n?u=RePEc:bog:wpaper:200&r=mac

repec.nep.mac

n?u=RePEc:bog:wpaper:200&r=mac

BANK OF GREECE

EUROSYSTEM

Working Paper

Credit-less recoveries: the role of

investment-savings imbalances

Hiona Balfoussia

Dimitris Malliaropulos

WORKINGPAPERWORKINGPAPERWORKINGPAPERWORKINGPAPERWORKI

NOVEMBER 2015

200


BANK OF GREECE

Economic Analysis and Research Department – Special Studies Division

21, Ε. Venizelos Avenue

GR-102 50 Athens

Τel: +30210-320 3610

Fax: +30210-320 2432

www.bankofgreece.gr

Printed in Athens, Greece

at the Bank of Greece Printing Works.

All rights reserved. Reproduction for educational and

non-commercial purposes is permitted provided that the source is acknowledged.

ISSN 1109-6691


CREDIT-LESS RECOVERIES: THE ROLE

OF INVESTMENT-SAVINGS IMBALANCES

Hiona Balfoussia

Bank of Greece

Dimitris Malliaropulos

Bank of Greece and University of Piraeus

Abstract

This paper argues that the investment–savings imbalances of households and

companies play an important role in determining the probability that an economy

experiences a credit-less recovery, following a recession. The investment–savings gap

determines the need for “external” finance of the private sector in the form of either

bank credit or capital market financing. Using a broad dataset covering 96 countries

and 272 recovery episodes, we provide empirical evidence that credit-less recoveries

are indeed associated with both low and declining financing needs of the private

sector, as proxied by the investment-savings gap at the trough of the recession and its

adjustment during the downturn. We show that this reflects a rebalancing of wealth

towards financial assets during the downturn which can subsequently be used to

finance real investment during the recovery stage, even in the absence of positive

bank credit flows. Lastly, we provide empirical evidence that, controlling for the

change in investment-savings imbalances, economies whose economic downturn was

preceded by a credit boom are more likely to experience a credit-less recovery.

JEL classification: E22, E32, E44, G01

Keywords: Credit-less recovery, crisis, savings-investment imbalance, financing

needs, flow of funds

Acknowledgments: We are most grateful to N. Sugawara and J. Zalduendo for making

their dataset available for use in the present research. We would also like to thank

Heather Gibson and the participants of a Bank of Greece seminar for their useful

comments. The views expressed in this paper are those of the authors and not

necessarily those of either the Bank of Greece or the Eurosystem.

Correspondence:

Dimitris Malliaropulos

Economic Analysis and Research Department,

Bank of Greece,

21 El. Venizelos Av.,

10250 Athens, Greece

Tel.:0030-210-3202380

Fax: 0030-210-3203939

Email: dmalliaropulos@bankofgreece.gr


1. Introduction

Credit-less recoveries have come into the mainstream of economic research

following the global financial crisis of 2008-2009 and the concurrent recession. The

related literature expanded rapidly in just a few years. Calvo et. al (2006) had set the

stage, coining the expression “phoenix miracle” to convey the puzzle inherently

associated with a rebound of economic activity in the absence of bank credit. Notable

more recent contributions include Claessens et al. (2009), Abiad et al. (2011),

Bijsterbosch and Dahlhaus (2011) and Sugawara and Zalduendo (2013), all of which

empirically explore the nature of credit-less recoveries and their possible

determinants.

Although a large battery of variables is considered in the aforementioned

literature, an aspect which has arguably been overlooked is the role of investmentsavings

imbalances in determining the probability of credit-less recoveries. From a

flow of funds perspective, imbalances between investment and savings of the private

non-financial sector reflect its needs for net external financing, i.e. financing from

sources other than private savings, such as bank credit, equity and bond market

financing. Thus, the lower the investment-savings gap, the lower the net external

financing needs of the private sector, and hence the more likely an economic recovery

without bank credit should be. From the perspective of national accounting, the

investment-savings gap of the private sector reflects the difference between the

current account deficit and the budget deficit of the government. The individual role

of large current account deficits and fiscal deficits in triggering economic crises has

been extensively studied in the literature. We reconsider the role of the current

account deficit in determining the probability of credit-less recoveries and argue that

the economic intuition commonly provided in the literature on this matter may be

misguided. We link the aforementioned two perspectives and reinterpret the role of

the two deficits in the occurrence of credit-less recoveries, from the prism of the

investment-savings imbalances and flow of funds approach.

Indeed, drawing on the large dataset employed in this paper, a substantial share

of credit-less recoveries seem to be associated with a low or declining investmentsavings

gap of the private sector. We use existing theoretical concepts to formalize the

hypothesis that such a decline increases the probability of the subsequent recovery

being credit-less. We formally test this hypothesis in a number of alternative

3


specifications and sample sizes and present empirical evidence in its favour. Next, we

explore how these changes in the investment-savings gap are actually financed, in

terms of flows of funds. We find that the decline in the investment-savings gap during

economic downturns which are followed by credit-less recoveries reflects an

increased accumulation of financial assets by households and companies. These

financial assets can subsequently be used to finance real investment (and thus the

rebound of economic activity) during the recovery stage, even in the absence of

positive credit flows. Finally, we provide empirical evidence that economies whose

downturn was preceded by a credit boom are more likely to experience a credit-less

recovery.

The remainder of the paper is structured as follows: Section 2 concisely presents

some stylized facts on credit-less recoveries and motivates our alternative perspective

on their possible determinants. Section 3 derives a proxy for the main variable of

interest, i.e. the private sector’s external financing needs, and outlines the dataset and

econometric methodology. Section 4 presents our empirical findings. Section 5

provides some robustness checks of our specification and findings. We conclude in

Section 6, where we set out some policy implications of our results.

2. Credit-less recoveries from a different angle

2.1 A review of the literature

Several empirical findings are already emerging as stylized facts in this strand

of the academic literature. In particular, a country is likelier to experience a credit-less

recovery when:

the preceding recession was deep, a large decline in output thought to imply

unused existing production capacity which can be readily tapped into to spur a

rebound of growth, even in the absence of credit expansion (Abiad et al., 2011

and Bijsterbosch and Dahlhaus, 2011, inter alia).

the recovery was preceded by a credit boom and/or a financial crisis (Abiad et

al., 2011 and Bijsterbosch and Dahlhaus, 2011)

a substantial REER devaluation occurred during the downturn, the intuition

being that a pick-up in exports may drive an economy out of recession without

much need for domestic credit (Darvas, 2013, inter alia).

4


a sharp improvement in the CA balance was recorded during the downturn,

thought to suggest a substantial restructuring of the economy towards

tradeable sectors, reflected in an increase in exports (Sugawara and

Zalduendo, 2013 and Darvas, 2013).

it is financially open, indicating a higher exposure to rapid capital outflows

(Sugawara and Zalduendo, 2013).

Additionally, Sugawara and Zalduendo (2013) were the first to systematically

consider policy variables, 1 and find that the probability of a credit-less recovery is

higher if, during the downturn:

there was an easing of fiscal policy, as the ensuing expansion would be driven,

in part, by public expenditure, implying a lower demand for bank credit during

the recovery phase

there was a tightening of monetary policy, as this would imply less funding

would be available to the private sector.

2.2 Why the economy’s financing needs may matter

The main theoretical premise of this paper stems from the savings-investment

imbalance concept. Investment is a key variable underpinning economic growth, as is

routinely reiterated in the reports of the World Bank and the IMF. Indeed, Ottonello

(2010) rightly points out that recoveries should be viewed as "phoenix miracles" only

to the extent that investment, their underlying driving force, is not contemporaneously

rebounding. The mirror image of investment is its financing, i.e. savings in the broad

sense. It is these two variables that we focus on in the present exploration of creditless

recoveries, within the savings-investment imbalance context.

Taking a broad view of conventional national accounting identities, the gap

between private investment and private savings, I-S, is thought to reflect the external

financing needs of an economy’s private sector. If I-S is positive, the private sector

needs external financing as its savings are not sufficient to cover investment.

Conversely, if S is sufficiently large to cover I, there is no need for external financing

1 Sugawara and Zalduendo (2013) also provide a thorough review of the theoretical and empirical

literature on credit-less recoveries, covering approaches from the micro-sectoral level to the broad

macro perspective.

5


of economic agents’ activities. By implication, as S increases, the need for credit

expansion to underpin economic growth should become less pressing.

In order to get some insight into the role of bank credit in financing private

investment, consider the aggregated flow of funds identity of households (HHs) and

non-financial companies (NFCs):

I+ΔFA=S+ΔL (1)

Equation (1) says that private sector investment, i.e. the sum of real investment,

I, and financial investment (the change in financial assets), ΔFA, is financed through

private savings, S, and bank credit, ΔL. 2 Hence, I-S equals the change in the net

financial liabilities of HHs and NFCs, ΔL-ΔFA, where ΔL denotes the change in

loans to the private sector and ΔFA the change in the private sector’s financial asset

holdings. This identity has the following implication: holding the rate of

accumulation/depletion of financial assets, ΔFA, constant, the higher the financing

needs of the private sector, I-S, the higher its dependence on bank credit. Conversely,

if private savings exceed private investment (I-S


declining investment-savings gap, Δ(I-S)


International Financial Statistics database and, secondarily, from the OECD’s

Quarterly National Accounts and the IMF’s World Economic Outlook database. 3

In line with the literature, economic troughs in SZ are identified from the

cyclical component of GDP and are defined as the points which fall one standard

deviation below trend, spaced at least 8 quarters apart. 4 A total of 272 recovery

episodes are thus identified in the sample, of which 105 in advanced economies and

113 in emerging economies, with 54 categorised as “other”. A credit-less recovery is

defined as one where average negative real credit growth was experienced over the 8

quarters following the trough. By this measure, roughly one in four economic

recovery episodes can be thought of as credit-less.

