errors are in parentheses.
*** p<0.01, ** p<0.05, * p<0.1.
TABLE 4
Credit Aggregates and Economic Growth
FE
(1) (2) (3)
Nonfinancial Credit
Credit stocks -0.030 ** -0.039 **
(0.014) (0.016)
Credit flows 0.104 0.129 *
(0.064) (0.065)
Initial GDPPC -4.469 -7.044 ** -4.765
(3.114) (2.767) (3.220)
Trade 0.009 0.013 * 0.008
(0.009) (0.008) (0.008)
Government -0.363 ** -0.362 ** -0.285
(0.161) (0.170) (0.177)
Inflation -0.102 -0.113 -0.109
(0.095) (0.100) (0.097)
Education 0.665 0.429 0.614
(0.554) (0.476) (0.513)
Institution 0.177 *** 0.183 *** 0.158 **
(0.064) (0.059) (0.059)
Time FE Yes Yes Yes
Observations 237 228 228
Number of id 46 46 46
[R.sup.2] 0.495 0.506 0.53
AR(2)
Overidentification
FE
(4) (5) (6)
Asset Market Credit
Credit stocks 0.002 -0.005
(0.008) (0.009)
Credit flows 0.073 0.079
(0.053) (0.055)
Initial GDPPC -6.778 ** -7.507 ** -7.429 **
(2.896) (2.835) (2.908)
Trade 0.017 * 0.018 * 0.017 **
(0.009) (0.008) (0.009)
Government -0.419 ** -0.388 ** -0.373 *
(0.170) (0.173) (0.186)
Inflation -0.118 -0.118 -0.116
(0.102) (0.102) (0.103)
Education 0.281 0.21 0.209
(0.427) (0.425) (0.430)
Institution 0.187 *** 0.188 *** 0.183 ***
(0.067) (0.064) (0.063)
Time FE Yes Yes Yes
Observations 228 219 219
Number of id 44 44 44
[R.sup.2] 0.497 0.517 0.517
AR(2)
Overidentification
System GMM
(7) (8) (9)
Nonfinancial Credit
Credit stocks -0.026 -0.013
(0.043) (0.014)
Credit flows 0.208 ** 0.189 *
(0.092) (0.103)
Initial GDPPC -2.803 * -2.360 ** -2.685 ***
(1.472) (0.948) (0.790)
Trade 0.007 ** 0.005 ** 0.006 **
(0.003) (0.002) (0.003)
Government -0.003 0.008 0.015
(0.069) (0.050) (0.051)
Inflation -0.116 -0.114 -0.125
(0.093) (0.083) (0.089)
Education 0.27 0.246 * 0.268 *
(0.221) (0.143) (0.140)
Institution 0.231 * 0.170 * 0.212 ***
(0.122) (0.089) (0.074)
Time FE Yes Yes Yes
Observations 237 228 228
Number of id 46 46 46
[R.sup.2]
AR(2) 0.415 0.807 0.748
Overidentification 0.408 0.331 0.598
System GMM
(10) (11) (12)
Asset Market Credit
Credit stocks -0.032 * -0.057 **
(0.016) (0.027)
Credit flows 0.018 0.052
(0.059) (0.097)
Initial GDPPC -1.582 ** -2.869 *** -1.587 *
(0.613) (0.879) (0.812)
Trade 0.006 *** 0.007 ** 0.006 *
(0.002) (0.003) (0.003)
Government -0.01 0.006 0.002
(0.041) (0.052) (0.056)
Inflation -0.093 -0.111 -0.091
(0.081) (0.091) (0.087)
Education 0.119 0.231 0.107
(0.099) (0.146) (0.129)
Institution 0.161 ** 0.201 ** 0.226 ***
(0.078) (0.081) (0.072)
Time FE Yes Yes Yes
Observations 228 219 213
Number of id 44 44 43
[R.sup.2]
AR(2) 0.569 0.707 0.804
Overidentification 0.627 0.386 0.607
Notes: This table presents the results using nonfinancial
credit and asset market credit based on Equations (3) and
(4), respectively. Columns (1)-(3) present the FE results,
columns (4)-(6) show the system-GMM results. The dependent
variable is the average growth rate of real GDP per capita
(constant 2005 US dollar) over each 3-year period. Credit
stocks and flows are defined as in Equations (1) and (2).
Initial GDPPC is real GDP per capita at the beginning of
each 3-year period. Trade is imports plus exports, divided
by GDP. Government is government consumption divided by GDP.
