BIS - The Real Effects of Debt

Cecchetti, Mohanty and Zampolli The real effects of debt
The real effects of debt

Stephen G Cecchetti, M S Mohanty and Fabrizio Zampolli
*

September 2011

Abstract
At moderate levels, debt improves welfare and enhances growth. But high levels can be
damaging. When does debt go from good to bad? We address this question using a new
dataset that includes the level of government, non-financial corporate and household debt in
18 OECD countries from 1980 to 2010. Our results support the view that, beyond a certain
level, debt is a drag on growth. For government debt, the threshold is around 85% of GDP.
The immediate implication is that countries with high debt must act quickly and decisively to
address their fiscal problems. The longer-term lesson is that, to build the fiscal buffer
required to address extraordinary events, governments should keep debt well below the
estimated thresholds. Our examination of other types of debt yields similar conclusions.
When corporate debt goes beyond 90% of GDP, it becomes a drag on growth. And for
household debt, we report a threshold around 85% of GDP, although the impact is very
imprecisely estimated.

* Cecchetti is Economic Adviser at the Bank for International Settlements (BIS) and Head of its Monetary and
Economic Department; Mohanty is Head of the Macroeconomic Analysis Unit at the BIS; and Zampolli is
Senior Economist at the BIS. This paper was prepared for the Achieving Maximum Long-Run Growth
symposium sponsored by the Federal Reserve Bank of Kansas City, Jackson Hole, Wyoming,
2527 August 2011. We thank Enisse Kharroubi for insightful discussions; Dietrich Domanski, Mathias
Drehmann, Leonardo Gambacorta, Eld Takts, Philip Turner and Christian Upper for suggestions;
participants at the Jackson Hole symposium for numerous comments; Christian Dembiermont, Marjorie
Santos and Denis Marionnet for their special efforts in putting together the dataset on non-financial sector
debt; and Jakub Demski, Jimmy Shek and Michela Scatigna for valuable statistical assistance. The views
expressed in this paper are those of the authors and not necessarily those of the BIS.
1/34

1. Introduction
Debt is a two-edged sword. Used wisely and in moderation, it clearly improves welfare. But,
when it is used imprudently and in excess, the result can be disaster. For individual
households and firms, overborrowing leads to bankruptcy and financial ruin. For a country,
too much debt impairs the governments ability to deliver essential services to its citizens.
High and rising debt is a source of justifiable concern. We have seen this recently, as first
private and now public debt have been at the centre of the crisis that began four years ago.
Data bear out these concerns and suggest a need to look comprehensively at all forms of
non-financial debt: household and corporate, as well as government. Over the past 30 years,
summing these three sectors together, the ratio of debt to GDP in advanced economies has
risen relentlessly from 167% in 1980 to 314% today, or by an average of more than
5 percentage points of GDP per year over the last three decades. Given current policies and
demographics, it is difficult to see this trend reversing any time soon. Should we be worried?
What are the real consequences of such rapid increase in debt levels? When does its
adverse impact bite?
Finance is one of the building blocks of modern society, spurring economies to grow. Without
finance and without debt, countries are poor and stay poor. When they can borrow and save,
individuals can consume even without current income. With debt, businesses can invest
when their sales would otherwise not allow it. And, when they are able to borrow, fiscal
authorities can play their role in stabilising the macroeconomy. But, history teaches us that
borrowing can create vulnerabilities. When debt ratios rise beyond a certain level, financial
crises become both more likely and more severe (Reinhart and Rogoff (2009)). This strongly
suggests that there is a sense in which debt can become excessive. But when?
We take an empirical approach to this question. Using a new dataset on debt levels in 18
OECD countries from 1980 to 2010 (based primarily on flow of funds data), we examine the
impact of debt on economic growth. Our data allow us to look at the impact of household,
non-financial corporate and government debt separately.
1
Using variation across countries
and over time, we examine the impact of the movement in debt on growth.
2

Our results support the view that, beyond a certain level, debt is bad for growth. For
government debt, the number is about 85% of GDP. For corporate debt, the threshold is
closer to 90%. And for household debt, we report a threshold of around 85% of GDP,
although the impact is very imprecisely estimated.
Our result for government debt has the immediate implication that highly indebted
governments should aim not only at stabilising their debt but also at reducing it to sufficiently
low levels that do not retard growth. Prudence dictates that governments should also aim to
keep their debt well below the estimated thresholds so that even extraordinary events are
unlikely to push their debt to levels that become damaging to growth.