For our empirical application, we proxy the investment-savings gap of the

private sector (henceforth denoted φ) as the difference between the current account

deficit and the fiscal deficit (both as a fraction of GDP), using the national accounts’

identity: 5

I-S ≡ current account deficit – fiscal deficit,

This identity forms the basis of the “twin deficits” literature (e.g. Corsetti and

Müller, 2006). The underlying intuition is that an excess of private investment over

private savings can be financed either by a higher current account deficit (i.e. capital

inflows from abroad) or by a lower budget deficit (i.e. freeing up private savings to

finance investment). 6

Guided by the above identity, we use data from the SZ dataset on the “twin

deficits” to proxy the investment-savings gap of the private sector. Note that the fiscal

deficit in the SZ dataset is cyclically adjusted, hence, our proxy is by construction less

than perfectly correlated with the actual investment-savings gap of the private sector.

Moreover, differences in methodology between national accounts data and flow of

funds data make the proxy naturally imperfect. However, we evaluated the ability of

our proxy to track the investment-savings gap of the private sector for a number of

3 For a complete description of the dataset see Sugawara and Zalduendo (2013).

4 The cyclical component of GDP is defined as the difference between the log of real GDP and trend

GDP, the latter computed using a Hodrick-Prescott filter.

5 This is because flow of funds data are not available for most countries in our dataset.

6 An alternative interpretation, which is more standard within the “twin deficits” context, is that budget

deficits must be financed by either an increase in private sector savings or a reduction in investment or

an increase in the current account deficit (i.e. by higher capital inflows from abroad), thus the link

between the two deficits.

8


European economies where flow of funds data are available, and find that it performs

well. 7 The estimations presented in Section 4 are probit regressions, specified as

follows:

The binary dependent variable takes the value 0 if a recovery is credit-supported

and 1 if it is credit-less. The function is the cumulative normal distribution and

is a vector of regressors that drives the incidence of credit-less events. We use the SZ

specifications as a benchmark, on the basis of which we proceed to test a series of

sequentially nested hypotheses stemming from our main theoretical thesis. Thus, our

estimates are directly comparable to the findings of SZ and extend them in several

ways.

Our main ex ante expectation is that our proxy φ for the external financing needs of

the private sector, which we include in alternative forms in the vector of regressors

, has explanatory power for the probability of a credit-less recovery. In line with

the literature, a number of additional variables have been included in , in order to

gain further insight or to confirm the aforementioned stylized facts. These are:

GDP growth (trough minus peak): the depth of the recession, measured as the peakto-trough

percentage change in real GDP.

R.E.E.R. (trough minus peak): the real exchange rate adjustment, measured as the

peak-to-trough percentage change in the IMF’s real effective exchange rate.

Exports (%GDP, peak): the country’s openness, measured as the share of exports in

GDP (both goods and services) at peak.

Capital Account Openness (peak): the country’s openness to capital flows at peak,

based on the Chinn and Ito (2008) index.

7 We evaluated the ability of our proxy to track the investment-savings gap of the private sector for

Germany, France, Portugal, Finland, Denmark, UK, Spain, Netherlands, Italy, Sweden and Greece

(data are available by the authors on request). In many cases there seems to be a roughly constant

difference between the two series, which can be accounted for by differences in methodology between

national account data and flow of fund data, and the fact that we use cyclically adjusted data for the

fiscal deficit. However, the proxy seems particularly good at capturing turning points, which is what is

essential for our regression specifications, given that they primarily involve peak-to-trough changes.

9


Fiscal Policy (easing, trough minus peak): the stance of fiscal policy, measured as the

change (trough minus peak) of the cyclically adjusted fiscal deficit as a percent

of GDP.

Current Account (%GDP, trough minus peak): the change (trough minus peak) of the

current account balance as a percent of GDP.

Monetary Policy (easing, trough minus peak): the stance of monetary policy,

measured as the change (trough minus peak) of the velocity of broad money

(broad money divided by GDP).

IMF program (dummy, trough): a dummy variable indicating whether the country was

under an IMF programme (i.e. making use of GRA credit) at trough.

Public Debt (%GDP, trough): the public debt to GDP ratio at trough.

Inflation (y-o-y, trough): the year-on-year rate of inflation at trough.

Real credit growth (trough minus peak): the peak-to-trough change in real credit

growth

Loans to GDP (trough minus peak): the peak-to-trough change in the ratio of private

sector credit to GDP

It is important to underline that, while our dependent variable is timed posttrough,

all explanatory variables are timed either at the trough or pre-trough (i.e.

peak-to-trough differences or values at peak). Thus, in all of our specifications, the

nature of the recovery is conditioned exclusively on information available before its

onset. This is, in a sense, a purist approach, given that much of the literature also

makes use of contemporaneous information on the recovery itself. Nonetheless, this

framework allows greater confidence in the empirical findings we present, as it

alleviates any question of endogeneity.

4. Our empirical findings

Our empirical findings are presented in Tables 2 – 6, in a progressive sequence

of tests. Table 2 presents a number of specifications which include our proxy for the

level of financing needs as a share of GDP at the trough of the recession, denoted φ t ,

along with various combinations of other variables which have been found significant

in determining the probability of a credit-less recovery. Estimation 1 corresponds to

the SZ benchmark specification. Estimation 2 corresponds to the SZ “Fiscal

Monetary” specification, augmented by φ t . Estimations 3 and 4 correspond to the SZ

10


“Economic Conditions 1 and 2” specifications, augmented not only by φ t but also by

the fiscal policy easing variable, as it is key to our analysis. Estimation 5 corresponds

to the most encompassing SZ specification, entitled “Economic Policies”. The

coefficient on φ t is estimated to be negative and is significant in most of the

specifications. Moreover, in terms of R 2 , these specifications compare favourably with

the corresponding SZ ones. In line with our theoretical hypothesis, the size of

financing needs as a share of GDP at the bottom of the downturn appears to be

important to the economy’s probability of a credit-less recovery. More specifically,

economies which, at the trough of a recession, have lower financing needs, appear to

have a higher probability to rebound without bank credit.

The fact that the peak-to-trough change in the current account deficit and the

fiscal deficit are also included in estimations 1-5 allows us to re-interpret the SZ

findings regarding these two variables from the prism of our theoretical premise.

Recall that, as discussed in Section 3, the difference between the two deficit variables

is, by definition, equal to the peak-to-trough change in our proxy φ. Hence, Table 2

provides insights as to which of the two deficits (fiscal or external) contributes more

to financing the change in the savings-investment gap of the private sector during the

downturn, when it is followed by a credit-less recovery.

Notably, the coefficient on fiscal easing peak-to-trough is positive and highly

significant throughout, echoing the SZ finding that expansionary fiscal policy during

the downturn increases the probability of a credit-less recovery, the recovery being

instead driven, presumably, by public expenditure. This is in line with the argument of

subsection 3.3 that an excess of private investment over private savings can be

financed by a lower budget deficit (i.e. by freeing up private savings to finance

investment). The inverse (i.e. I-S


The other explanatory variables have, by and large, the same sign and level of

significance as in SZ and others: the probability of a credit-less recovery is higher the

greater the peak-to-trough decline in real GDP, the lower the economy’s trade

openness and the higher its financial openness at peak, while tighter monetary policy

peak-to-trough, higher inflation, higher indebtedness and the absence of an IMF

program at trough also have the same effect.

In Table 3, we decompose the level of financing needs at trough φ t into the sum

of the level at the preceding peak and its peak-to-trough change, which we denote φ p

and Δ t-p φ, respectively. This decomposition allows us to assess whether the

adjustment of the investment-savings gap during the downturn plays an important role

in determining the probability of a credit-less recovery, over and above its initial

level. 9 This yields a more intuitive specification, as the probability of a recovery being

credit-less is essentially conditioned on the nature and characteristics of the downturn,

and the situation of the economy before it, abstracting from the exact conditions at the

trough. We test, in particular: i) whether a decline in a country’s external financing

needs of the private sector from peak to trough (i.e. a negative value of Δ t-p φ) is

associated with a statistically significant increase in the probability of a credit-less

recovery, and ii) whether the size of the investment-savings imbalance prior to the

onset of the recession affects the probability that the recovery may be credit-less. On

the basis of the estimates presented in Table 3, it seems that both the initial level and

the peak-to-trough change of φ significantly affect the probability that an economy

experiences a credit-less recovery, the latter appearing significant irrespective of the

exact specification or sample size. This probability is higher, the lower the private

sector’s initial financing needs and the greater their decline during the downturn. In

economic terms, our estimates suggest that the change of φ during the downturn is

roughly twice as important as the initial level of φ. Interestingly, the most

encompassing model specification (estimation 5) suggests that the probability of a

creditless recovery is affected by the change in φ during the downturn independently

improvement in this variable may be thought to suggest a substantial rebalancing of the economy

towards exports (see inter alia SZ and Darvas, 2013). However, this argument is not very plausible, as

any substantial restructuring towards tradable sectors of the economy would require considerable time

to materialize and bear fruit in the form of higher export performance. In fact, it is all too common that

a current account improvement during a recession reflects not an increase in exports but rather a

decline in imports as a result of curtailed domestic demand.