Inflation is the change in CPI. Education is average years
of schooling. Institutions is the ICRG composite country
risk measure. AR(2) is the Arellano-Bond serial correlation
test (we report the p value). Overidentification is the
Hansen J statistic (we report the p value). All
specifications include time dummies (coefficients not
reported). Coefficients for the constant are not reported.
Robust standard errors are in parentheses.
*** p<0.01, ** p<0.05, * p<0.1.
TABLE 5
Credit and Economic Growth: Industry-Level Results
(1) (2) (3)
Total credit
Credit stocks -0.030 ** -0.044 ***
(0.013) (0.014)
ED * credit stocks 0.053 *** 0.054 ***
(0.019) (0.021)
Credit flows 0.134 ** 0.175 ***
(0.053) (0.056)
ED * credit flows 0.016 -0.021
(0.090) (0.095)
Initial share 0.473 * 0.543 ** 0.539 **
(0.260) (0.261) (0.261)
Observations 5,415 5,182 5,182
Number of countries 41 41 41
[R.sup.2] 0.447 0.457 0.459
Marginal effects of credit stocks
for high dependence -0.006 -0.019
industry
for low dependence -0.026 -0.039
industry
Implied differential effect 0.02 0.02
(4) (5) (6)
Nonfinancial credit
Credit stocks -0.034 -0.064 **
(0.027) (0.028)
ED * credit stocks 0.089 ** 0.099 **
(0.040) (0.042)
Credit flows 0.387 *** 0.429 ***
(0.121) (0.124)
ED * credit flows -0.007 -0.062
(0.186) (0.192)
Initial share 0.474 * 0.537 ** 0.534 **
(0.259) (0.260) (0.260)
Observations 5,415 5,182 5,182
Number of countries 41 41 41
[R.sup.2] 0.446 0.459 0.46
Marginal effects of credit stocks
for high dependence 0.007 -0.018
industry
for low dependence -0.027 -0.056
industry
Implied differential effect 0.02 0.038
(7) (8) (9)
Asset market credit
Credit stocks -0.069 *** -0.084 ***
(0.021) (0.024)
ED * credit stocks 0.095 *** 0.094 **
(0.036) (0.040)
Credit flows 0.067 0.154 *
(0.077) (0.089)
ED * credit flows 0.056 -0.022
(0.154) (0.164)
Initial share 0.234 0.303 0.3
(0.226) (0.226) (0.228)
Observations 5,306 5,073 5,073
Number of countries 41 41 41
[R.sup.2] 0.417 0.425 0.427
Marginal effects of credit stocks
for high dependence -0.025 -0.041
industry
for low dependence -0.061 -0.076
industry
Implied differential effect 0.036 0.035
Notes: This table presents the industry-level evidence based
on Equation (5). Columns (1)-(3) presents the results for
total credit, columns (4)-(6) for "nonfinancial credit" (the
sum of nonfinancial business and consumption credit) and
columns (7)-(9) for "asset market credit" (the sum of
financial business and mortgage credit). The dependent
variable is the average growth rate of real value added over
each 3-year period. Credit stocks and flows are defined as
in Equations (1) and (2). ED is external dependence on
finance, taken from Rajan and Zingales (1998). Initial share
is the share of each industry in a country's s total
manufacturing value added at the beginning of each 3-year
period. All estimations include a constant and country,
year, industry, industry-year, and industry-country dummies
(coefficients not reported). Country-time controls include
initial GDP per capita at the beginning of each 3-year
period, trade openness, government spending, inflation,
education and institution, as in the country-level
regressionsin table 4. The last three rows show the marginal
growth effect of credit stocks 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.