9 We take care to adjust the vector of regressors appropriately, so as to avoid potential collinearity

problems. In this case, the peak-to-trough changes in the current account and fiscal deficit have been

excluded, as Δ t-p φ is a linear combination of the two.

12


of its initial level. Table 4 presents the corresponding estimates, including only the

peak-to-trough change in the private sector’s financing needs, i.e. adhering to the

approach adopted for the other macroeconomic variables and the standard

specification in SZ and much of the related literature which focuses exclusively on

peak-to-trough developments as regressors. The results are much the same, but the

coefficient on Δ t-p φ is now even more significant.

Having already established that a decline in the private sector’s financing needs

during the economic downturn significantly increases the probability of the ensuing

recovery being credit-less, we now turn to the question of what this decline means, in

terms of flows of funds. Given the aggregate flow of funds identity (1), the private

sector’s financing needs equal the change in the net financial liabilities of HHs and

NFCs, ΔL-ΔFA. Thus, a decline in the private sector’s financing needs during a

downturn could reflect either a decline in the flow of bank loans to the private sector,

i.e. Δ 2 L0. In order to test

which of these two financing channels is predominantly associated with the

occurrence of credit-less recoveries, we construct a proxy for Δ 2 FA by subtracting Δ t-

p φ from the peak-to-trough change in the ratio of loans to GDP. In other words, based

on equation (1), we decompose Δ t-p φ into two flow of funds components

corresponding to Δ 2 L (referred to as “Loans to GDP” in the relevant table) and Δ 2 FA

(referred to as “Fin. assets").

Table 5 presents a battery of estimations exploring this decomposition.

Throughout them all, the coefficient on the change in financial asset holdings (“Fin.

assets”) is highly significant and positive, in line with the negative sign of the

coefficient of the change in φ reported in Tables 3 and 4. This is strong evidence in

support of the hypothesis that credit-less recoveries are associated with a rebalancing

of the private sector’s financing means towards financial assets over the downturn.

Our results suggest that credit-less recoveries are likelier if, during the downturn, the

private sector has accumulated net financial assets, thus implying availability of

private sector financial assets to cover its financing needs for real investment, and

thus to finance a recovery, even in the absence of positive credit flows. The

coefficients on the change in loans to GDP is negative, as expected, confirming the

overall mechanics of the flow of funds identity, but is only significant in the most

extended specification, a sign that perhaps changes in this variable have a less clear

13


association with the likelihood of a credit-less recovery occurring but, rather, are

associated with downturns per se.

In sum, our results suggest that credit-less recoveries are strongly associated

with a decline in the external financing needs (investment-savings gap) of the private

sector during the economic downturn. In turn, this decline in the financing needs for

real investment is related to a rebalancing of wealth from real assets to financial

assets.

Finally, in Table 6 we explore whether the recent trajectory of bank credit

growth matters. There is ample empirical evidence that credit booms precipitate sharp

busts in credit and economic activity (see for example Gourinchas et al., 2001 and

Schularick and Taylor, 2012). There is also evidence that financial crises increase the

probability of a credit-less recovery (Abiad et al 2011, Bijsterbosch and Dahlhaus

2011). In an effort to construct a suitable proxy, we consider the peak-to-trough

change in real credit growth as an additional regressor to some of the most extended

specifications of Tables 2-4. A sharp decline in this variable could be seen as an

indication that a credit boom preceded the downturn and/or that a financial crisis

occurred during the downturn. The estimated coefficient is negative and significant

when added to the specification including φ t , without affecting the significance of φ t

itself or any of the other variables (estimation 1). This implies that the sharper the

peak-to-trough decline in real credit growth, the higher the probability that the

ensuing recovery will be credit-less. When φ t is decomposed into φ p and Δ t-p φ

(estimation 2), or when only Δ t-p φ is included (estimation 3), the change in real credit

growth loses its significance. However, when the change in real credit growth is

included in interaction with φ t , it becomes significant again and leads to a marked

increase in R 2 . Furthermore, the interaction term has an intuitive coefficient: given

that, as seen in Figure 1.b, φ t is mostly negative in credit-less recovery episodes, a

positive coefficient of the interaction term implies that the sharper the peak-to-trough

decline in real credit growth, given negative financing needs at trough, the higher the

likelihood that the recovery will be creditless, over and above the effect of Δ t-p φ. 10

10 We also experimented with other proxies. We included credit growth at peak, whose estimated

coefficient was negative, an indication perhaps that high credit growth at the peak of the cycle does not

in itself imply a credit boom. The inclusion of private sector credit to GDP at trough and at peak,

measures of either bank dependence or of a preceding credit boom, was positive. This is intuitive, as ex

ante, bank-based economies, which have a developed banking system and rely on it heavily for the

financing of private sector activities, are thought to be likelier candidates for a credit-less recovery

14


This would suggest that it is not the past trajectory of bank credit growth per se that

matters in this context (as proposed by Abiad et al., 2011, inter alia), but rather its

coincidence with a rebalancing of the economy over the downturn, in the form of a

decline in the investment-savings gap.

5. Robustness checks

While the results are not reported here in the interest of brevity, sub-sample

estimations along the advanced-emerging economy distinction yield much the same

results. 11 Moreover, the inclusion of an advanced economy dummy turns out to be

mostly insignificant in interaction with other variables, other estimates remaining

largely unaffected. Moving to more parsimonious specifications also does not

qualitatively alter our findings. These robustness checks suggest that our estimates are

stable across specifications.

As an alternative approach to checking the robustness of our findings, we

abstract from the somewhat limiting concept of credit-less recoveries and consider

instead the importance of financing needs φ to developments in credit during

recoveries in general. To this aim, Table 7 presents OLS estimations, where the

average real credit growth over the 8 quarters following a trough has been regressed

on the average real output growth recorded over the same period, and on the variables

employed in our earlier estimations. Note that the dependent variable is constructed so

as to precisely match the definition of a credit-less recovery adopted previously,

which also referred to credit developments 8 quarters following the trough. Moreover,

the only contemporaneous variable is output growth, while all other variables are

timed either at the trough or before it, as peak-to-trough differences, so that these

estimations may best serve as robustness tests of our probit results. In brief, these

regressions are designed to explore whether the factors considered in Section 4 affect

episode (see Reichlin, 2013, and Allard and Blavy, 2011), while economies which depend more on

capital market financing are less likely to undergo a very sharp liquidity squeeze. Finally, we

considered a relative measure, the ratio of real private sector credit growth to real GDP growth at peak,

designed to capture whether, at peak, credit was increasing at rates disproportionate to the real growth

rate. According to our estimates, if credit is growing faster than output at the peak of the cycle (a

possible indication of a preceding credit boom) it is relatively likelier that a credit-less recovery will

follow the downturn. However, these measures were significant in some of the more parsimonious

specifications but not significant when monetary and other policy variables were included (results

available upon request).

11 Results available from the authors upon request.

15


post-trough credit growth in general and, more specifically, to see whether financing

needs, as captured by our proxy, play a role in this respect.

The estimated coefficients on the change in financing needs Δ t-p φ is positive as

expected and highly significant in all Table 7 estimations, as is the coefficient on

post-trough average output growth. Hence, we find that an increase (decline) in the

private sector’s financing needs during the downturn implies higher (lower) credit

growth during the recovery phase. This corroborates the main finding of Section 4.

The coefficient on φ t itself is also positive, though insignificant. The estimated

coefficient on loans to GDP at peak is highly significant and negative, as is the one on

credit growth relative to GDP growth, confirming the earlier finding that recessions in

bank dependent economies, preceded by credit booms are associated with relatively

slower credit growth during the rebound. Finally, the coefficients on the other

variables are also intuitive, implying that average real credit growth during the

rebound is positively related to average real output growth over the same period, an

appreciation of the domestic currency and monetary policy easing during the

downturn and negatively related to inflation at the trough of the recession.

6. Conclusions and policy implications

We argue that a decline in the private sector’s financing needs –the savingsinvestment

imbalance– during a recession increases the probability that the ensuing

recovery may be credit-less. We construct a suitable proxy and obtain empirical

evidence in support of this hypothesis. We show that such a decline reflects a

rebalancing of wealth towards financial assets on behalf of the private sector during

the downturn, which frees up funds to finance real investment during the recovery

stage, even in the absence of positive bank credit flows. Finally, we provide evidence

that economies whose downturn was preceded by a bank credit boom or was

associated with a sharp decline in bank credit are more likely to experience a creditless

recovery. These insights are potentially valuable from a broader policy

perspective, as they suggest that a rapid rebalancing of the economy is pivotal to

ensuring an economic rebound, in situations where the banking sector is unable to

provide sufficient credit.

16


7. Bibliography

Abiad, A., G. Dell’Ariccia and B. Li (2011), “Creditless Recoveries”, IMF Working Paper

11/58.

Allard, J. and R. Blavy, (2011), “Market Phoenixes and Banking Ducks - Are Recoveries

Faster in Market-Based Economies?”, IMF Working Paper 11/213

Bijsterbosch, M. and T. Dahlhaus (2011), “Determinants of Credit-Less Recoveries”, ECB

Working Paper 1358.

Calvo, G., A. Izquierdo, and E. Talvi (2006), “Sudden Stops and Phoenix Miracles in

Emerging Markets”, Τhe American Economic Review, Papers and Proceedings, Vol. 96, No.

2, 405-10.

Chinn, Menzie D., and Hiro Ito. (2008), “A New Measure of Financial Openness” Journal of

Comparative Policy Analysis 10 (3): 309–322.

Claessens, S., K.M. Ayhan and M.E. Terrones (2009), “What Happens during Recessions,

Crunches and Busts?”, Economic Policy, 24, 60, 653–700.

Coricelli, F. and I. Roland, (2011), “How do credit conditions shape economic recoveries?”,

Discussion Paper 8325, CEPR

Corsetti, G., and G. Müller, (2006), “Twin deficits: squaring theory, evidence and common

sense” Economic Policy 48, 597–638

Darvas, Z., (2013), “Can Europe recover without credit?”, Bruegel policy contribution Issue

2013/03

Gourinchas, P, R. Valdes and O. Landerretche, (2001), “Lending Booms: Latin America and

the World,” Economia, Vol. 1, No. 2, pp. 47-99

Ottonello, P., (2010), “The Credit Impulse in Economic Recoveries: Still a Phoenix Miracle”,

mimeo

Reichlin L., (2013), "The ECB and the Banks: The Tale of Two Crises", CEPR Discussion

Paper No. 9647

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

Leverage Cycles, and Financial Crises, 1870-2008.” American Economic Review, 102(2):

1029-61

Sugawara, N. and J. Zalduendo (2013), “Credit-less recoveries: neither a rare nor an

insurmountable challenge”, World Bank Policy Research Working Paper no. 6459.

17


8. Tables and figures

Table 1

Savings and investment of private non-financial sector

before and after the great recession – in % of GDP

Greece Spain Germany France Italy Netherlands Portugal Finland Netherlands UK

S I S I S I S I S I S I S I S I S I S I

2007 0.120 0.230 0.120 0.270 0.220 0.180 0.180 0.180 0.170 0.190 0.220 0.170 0.090 0.200 0.175 0.195 0.155 0.200 0.125 0.145

2009* 0.110 0.110 0.230 0.180 0.230 0.140 0.180 0.140 0.170 0.170 0.210 0.130 0.150 0.140 0.165 0.160 0.140 0.110 0.160 0.110

*: Numbers for Greece and Portugal refer to 2013, when GDP reached a trough. Source: ECB data warehouse.

18


est1 est2 est3 est4 est5

b/z/_star b/z/_star Table b/z/_star 2

b/z/_star b/z/_star

The impact of financing needs on the probability of

a credit-less recovery: financing needs at trough (φ t )

CRE_1

est1 est2 est3 est4 est5

D_RGDPG b/z/_star -0.054 b/z/_star -0.060 b/z/_star -0.121 b/z/_star -0.129 b/z/_star -0.161

-2.56 -2.39 -4.11 -3.85 -4.59

CRE_1

est1

**

est2

**

est3

***

est4

***

est5

***

D_RGDPG D_REER

b/z/_star -0.054 -0.022

b/z/_star -0.060 -0.026

b/z/_star -0.121 -0.007

b/z/_star -0.129 0.005

b/z/_star -0.161 -0.004

-2.56 -1.76 -2.39 -1.97 -4.11 -0.52 -3.85 0.33 -4.59 -0.22 est1

CRE_1

*** **

b/z/_star

*** *** ***

D_REER D_RGDPG

XPY_PK growth

-0.022 -0.054

-0.017 -0.026 -0.060

-0.018 -0.007 -0.121

-0.026

-0.129

-0.028 0.005 -0.004 -0.161

-0.029

(trough minus CRE_1

-1.76 -2.56

-3.46 -1.97 -2.39

-3.29 -0.52 -4.11

-4.03

-3.85

-3.99 0.33 -0.22 -4.59

-4.24

peak)

D_RGDPG ***

*** *** ** ***

***

***

*** -0.106

***

***

XPY_PK D_REER

KAOPEN_PK R.E.E.R.

-0.017 -0.022

0.193 -0.018 -0.026

0.176 -0.026 -0.007

0.293 -0.028 0.005

0.508 -0.029 -0.004

0.533

-3.80

(trough minus

-3.46 -1.76

1.79 -3.29 -1.97

1.60 -4.03 -0.52

2.10 -3.99 0.33

3.18 -4.24 -0.22

3.16

peak)

***

*** * *** *** *** ***

KAOPEN_PK XPY_PK

FBSTYminus~Y Exports D_REER

-0.017

-0.065 0.193 -0.018

-0.058 0.176 -0.026

-0.069 0.293 -0.028

-0.097 0.508 -0.029

-0.046 -0.006

0.533

(%GDP, peak)

-3.46

-2.45 1.79 -3.29

-2.10 1.60 -4.03

-2.15 2.10 -3.99

-2.42 3.18 -4.24

-1.02 -0.42

3.16

***

** * ***

**

*** ** *** ***

FBSTYminus~Y KAOPEN_PK

D_CUAY Capital Acc. XPY_PK -0.065 0.193

0.051 -0.058 0.176

0.050 -0.069 0.293

0.058 -0.097 0.508

0.063 -0.046 0.533

0.096 -0.025

Openness (peak)

-2.45 1.79

1.80 -2.10 1.60

1.57 -2.15 2.10

1.78 -2.42 3.18

1.86 -1.02 3.16

2.05 -2.82

** * ** ** *

*** *

***

**

***

D_CUAY FBSTYminus~Y

D_FBSTY φ t

-0.065 0.051 0.135

-0.058 0.050 0.135

-0.069 0.058 0.134

-0.097 0.063 0.182

-0.046 0.096 0.196

KAOPEN_PK 0.324

-2.45 1.80 2.40

-2.10 1.57 2.38

-2.15 1.78 2.21

-2.42 1.86 2.60

-1.02 2.05 2.71

**

** * **

**

**

** * ***

** * ***

2.42

D_FBSTY D_CUAY

o.Table_2 Current Acc. 0.135 0.051

0.000 0.135 0.050

0.000 0.134 0.058

. 0.182 0.063

. 0.196 0.096

. **

(%GDP, trough DFminusC 2.40 1.80

. 2.38 1.57

. 2.21 1.78

. 2.60 1.86

. -0.140

2.71 2.05

.

minus peak)

*** ** *** **** ***

-2.87

o.Table_2 D_FBSTY

D_VEL Fiscal Pol.

0.000 0.135

. -0.006 0.000 0.135 0.134 .

0.182 . -0.019

0.196 . ***

(easing, trough

2.40 . -1.64

2.38 . 2.21 .

2.60 . -3.30

minus peak) FBSTYminus~K -0.080

2.71 .

** ** ** *** ***

***

D_VEL o.Table_2 0.000 -0.006 0.000 . . -0.019 -2.47

Monetary Pol.

.

(easing, trough

**

. -1.64 . -1.56 . -2.34 . -3.30 -2.82

minus peak)

.

D_VEL ** -0.007

***

FUNDCRED D_VEL

PDEBTY IMF program

.

-0.006 . -0.804 0.012

. -1.328 0.008

. -1.808 -0.019

0.013 -2.05

(dummy, trough)

.

-1.64 . -1.56 1.97

. -2.34 1.27

. -2.82 -3.30

1.63 **

o.Table_3 **

** ***

0.000

PDEBTY FUNDCRED

o.Table_3 Public Debt

. .

-0.804 0.012 0.000

-1.328 0.008 0.000

-1.808 0.013 0.000 .

(%GDP, trough)

. .

-1.56 1.97 .

-2.34 1.27 .

-2.82 1.63.

**

** ***

o.Table_3 PDEBTY

CPIG Inflation FUNDCRED .

. . 0.000 0.012

. 0.000 0.008

0.108 0.000 0.013

0.123

(y-o-y, trough)

. .

1.97 .

1.27

2.95 . 1.63

3.12 .

.

**

*** ***

CPIG o.Table_3 Constant PDEBTY -1.264 . . 0.000 . 0.108 0.000 0.123 0.000

.

-4.46 . -4.12 . -4.04 . -4.61 2.95 .

-4.62 3.12.

.

*** *** *** *** ***

_cons CPIG CPIG -1.264. -1.147. -2.237. -2.924 0.108 -3.157 0.123 .

N Episodes

109.000 -4.46. 109.000 -4.12. 93.000 -4.04. 93.000 -4.61 2.95

93.000 -4.62 3.12

p

Prob.

.

0.000 *** 0.000 *** 0.000 *** 0.000 *** 0.000 ***

_cons

r2_p

Pseudo R 2

-1.264

0.272

-1.147

0.292

-2.237

0.359

-2.924

0.420

-3.157

0.472

N

_cons

109.000 -4.46 109.000 -4.12 93.000 -4.04 93.000 -4.61 93.000

-1.399

-4.62

p 0.000 *** 0.000 *** 0.000 *** 0.000 *** 0.000 ***

-3.60

r2_p 0.272 0.292 0.359 0.420 0.472 ***

N Note: Statistical 109.000 significance at 109.000 the 1%, 5% and 10% 93.000 is denoted by ***, 93.000 ** and * 93.000

p

respectively, using

0.000

robust standard

0.000

errors for the z-statistics.