TABLE 6
Credit and Economic Growth: Stock and Flow Interaction Effects
FE
(1) (2) (3)
Total credit Nonfinancial Asset market
Credit stocks -0.005 -0.028 0.012
(0.007) (0.017) (0.009)
Credit flows 0.184 *** 0.299 *** 0.304 ***
(0.041) (0.070) (0.085)
Stocks * flows -0.001 *** -0.003 ** -0.002 **
(0.0002) (0.001) (0.001)
Initial GDPPC -6.787 ** -5.089 -8.049 ***
(2.850) (3.232) (2.678)
Trade 0.013 0.009 0.020 **
(0.008) (0.008) (0.009)
Government -0.256 -0.26 -0.348 *
(0.177) (0.176) (0.178)
Inflation -0.114 -0.117 -0.122
(0.095) (0.096) (0.097)
Education 0.499 0.648 0.155
(0.453) (0.493) (0.385)
Institutions 0.157 *** 0.146 ** 0.180 ***
(0.054) (0.055) (0.059)
Observations 228 228 219
Number of id 46 46 44
[R.sup.2] 0.548 0.548 0.552
AR(2)
Overidentification
System-GMM
(4) (5) (6)
Total credit Nonfinancial Asset market
Credit stocks 0.01 0.011 0.004
(0.010) (0.024) (0.021)
Credit flows 0.141 0.239 * 0.124
(0.085) (0.131) (0.242)
Stocks * flows -0.0002 -0.001 * -0.0003
(0.0003) (0.0008) (0.002)
Initial GDPPC -4.124 ** -3.817 *** -2.938 **
(1.624) (1.417) (1.233)
Trade 0.006 0.006 0.006' *
(0.004) (0.004) (0.003)
Government 0.045 0.048 0.007
(0.083) (0.081) (0.055)
Inflation -0.127 -0.131 -0.111
(0.102) (0.102) (0.093)
Education 0.402 0.393 * 0.231
(0.248) (0.226) (0.177)
Institutions 0.304 ** 0.311 ** 0.198 *
(0.149) (0.129) (0.099)
Observations 228 228 219
Number of id 46 46 44
[R.sup.2]
AR(2) 0.979 0.962 0.892
Overidentification 0.288 0.483 0.415
Notes: This table reports results including the interactions
of credit stocks and flows. Columns (1) and (4) use total
credit, whereas columns (2), (5) and (3), (6) use
"nonfinancial credit" (the sum of nonfinancial business and
consumption credit) and "asset market credit" (the sum of
financial business and mortgage credit), respectively.
Columns (1)-(3) present the FE results, columns (4)-(6) show
the system-GMM results. The dependent variable is the
average growth rate of real GDP per capita (constant 2005 US
dollar) over each 3-year period. Credit stocks and flows are
defined as in Equations (1) and (2). Initial GDPPC is the
real GDP per capita at the beginning of each 3-year period.
Trade is imports plus exports, divided by GDP. Government is
government consumption divided by GDP. Inflation is the
change in CPI. Education is average years of schooling.
Institutions is the ICRG composite country risk measure.
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 constants
and time dummies (coefficients are not reported). Robust
standard errors in parentheses.
*** p <0.01, ** p <0.05, * p <0.1.
TABLE 7
Robustness Analyses
(la) (2a) (3a) (4a)
The Four Categories of Credit
Panel A: FE NonFin Consumer Mortgage FinBus
Credit stocks -0.060 *** -0.059 -0.005 -0.004
(0.017) (0.040) (0.009) (0.028)
Credit flows 0.198 *** 0.076 0.067 0.099 *
(0.069) (0.151) (0.076) (0.050)
Crisis
Crisis * Credit stocks
Time FE Yes Yes Yes Yes
Observations 228 199 219 183
Number of id 46 39 44 37
[R.sup.2] 0.542 0.48 0.512 0.506
(5a) (6a) (7a)
Country Outliers
Panel A: FE NF AM NF_HD
Credit stocks -0.070 *** -0.009 -0.032 *
(0.018) (0.013) (0.016)
Credit flows 0.223 *** 0.169 *** 0.103 *
(0.054) (0.056) (0.057)
Crisis
Crisis * Credit stocks
Time FE Yes Yes Yes
Observations 204 197 173
Number of id 40 39 33
[R.sup.2] 0.556 0.522 0.632
(8a) (9a) (10a)
Level of Credit Stocks
Panel A: FE NF_LD AMJHD AMJLD
Credit stocks -0.070 ** 0.004 0.001
(0.026) (0.009) (0.023)
Credit flows 0.251 *** 0.036 0.264 **
(0.060) (0.039) (0.099)
Crisis
Crisis * Credit stocks
Time FE Yes Yes Yes
Observations 176 166 167
Number of id 37 32 35
[R.sup.2] 0.573 0.64 0.555
(11a) (12a) (13a) (14a)
The Role of Crisis
Panel A: FE NF AM NF AM
Credit stocks -0.067 *** -0.023 -0.062 *** 0.014
-0.018 -0.023 (0.014) (0.013)