N 0.000 0.000 0.000 93.000

r2_p p 0.272 0.292 0.359 0.420 0.472 0.006

r2_p 0.339

19


est1 est2 est3 est4 est5

b/z/_star b/z/_star b/z/_star b/z/_star b/z/_star

Table 3

CRE_1 The impact of financing est1 needs est2 on the probability est3 of a credit-less est4 recovery: est5

D_RGDPG

financing b/z/_star -0.052

needs at-peak b/z/_star -0.058

(φ p ) and their b/z/_star -0.121

peak-to-trough b/z/_star -0.127

change (Δ t-p b/z/_star -0.160

φ)

-2.48 -2.35 -4.06 -3.69 -4.56

CRE_1

est1 ** est2 ** est3 *** est4 *** est5 ***

D_REER D_RGDPG b/z/_star -0.018 -0.052 b/z/_star -0.022 -0.058 b/z/_star -0.006 -0.121 b/z/_star -0.127 0.004 b/z/_star -0.004 -0.160

-1.44 -2.48 -1.65 -2.35 -0.38 -4.06 -3.69 0.25 -0.24

est1

-4.56

CRE_1

** ** *

b/z/_star

*** *** ***

D_RGDPG D_REER growth

-0.052 -0.018 -0.058 -0.022 -0.121 -0.006 -0.127 0.004 -0.160 -0.004

(trough minus CRE_1

-2.48 -3.25 -1.44 -2.35 -3.06 -1.65 -4.06 -3.51 -0.38 -3.69 -3.22 0.25 -4.56 -3.56 -0.24

peak)

D_RGDPG *** *** *

*** *** -0.106

***

D_REER KAOPEN_PK XPY_PK R.E.E.R.

-0.018 -0.017 0.207 -0.022 -0.018 0.188 -0.006 -0.026 0.309 -0.027 0.004 0.507 -0.004 -0.030 0.522

-3.80

(trough minus -1.44 -3.25 1.95 -1.65 -3.06 1.76 -0.38 -3.51 2.23 -3.22 0.25 3.04 -0.24 -3.56 3.15

peak)

***

*** * *** * *** *** ***

D_REER -0.006

XPY_PK DFminusC KAOPEN_PK Exports

-0.017 -0.131 0.207 -0.018 -0.125 0.188 -0.026 -0.140 0.309 -0.027 -0.170 0.507 -0.030 -0.154 0.522

(%GDP, peak) -3.25 -3.51 1.95 -3.06 -3.08 1.76 -3.51 -3.08 2.23 -3.22 -3.05 3.04 -3.56 -2.52 -0.42

3.15

**** **** *** *** ***

KAOPEN_PK FBSTYminus~K DFminusC Capital Acc. XPY_PK -0.065 -0.131 0.207 -0.058 -0.125 0.188 -0.070 -0.140 0.309 -0.090 -0.170 0.507 -0.035 -0.154 0.522 -0.025

Openness (peak) -2.35 -3.51 1.95 -1.98 -3.08 1.76 -2.05 -3.08 2.23 -2.07 -3.05 3.04 -0.78 -2.52 3.15 -2.82

*** * *** * *** *** ***

***

DFminusC o.Table_2 FBSTYminus~K Δ t-p φ

-0.131 -0.065 0.000 -0.125 -0.058 0.000 -0.140 -0.070 . -0.170 -0.090 . -0.154 -0.035 .

KAOPEN_PK 0.324

-3.51 -2.35 . -3.08 -1.98 . -3.08 -2.05 . -3.05 -2.07 . -2.52 -0.78 .

2.42

*** *** *** *** **

FBSTYminus~K D_VEL o.Table_2 φ p

-0.065 0.000 . -0.058 -0.005 0.000 -0.070. -0.090. -0.035 -0.018 **

.

DFminusC

-2.35. -1.98 -1.70. -2.05. -2.07. -0.78 -3.50 -0.140

.

** *** ** **

*** -2.87

o.Table_2 D_VEL Monetary Pol. 0.000 -0.005 0.000 -0.873 . -1.454 . -1.987 -0.018 . ***

(easing, trough FBSTYminus~K . -1.70 . -1.73 . -2.58 . -3.23 -3.50 -0.080

.

minus peak)

* * *** ***

-2.47

D_VEL PDEBTY FUNDCRED IMF program

. -0.005. -0.873 0.010 . -1.454 0.006 . -0.018 -1.987 0.013

**

(dummy, trough)

. -1.70. -1.73 1.72 . -2.58 0.98 . -3.50 -3.23 1.69

D_VEL

* * -0.007

*** *** *

FUNDCRED o.Table_3 PDEBTY Public Debt

. . -0.873 0.000 0.010 -1.454 0.000 0.006 -1.987 0.000 -2.05

0.013

(%GDP, trough)

. . -1.73 1.72 . -2.58 0.98 . -3.23 1.69 . **

o.Table_3 * *** *** 0.000

*

PDEBTY CPIG Inflation

o.Table_3 . . 0.010 0.000 . 0.006 0.091 0.000 0.013 0.109 0.000 .

(y-o-y, trough)

. . 1.72 . 0.98 2.40 . 1.69 2.81.

o.Table_3 CPIG

Constant

* ** ***

FUNDCRED *

. . 0.000 . 0.000 0.091 .

0.000 0.109

-4.27 . -3.88 . -3.54 . -3.69 2.40 . -3.97

.

2.81 .

*** *** *** *** ***

CPIG _cons PDEBTY -1.229 . -1.110 . -2.108 . -2.603 0.091 -2.971 0.109 .

N Episodes

109.000 -4.27 . 109.000 -3.88 . 93.000 -3.54 . 93.000 -3.69 2.40 93.000 -3.97 2.81 .

p

Prob.

0.001 *** 0.004 *** 0.001 *** 0.003 *** 0.001

***

Pseudo R

_cons r2_p CPIG -1.229 0.259 -1.110 0.280 -2.108 0.350 -2.603 0.402 -2.971 0.462

.

N 109.000 -4.27 109.000 -3.88 93.000 -3.54 93.000 -3.69 93.000 -3.97 .

p 0.001 *** 0.004 *** 0.001 *** 0.003 *** 0.001 ***

r2_p Note: Statistical 0.259 significance at the 0.280 1%, 5% and 10% 0.350 is denoted by ***, 0.402 ** and * 0.462

N respectively, _cons

109.000 using robust standard 109.000 errors for the z-statistics. -1.399

93.000 93.000 93.000

p 0.001 0.004 0.001 0.003 0.001

-3.60

r2_p 0.259 0.280 0.350 0.402 0.462 ***

N 93.000

p 0.006

r2_p 0.339

20


est1 est2 Table 4

b/z/_star b/z/_star b/z/_star b/z/_star b/z/_star

The impact of financing needs on the probability of

CRE_1

est1 est2 est3 est4

a credit-less recovery: the peak-to-trough change of financing needs (Δ t-p est5

φ)

D_RGDPG b/z/_star -0.049 b/z/_star -0.056 b/z/_star -0.118 b/z/_star -0.118 b/z/_star -0.166

-2.46 -2.33 -4.28 -3.72 -4.65

CRE_1

est1 ** est2 ** est3 *** est4 *** est5 ***

D_REER D_RGDPG b/z/_star -0.021 -0.049 b/z/_star -0.025 -0.056 b/z/_star -0.007 -0.118 b/z/_star -0.004 -0.118 b/z/_star -0.006 -0.166

-1.82 -2.46 -1.86 -2.33 -0.56 -4.28 -0.26 -3.72 -0.37 est1

-4.65

CRE_1

*** ***

*** *** b/z/_star

***

XPY_PK D_RGDPG D_REER growth

-0.014 -0.049 -0.021 -0.016 -0.056 -0.025 -0.022 -0.118 -0.007 -0.022 -0.118 -0.004 -0.030 -0.166 -0.006

(trough minus

CRE_1

-2.83 -2.46 -1.82 -2.73 -2.33 -1.86 -3.66 -4.28 -0.56 -3.39 -3.72 -0.26 -3.53 -4.65 -0.37

peak)

D_RGDPG *** * *** *

*** *** -0.106

***

KAOPEN_PK D_REER XPY_PK R.E.E.R.

-0.021 -0.014 0.188 -0.025 -0.016 0.169 -0.007 -0.022 0.265 -0.004 -0.022 0.415 -0.006 -0.030 0.507

-3.80

(trough minus

-1.82 -2.83 1.83 -1.86 -2.73 1.63 -0.56 -3.66 2.06 -0.26 -3.39 2.80 -0.37 -3.53 3.13

peak)

***

**** *** *

*** *** ***

DFminusC XPY_PK KAOPEN_PK Exports

D_REER

-0.077 -0.014 0.188 -0.077 -0.016 0.169 -0.079 -0.022 0.265 -0.092 -0.022 0.415 -0.132

-0.006

-0.030 0.507

(%GDP, peak) -2.99 -2.83 1.83 -2.91 -2.73 1.63 -2.53 -3.66 2.06 -2.72 -3.39 2.80 -3.22 -3.53 3.13 -0.42

*** * *** *** *** ***

o.Table_2 KAOPEN_PK DFminusC Capital Acc. XPY_PK -0.077 0.000 0.188 -0.077 0.000 0.169 -0.079 0.265 . -0.092 0.415 . -0.132 0.507 -0.025

.

Openness (peak)

-2.99 1.83 . -2.91 1.63 . -2.53 2.06 . -2.72 2.80 . -3.22 3.13 -2.82

.

**** *** ** *** *** ***

D_VEL DFminusC o.Table_2 Δ t-p φ

-0.077 0.000 . -0.006 -0.077 0.000 -0.079 . -0.092 . -0.020 -0.132.

KAOPEN_PK 0.324

-2.99 . -1.91 -2.91. -2.53 . -2.72 . -3.56 -3.22.

*** *** * ** *** ***

2.42

FUNDCRED o.Table_2 D_VEL Monetary Pol.

0.000 . -0.006 0.000 . -0.947 . -1.475 . -2.145 -0.020 **

.

(easing, trough

DFminusC . -1.91 . -1.80 . -2.44 . -3.34 -3.56 -0.140

minus peak)

.

* * ** ***

-2.87

PDEBTY D_VEL FUNDCRED IMF program

. -0.006 . -0.947 0.015 . -1.475 0.012 . -0.020 -2.145 0.016 ***

(dummy, trough)

FBSTYminus~K

. -1.91 . -1.80 2.47 . -2.44 1.90 . -3.56 -3.34 -0.080

2.30

* *** *** ***

-2.47

o.Table_3 FUNDCRED PDEBTY Public Debt

. . -0.947 0.000 0.015 -1.475 0.000 0.012 -2.145 0.000 0.016

(%GDP, trough)

. . -1.80 2.47 . -2.44 1.90 . -3.34 2.30 .

**

D_VEL ** * *** -0.007

***

CPIG PDEBTY o.Table_3 Inflation

. . 0.015 0.000 . 0.063 0.012 0.000 0.106 0.016 0.000 -2.05

(y-o-y, trough)

. . 2.47 . 2.14 1.90 . 3.04 2.30.

**

_cons o.Table_3 CPIG

Constant

o.Table_3 ** *** ***

0.000

-1.133 . -1.009 . -2.229 0.000 . -2.561 0.000 0.063 -3.071 0.000 0.106 .

-4.16 . -3.68 . -3.86 . -3.81 2.14 . -4.09 3.04 .

*** *** *** *** ***

FUNDCRED .

CPIG _cons -1.133 . -1.009 . -2.229 . -2.561 0.063 -3.071 0.106

N Episodes

109.000 -4.16 . 109.000 -3.68 . 93.000 -3.86 . 93.000 -3.81 2.14 93.000 -4.09 .

3.04

p Prob.

0.000 *** 0.002 *** 0.000 *** 0.001 *** 0.000

***

r2_p Pseudo R 2

_cons PDEBTY -1.133 0.224 -1.009 0.254 -2.229 0.315 -2.561 0.358 -3.071 0.457 .

N 109.000 -4.16 109.000 -3.68 93.000 -3.86 93.000 -3.81 93.000 -4.09 .

p 0.000 *** 0.002 *** 0.000 *** 0.001 *** 0.000 ***

r2_p Note: Statistical 0.224 significance at the 0.254 1%, 5% and 10% 0.315 is denoted by ***, 0.358 ** and * 0.457

CPIG .

N respectively, 109.000 using robust standard 109.000 errors for the z-statistics. 93.000 93.000 93.000

.

p 0.000 0.002 0.000 0.001 0.000

r2_p 0.224 est1 0.254 est2 0.315 est3 0.358 est4

0.457

_cons b/z/_star b/z/_star b/z/_star b/z/_star -1.399

-3.60

CRE_1

***

D_RGDPG -0.106 -0.121 -0.127 -0.160

-3.80 -4.06 -3.69 -4.56

N 93.000

*** *** *** ***

D_REER p -0.006 -0.006 0.004 -0.004 0.006

r2_p -0.42 -0.38 0.25 -0.24 0.339

XPY_PK -0.025 -0.026 -0.027 -0.030

-2.82 -3.51 -3.22 -3.56

*** *** *** ***

KAOPEN_PK 0.324 0.309

21

2.42 2.23 3.04 3.15

** ** *** ***

DFminusC -0.140 -0.140 -0.170 -0.154

-2.87 -3.08 -3.05 -2.52

*** *** *** **


Table 5

The impact of financing needs on the probability of a credit-less recovery:

a flow of funds decomposition of the investment-savings gap

est1 est2 est3 est4 est5 est6

b/z/_star b/z/_star b/z/_star b/z/_star b/z/_star b/z/_star

CRE_1

est1 est2 est3 est4 est5 est6

D_RGDPG b/z/_star -0.081 b/z/_star -0.103 b/z/_star -0.120 b/z/_star -0.153 b/z/_star -0.120 b/z/_star -0.185

est

-3.34 -3.93 -4.16 -4.89 -3.62 -4.50

b/z/_sta

CRE_1

est1 *** est2 *** est3 *** est4 *** *** ***

D_RGDPG D_REER growth b/z/_star -0.081 -0.010 b/z/_star -0.103 -0.012 b/z/_star -0.120 -0.008 b/z/_star -0.153 -0.010 -0.120 -0.004 -0.185 -0.009

(trough minus CRE_1

-3.34 -0.76 -3.93 -0.79 -4.16 -0.58 -4.89 -0.61 -3.62 -0.29 -4.50 -0.55

peak)

CRE_1 D_RGDPG *** *** *** *** *** -0.10

***

D_RGDPG D_REER XPY_PK R.E.E.R.

-0.106 -0.010 -0.018 -0.121 -0.012 -0.020 -0.127 -0.008 -0.022 -0.160 -0.010 -0.027 -0.004 -0.022 -0.009 -0.031 -3.8

(trough minus -3.80 -0.76 -2.70 -4.06 -0.79 -2.70 -3.69 -0.58 -3.69 -4.56 -0.61 -4.20 -0.29 -3.41 -0.55 -3.61

peak)

**

*** *** *** *** *** *** *** *** *** ***

D_REER -0.00

D_REER XPY_PK KAOPEN_PK Exports

-0.006 -0.018 0.280 -0.006 -0.020 0.293 0.004 -0.022 0.277 -0.004 -0.027 0.276 -0.022 0.424 -0.031 0.574

(%GDP, peak) -0.42 -2.70 2.26 -0.38 -2.70 2.32 0.25 -3.69 2.12 -0.24 -0.4

-4.20 2.04 -3.41 2.78 -3.61 3.15

*** *** *** *** *** ***

XPY_PK KAOPEN_PK D_FA Capital Acc. XPY_PK -0.025 0.280 0.069 -0.026 0.293 0.075 -0.027 0.277 0.080 -0.030 0.276 0.098 0.424 0.093 -0.02

0.574 0.145

Openness (peak) -2.82

2.26

-3.51

2.32

-3.22

2.12

-3.56

2.35 2.33 2.49 2.04 3.20 2.78 2.67 3.15 3.38 -2.8

*** *** *** ***

KAOPEN_PK ** ** ** *** *** *** **

D_FA D_RPSCY Fin. assets

0.069 0.075 0.080 0.098 0.093 0.145

KAOPEN_PK 0.324 -0.105 0.309 -0.139 0.507 -0.113 0.522 -0.186 -0.116 -0.286 0.32

(%GDP, trough 2.42 2.23 3.04 3.15

-1.00 2.35 -1.19 2.33 -0.94 2.49 -1.42 3.20 -1.09 2.67 -2.29 3.38

minus peak)

** ** *** ***

2.4

** ** ** *** *** ***

DFminusC -0.140 -0.140 -0.170 -0.154

D_RPSCY -0.105 -0.139 -0.113 -0.186 -0.116 -0.286 *

o.Table_3 Loans to GDP

DFminusC -2.87 0.000 -3.08 0.000 -3.05 0.000 -2.52 0.000 0.000 0.000

(trough minus -1.00. -1.19. -0.94. -1.42. -1.09. -0.14

-2.29.

peak)

*** *** *** **

FBSTYminus~K -0.080 -0.070 -0.090 -0.035 -2.8

**

o.Table_3 D_VEL Monetary Pol. -2.47 0.000 . -2.05 -0.008 0.000 -2.07 0.000 . -0.78 -0.011 0.000 0.000 . -0.021 0.000 **

(easing, trough

FBSTYminus~K ** . -2.12 ** . **

. -2.72 . . -0.08

-3.72 .

minus peak)

D_VEL -0.007 .** . -0.018 *** ***

-2.4

D_VEL FUNDCRED IMF program

-2.05 . -0.008 . . -0.954 . . -3.50 -0.011 -1.077 -1.474 . -0.021 -2.334

(dummy, trough)

*

** . -2.12. -1.80 . *** -2.72 -1.99 -2.42 . -3.72 -3.21

o.Table_3 D_VEL 0.000 0.000 ** 0.000 * 0.000 *** **

-0.00

***

FUNDCRED PDEBTY Public Debt

. . . . -0.954 0.015 . -1.077 0.019 .