Credit flows 0.236 *** 0.105 0.174 *** 0.101 *
-0.063 -0.087 (0.048) (0.054)
Crisis -1.117 -0.385
(0.803) (0.809)
Crisis * Credit stocks 0.031 ** 0.011
(0.013) (0.010)
Time FE Yes Yes Yes Yes
Observations 182 175 228 219
Number of id 46 44 46 44
[R.sup.2] 0.392 0.308 0.543 0.52
(lb) (2b) (3b) (4b)
Panel B: System-GMM NonFin Consumer Mortgage FinBus
Credit stocks -0.017 -0.057 -0.033 * -0.041
(0.019) (0.041) (0.019) (0.048)
Credit flows 0.273 *** 0.096 0.023 -0.139
(0.097) (0.133) (0.047) (0.206)
Crisis
Crisis * Credit stocks
Observations 228 199 219 183
Number of id 46 39 44 37
AR(2) 0.696 0.909 0.755 0.766
Overidentification 0.515 0.579 0.57 0.19
(5b) (6b) (7b)
Panel B: System-GMM NF AM NF_HD
Credit stocks -0.054 -0.015 -0.023
(0.036) (0.021) (0.034)
Credit flows 0.203 ** 0.004 0.058
(0.090) (0.073) (0.079)
Crisis
Crisis * Credit stocks
Observations 204 197 173
Number of id 40 39 33
AR(2) 0.931 0.851 0.268
Overidentification 0.384 0.448 0.211
(8b) (9b) (10b)
Panel B: System-GMM NFJLD AM_HD AM_LD
Credit stocks -0.031 -0.007 -0.054
(0.027) (0.013) (0.044)
Credit flows 0.211 ** -0.021 -0.013
(0.086) (0.051) (0.241)
Crisis
Crisis * Credit stocks
Observations 176 166 167
Number of id 37 32 35
AR(2) 0.875 0.187 0.821
Overidentification 0.459 0.356 0.536
(11b) (12b) (13b) (14b)
Panel B: System-GMM NF AM NF AM
Credit stocks -0.035 -0.056 * -0.068 ** -0.035
-0.023 -0.029 (0.027) (0.023)
Credit flows 0.332 *** 0.112 0.165' * 0.109
-0.088 -0.188 (0.08) (0.078)
Crisis -2.438 * -1.246
(1.275) (1.012)
Crisis * Credit stocks 0.055 ** 0.030 *
(0.022) (0.018)
Observations 182 175 228 219
Number of id 46 44 46 44
AR(2) 0.793 0.641 0.88 0.984
Overidentification 0.527 0.492 0.313 0.45
Notes: This table presents the robustness analyses. Panel A
reports FE results, Panel B reports corresponding system-
GMM results. NF denotes "nonfinancial credit" (the sum of
nonfinancial business and consumption credit) and AM denotes
"asset market credit" (the sum of financial business and
mortgage credit). HD denotes the "high-credit-stocks"
subsample, whereas LD denotes the "low-credit-stocks"
subsample. Columns (la)-(4a) and (lb)-(4b) examine the
relations between each of the four underlying credit
categories, namely nonfinancial business, consumer,
mortgage, and financial business credit and growth. Columns
(5a)-(6a) and (5b)-(6b) drop outlier countries with high-
credit stocks-low growth (i.e., Denmark, Spain, and
Switzerland) and low credit levels-high growth (i.e.,
Armenia, India, and Uruguay). Columns (7a)-(10a) and (7b)-(1
Ob) consider the differential credit-growth relations in
countries with high and low levels of credit stocks. Columns
(1 la)-(14a) and (11 b)-(14b) examine the role of crisis.
The dependent variable is the average growth rate of real
GDP per capita (constant 2005 US dollar) over each 3-year
period. Credit stocks and flows are defined as in Equations
(1) and (2). Initial GDPPC is real GDP per capita at the
beginning of each 3-year period. Trade is imports plus
exports, divided by GDP. Government is government
consumption divided by GDP. Inflation is the change in CPI.
Education is average years of schooling. Institutions is the
ICRG composite country risk measure. Crisis is a dummy
variable that takes the value of one if a country was in
crisis for at least 1 year during a 3-year period, and zero
otherwise. The identification of crisis is based on Laeven
and Valencia (2013). All specifications include control
variables, constants, and time dummies (coefficients not
reported). AR(2) is the Arellano-Bond serial correlation
test (p value is reported). Overidentification is the Hansen
J statistic (p value is reported). Robust standard errors in
parentheses.
*** p < 0.01, ** p < 0.05, * p < 0.1.
COPYRIGHT 2016 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.