-1.474 0.012 -2.334 0.016 -2.0

(%GDP, trough)

. . -1.80 2.49 -1.99 3.15 -2.42 1.91 -3.21 2.32 *

FUNDCRED . -0.873 -1.454

o.Table_3 *** -1.987

*** *** *** 0.00

PDEBTY

.

CPIG -1.73

. -2.58

. 0.015

-3.23

Inflation

. 0.019 . 0.012 0.063 0.016 0.110

* *** ***

(y-o-y, trough)

. . 2.49 . 3.15 . 1.91 2.10 2.32 2.88

PDEBTY . 0.010 0.006 0.013

** *** *** ***

FUNDCRED

. 1.72 0.98 1.69

CPIG _cons Constant

-1.277 . -1.227 . -2.219 . -2.372 . -2.555 0.063 -3.151 0.110

* *

-3.77 . -3.44 . -3.87 . -4.29 . -3.80 2.10 -4.04 2.88

CPIG . . 0.091 0.109

*** *** *** *** *** . . 2.40 2.81

***

_cons PDEBTY

-1.277 -1.227 -2.219

** ***

-2.372 -2.555 -3.151

Episodes _cons N -1.399 93.000 -3.77 -2.108 93.000 -3.44 -2.603 93.000 -3.87 -2.971 93.000 -4.29 93.000 -3.80 93.000 -4.04

Prob. p -3.60 0.001 *** -3.54 0.002 *** -3.69 0.001 *** -3.97 0.000 *** 0.002 *** 0.001 ***

Pseudo r2_p R 2

0.259 *** 0.292 *** *** 0.316 *** 0.369 0.358 0.467

CPIG

N 93.000 93.000 93.000 93.000 93.000 93.000

N p 93.000 0.001 93.000 0.002 93.000 0.001 93.000 0.000 0.002 0.001

p r2_p Note: Statistical significance 0.006 0.259 at the 1%, 0.001 0.292 5% and 10% is 0.003 denoted 0.316 by ***, 0.001 ** and 0.369 * respectively, 0.358 using robust 0.467

r2_p standard errors for _cons the 0.339 z-statistics. 0.350 0.402 0.462 -1.39

-3.6

**

N 93.00

p 0.00

r2_p 0.33

22


Table 6

The impact of financing needs on the probability of

a credit-less recovery: the role of credit shortages

est1 est2 est3 est4 est5

b/z/_star b/z/_star b/z/_star b/z/_star b/z/_star

est1 est2 est3 est4 est5

CRE_1

b/z/_star b/z/_star b/z/_star b/z/_star b/z/_star

est1

D_RGDPG -0.139 -0.162 -0.162 -0.160 -0.159

b/z/_star

CRE_1

D_RGDPG est1 -3.67 -0.139 est2 -3.88 -0.162 est3 -4.19 -0.162 est4 -4.18 -4.34

growth

b/z/_star *** b/z/_star *** b/z/_star *** b/z/_star -0.160 *** -0.159 ***

(trough

D_REER CRE_1

minus

-3.67 -3.88 -4.19 -4.18 -4.34

peak)

0.003 -0.008 -0.008 0.004 0.005

CRE_1 D_RGDPG 0.18 *** -0.44 *** -0.47 *** 0.23 *** 0.23 *** -0.106

D_RGDPG D_REER R.E.E.R. -0.1060.003 -0.121-0.008 -0.127 -0.008 -0.160 0.004 0.005 -3.80

(trough minus

XPY_PK -3.80 -0.023 0.18 -4.06-0.028 -0.44 -3.69 -0.028 -0.47 -4.56 0.23 0.23

peak)

-0.028 -0.028 ***

***-3.23 *** -4.23 *** -4.16 ***

XPY_PK D_REER -4.04 -3.95

-0.023 -0.028 -0.028 -0.028 -0.028 -0.006

D_REER Exports

-0.006 -0.006 0.004 -0.004

*** *** *** *** ***

(%GDP, peak) -0.42-3.23 -0.38 -4.23 0.25 -4.16 -0.24 -0.42

KAOPEN_PK 0.509 0.533 0.530 0.531 -4.04 0.528 -3.95

2.78 *** 2.99 *** 3.00 *** 2.99 *** 3.07 ***

XPY_PK KAOPEN_PK Capital XPY_PK Acc. -0.025 0.509 -0.026 0.533 -0.027 0.530 -0.030

0.531 0.528 -0.025

*** *** *** *** ***

D_VEL Openness (peak) -2.82 -0.015 -3.51

2.78 -3.22

2.99 -3.56

3.00 2.99 3.07 -2.82

*** *** *** *** *** *** ***

-2.75 -3.13 -3.13 -1.98 *** -1.97 *** ***

KAOPEN_PK D_VEL 0.324 -0.015 0.309 -0.022 0.507 -0.022 0.522

Monetary KAOPEN_PK Pol.

2.42 *** 2.23 *** 3.04 *** 3.15

-0.014 ** -0.014 ** 0.324

(easing, trough

FBSTYminus~Y -0.084 -2.75 -3.13 . -3.13 . -1.98 . -1.97 .

minus peak)

** *** ** *** *** *** *** 2.42

DFminusC -0.140-1.75 -0.140 . -0.170 . -0.154 ** . ** .

FBSTYminus~Y -0.084 . . . . **

φ t

FUNDCRED -2.87 * -3.08 -3.05 -2.52

*** -1.846 -1.75 ***-2.515 . *** -2.527. ** -2.514. -2.523

.

FBSTYminus~K -0.080-2.88 * -0.070 -3.61 -0.090 -3.59 -0.035 -2.87

-3.10 -3.17

FUNDCRED IMF program -2.47 -1.846 *** -2.05 -2.515 *** -2.07 -2.527 *** -0.78 -2.514 *** -2.523 *** ***

(dummy, trough)

PDEBTY FBSTYminus~K **

0.009 -2.88 **

0.019 -3.61 **

0.019 -3.59 0.014 -3.10 0.014 -3.17 -0.080

D_VEL -0.007

1.37 *** .

1.90 *** .

2.53 *** -0.018

1.84 *** 1.84 *** -2.47

PDEBTY Public Debt -2.05 0.009 . 0.019 . 0.019 -3.50 0.014 0.014 **

(%GDP, trough) **

CPIG 0.128 1.37 0.161 1.90 0.161 2.53

***

0.173 1.84 0.173 1.84

o.Table_3 D_VEL 0.000 0.000

2.50 3.07 * 0.000

3.08 ** 0.000

3.26 * -0.007

3.26 *

CPIG . 0.128 .

** 0.161 .

*** 0.161 .

*** 0.173 *** 0.173 -2.05

Inflation

***

(y-o-y, trough)

D_RPSCG -0.042 2.50 -0.007 3.07 -0.006 3.08 0.002 3.26 3.26 **

FUNDCRED o.Table_3 . .

-1.98 ** -0.873

-0.26 *** -1.454

-0.29 *** -1.987

0.09 *** *** .

D_RPSCG -0.042 . -1.73

** -0.007 -2.58 -0.006 -3.23

Credit growth

* *** *** 0.002 . .

(change, trough

PDEBTY o.Table_3 . 0.000 -1.98 0.010 0.000 -0.26 0.006 0.000 -0.29 0.013 0.000 0.09 0.000.

minus

0.98 . 1.69 . . .

peak)

FUNDCRED . ** . 1.72 .

o.Table_3 0.000 * 0.000 0.000 * 0.000 0.000

CPIG FBSTYminus~K φ p

. . .-0.005 . 0.091 . 0.109 . .

.

.

FBSTYminus~K . -0.005 ** . *** . .

Δ t-p φ

-1.399 . -0.147 -0.07 . . . .

-3.60 -3.54 -3.69 -3.97

. -2.43 -3.00 -2.41 -2.58

DFminusC ***

CPIG . *** -0.147 ***

** -0.144 ***

*** -0.132 ** -0.130 *** .

PDEBTY

. . . -0.07 2.40 . 2.81 . .

I_6 Interaction term

. -2.43 . -3.00 . 0.011 -2.41 0.011 -2.58 .

N 93.000 93.000

p . ** 93.000

. *** 93.000

Δ t-p (Credit gr.)

. 2.28 ** 2.11 ***

r2_p I_6 0.006

_cons 0.339 . 0.001

0.350 . 0.003

0.402 . 0.001

with φ t

0.462

0.011 ** 0.011 -1.399

**

_cons Constant

-2.847. -3.552. -3.565. -3.515 2.28 -3.512 2.11

-4.11 -3.89 -3.97 -4.25 ** -4.26 ** -3.60

_cons -2.847 *** -3.552 *** -3.565 *** -3.515 *** -3.512 ***

***

-4.11 -3.89 -3.97 -4.25 -4.26

Episodes N N 90.000 *** 90.000 *** 90.000 *** 90.000 *** 90.000 *** 93.000

Prob.

p p 0.002 0.003 0.003 0.000 0.000 0.006

Pseudo N R

r2_p r2_p 90.000 0.429 90.000 0.492 90.000 0.492 90.000 0.527 90.000 0.527 0.339

p 0.002 0.003 0.003 0.000 0.000

r2_p 0.429 0.492 0.492 0.527 0.527

Note: Statistical significance at the 1%, 5% and 10% is denoted by ***, ** and * respectively, using robust

standard errors for the z-statistics.

23


Table 7

OLS robustness tests

Dependent variable: est1 est2 est3 est4 est5

average real credit b/z/_star growth over b/z/_star the 8 quarters b/z/_star following b/z/_star the trough b/z/_star

RGDPG_t8 1.196 est1 1.150 est2 1.177 est3 1.209 est4 1.373 est5

b/z/_star 5.10 b/z/_star 5.26 b/z/_star 5.31 b/z/_star 5.91 b/z/_star 6.53

est1

*** *** *** *** ***

RGDPG_t8 D_RGDPG growth

b/z/_star

1.196 0.355 1.150 0.452 1.177 0.387 1.209 0.538 1.373 0.591

(mean, 2 years 5.10 0.94 5.26 1.23 5.31 1.13 5.91 1.49 6.53 1.26

after trough)

CRE_1 *** *** *** *** ***

D_RGDPG D_REER growth D_RGDPG 0.355 0.524 0.452 0.508 0.387 0.425 0.538 0.288 0.591 0.040 -0.106

(trough minus 0.94 2.15 1.23 2.10 1.13 1.97 1.49 1.12 1.26 0.14

peak)

-3.80

** ** *

***

D_REER XPY R.E.E.R.

0.524 0.041 0.508 0.082 0.425 0.052 0.288 0.054 0.040 0.035

(trough minus D_REER -0.006

2.15 0.95 2.10 1.73 1.97 1.10 1.12 1.05 0.14 0.77

peak)

** ***

* -0.42

XPY KAOPEN_PK Exports

-2.885 0.041 -2.034 0.082 -0.546 0.052 0.054 0.158 0.035 1.152

(%GDP, peak) XPY_PK -1.58 0.95 -1.19 1.73 -0.37 1.10 1.05 0.09 -0.025

0.77 0.66

*

-2.82

KAOPEN_PK DFminusC Capital Acc. -2.885 2.065 -2.034 1.898 -0.546 1.722 0.158 1.779 1.152 1.879 ***

Openness (peak) -1.58 5.12 -1.19 5.15 -0.37 4.60 0.09 4.18 0.66 4.73

KAOPEN_PK 0.324

*** *** *** *** ***

DFminusC FBSTYminus~Y Δ t-p φ

2.42

2.065 0.060 -0.070 1.898 1.722 0.176 1.779 0.407 1.879 0.323

5.12 0.18 -0.22 5.15 4.60 0.51 4.18 0.82 4.73 0.59 **

DFminusC *** *** *** *** -0.140

***

FBSTYminus~Y o.Table_1 φ t

0.060 0.000 -0.070 0.000 0.176 0.000 0.407 0.000 0.323 0.000 -2.87

0.18 . -0.22. 0.51 . 0.82 . 0.59.

***

FBSTYminus~K -0.080

o.Table_1 RPSCY_PK Loans to GDP

0.000 . -0.774 0.000 -0.721 0.000 -0.549 0.000 -0.215 0.000

(peak)

-2.47

. -2.99 . -2.87 . -1.72 . -0.62 .

**

*** *** *

RPSCY_PK BIGGSG_PK Credit growth D_VEL . -0.774. -0.721 -3.282 -0.549 -4.901 -0.215 -3.669 -0.007

(relative, peak)

. -2.99. -2.87 -2.56 -1.72 -3.44 -0.62 -2.10 -2.05

*** *** ****

** **

BIGGSG_PK D_VEL Monetary Pol.

o.Table_3 . . -3.282. -4.901 0.121 -3.669 0.163 0.000

(easing, trough

. . -2.56. -3.44 2.87 -2.10 3.85

minus peak)

.

** *** ***

D_VEL FUNDCRED IMF program

. . . 0.121 . 12.808 0.163

(dummy, trough) FUNDCRED . . . 2.87 . 3.85 1.60 .

*** *** .

FUNDCRED PDEBTY Public Debt

. . . . 12.808 -0.077

(%GDP, trough) PDEBTY . . . . -0.91 1.60 .

.

PDEBTY CPIG Inflation

. . . . -0.077 -0.795

(y-o-y, trough)

. . . . -0.91 -2.10

CPIG .

**

CPIG _cons Constant

13.716 . 19.612 . 24.037 . 23.042 . -0.795 20.825 .

2.81 . 3.50 . 3.89 . 3.30 . -2.10 2.80

_cons *** *** *** *** -1.399

***

_cons 13.716 19.612 24.037 23.042 20.825 -3.60

N Episodes

129.000 2.81 129.000 3.50 126.000 3.89 91.000 3.30 77.000 2.80

p Prob.

0.000 *** 0.000 *** 0.000 *** 0.000 *** 0.000 ***

***

r2 R 2

0.439 0.462 0.513 0.610 0.654

N N 129.000 129.000 126.000 91.000 77.000 93.000

p p 0.000 0.000 0.000 0.000 0.000 0.006

r2 Note: Statistical r2_p significance 0.439 at the 1%, 5% 0.462 and 10% is denoted 0.513 by ***, ** and 0.610 * respectively, using 0.654 0.339

robust

standart errors for the t-statistics.

24


0 5

0

Percent

10 15 20

Percent

10 20 30

0

0

Percent

20 40 60 80

Percent

10 20 30

Figure 1.a: The distribution of the financing needs proxy at trough, φ t ,

for credit-less recoveries:

70.69

12.07

10.34

3.448

3.448

-30 -20 -10 0 10

DFminusC

-40 -20 0 20 40

FBSTYminusCUAY

for credit-supported recoveries:

19.44

20.83

15.28

15.97

9.722

5.556

3.472

2.778

2.083

.6944

2.083 2.083

-30 -20 -10 0 10

DFminusC

-20 -10 0 10 20

FBSTYminusCUAY

25


0 5

0

Percent

10 15 20 25

Percent

10 20 30

0

0

Percent

10 20 30

Percent

10 20 30

Figure 1.b: The distribution of the peak-to-trough change

in the financing needs proxy, Δ t-p φ,

for credit-less recoveries:

32.08

22.64

16.98

11.32

7.547

3.774

5.66

-30 -20 -10 0 10

DFminusC

-30 -20 -10 0 10

DFminusC

for credit-supported recoveries:

24.43

23.66

6.87

22.9

10.69

4.58

2.29

2.29

.7634

-30 -20 -10 0 10

DFminusC

.7634 .7634

-20 -10 0 10 20

DFminusC

26


Figure 2: Investment-savings gap

of households and non-financial companies - % of GDP

0.13

0.08

0.03

-0.02

-0.07

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Greece Spain Portugal UK Denmark

0.1

0.05

0

-0.05

-0.1

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Germany France Finland

27


BANK OF GREECE WORKING PAPERS

185. Adam, A., and T., Moutos, “Industry-Level Labour Demand Elasticities

Across the Eurozone: Will There Be Any Gain After the Pain of Internal

Devaluation?” July, 2014.

186. Tagkalakis, O.A., “Fiscal Policy, Net Exports, and the Sectoral Composition

of Output in Greece”, September 2014.

187. Hondroyiannis, G. and D., Papaoikonomou, “When Does it Pay To Tax?

Evidence from State-Dependent Fiscal Multipliers in the Euro Area”, October

2014.

188. Charalambakis, C. E., “On Corporate Financial Distress Prediction: What Can

we Learn From Private Firms in a Small Open Economy?, November 2014.

189. Pagratis, S., E., Karakatsani and E. Louri, “Bank Leverage and Return on

Equity Targeting: Intrinsic Procyclicality of Short-Term Choices”, November

2014.

190. Evgenidis, A. and C., Siriopoulos, “What are the International Channels

Through Which a US Policy Shock is Transmitted to the World Economies?

Evidence from a Time Varying Favar, January 2015.

191. Louzis, D. P., and A.T., Vouldis, “Profitability in the Greek Banking System:

a Dual Investigation of Net Interest and Non-Interest Income”, February 2015.

192. Papaspyrou, S.T, “EMU 2.0 - Drawing Lessons From the Crisis - a New

Framework For Stability and Growth”, March 2014.

193. Litina, A and T, Palivos, “Corruption and Tax Evasion: Reflections on Greek

Tragedy”, June 2015.

194. Balfoussia, H. and H.D. Gibson, “Financial Conditions and Economic

Activity: The Potential Impact of the Targeted Longer-Term Refinancing

Operations (TLTROS)”, July 2015.

195. Louzis, P. D., “Steady-State Priors and Bayesian Variable Selection in VAR

Forecasting”, July 2015.

196. Zografakis, S. and A., Sarris, “The Distributional Consequences of the

Stabilization and Adjustment Policies in Greece During the Crisis, with the

Use of A Multisectoral Computable General Equilibrium Model”, August

2015.

197. Papageorgiou, D. and E., Vourvachaki, “The Macroeconomic Impact of

Structural Reforms in Product and Labour Markets: Trade-Offs and

Complementarities”, October 2015.

198. Louri, H., and P., M., Migiakis, “Determinants of Euro-Area Bank Lending

Margins: Financial Fragmentation and ECB Policies”, October 2015.

199. Gibson, D. H, S.G. Hall, and G. S. Tavlas, “The effectiveness of the ECB’s

asset purchase programs of 2009 to 2012”, November 2015.

28

More magazines by this user
Similar magazines