1
Flow of funds data should provide a more accurate picture of indebtedness than bank credit data, which
exclude several forms of debt including securitised debt, corporate bonds and trade credit. The difference is
likely to matter in countries such as the United States, where a large fraction of credit is granted by non-bank
intermediaries.
2
Recent empirical studies of the effect of public debt on growth using panel data include Checherita and Rother
(2010) and Kumar and Woo (2010). Unlike these studies, ours investigates the impact on growth of household
and non-financial corporate debt too.
2/34

Taking a longer-term perspective, reducing debt to lower levels represents a severe test for
the advanced economies. Here, the challenge is compounded by unfavourable
demographics. Ageing populations and rising dependency ratios have the potential to slow
growth as well, making it even more difficult to escape the negative debt dynamics that are
now looming.
The remainder of the paper is organised in four sections. In Section 2, we discuss why we
believe that high levels of debt create volatility and are bad for growth. Formal models of this
phenomenon are still at very early stages, so all we can offer is some intuition. We go on, in
Section 3, to a preliminary examination of the data and the main facts about the build-up of
non-financial sector debt in advanced economies. Section 4 contains our main empirical
results. These are based on a series of standard growth regressions, augmented with
information about debt levels. It is here that we report our estimates of the thresholds beyond
which debt becomes a drag on growth. Section 5 discusses these results in the context of
the inescapable demographic trends. Section 6 concludes.
2. Why debt matters
For a macroeconomist working to construct a theoretical structure for understanding the
economy as a whole, debt is either trivial or intractable. Trivial because (in a closed
economy) it is net zero the liabilities of all borrowers always exactly match the assets of all
lenders. Intractable because a full understanding of debt means grappling with a world in
which the choice between debt and equity matters in some fundamental way. That means
confronting, among other things, the intrinsic differences between borrowers and lenders;
non-linearities, discontinuities, and constraints in which bankruptcy and limits on borrowing
are key; taxes, where interest paid to lenders is treated differently from dividends paid to
shareholders; differences between types of borrowers, so household, corporate and
government debt are treated separately; and externalities, since there are times when
financial actors do not bear (or are able to avoid) the full costs of their actions.
As modern macroeconomics developed over the last half-century, most people either ignored
or finessed the issue of debt. With few exceptions, the focus was on a real economic system
in which nominal variables prices or wages, and sometimes both were costly to adjust.
The result, brought together brilliantly by Michael Woodford in his 2003 book, is a logical
framework where economic welfare depends on the ability of a central bank to stabilise
inflation using its short-term nominal interest rate tool. Money, both in the form of the
monetary base controlled by the central bank and as the liabilities of the banking system, is a
passive by-product. With no active role for money, integrating credit in the mainstream
framework has proven to be difficult.
3

Yet, as the mainstream was building and embracing the New Keynesian orthodoxy, there
was a nagging concern that something had been missing from the models. On the fringe
were theoretical papers in which debt plays a key role, and empirical papers concluding that
the quantity of debt makes a difference.
4
The latest crisis has revealed the deficiencies of the
mainstream approach and the value of joining those once seen as inhabiting the margin.
In response to the challenge, macroeconomists are now working feverishly to put financial
stability policy on the same theoretical footing that exists for conventional monetary policy.

3
Indeed, there has been little significant progress in modelling financial frictions and credit since the model of
the financial accelerator of Bernanke et al (1999).
4
See eg Friedman (1987), Kiyotaki and Moore (1997) and Borio et al (2001).
3/34

They are working not only to understand the sources of systemic risk, but also on how to
measure it and mitigate it.
5
That means writing down models in which debt truly matters and
working through the implications.
Like a cancer victim who cannot wait for scientists to find a cure, policymakers cannot wait
for academics to deliver the synthesis that will ultimately come. Instead, authorities must do
the best they can with the knowledge they have. As they make their day-to-day policy
decisions, central bankers, regulators and supervisors need some understanding of the role
of debt in the economy. When is debt excessive? When should we worry about its level,
growth rate and composition?
Starting with the basics, once one begins thinking about fixed non-state-contingent
obligations bonds, loans and the like things get very complicated very fast. Why are loans
and bonds by far the most prevalent mechanism for shifting resources over time? Why arent
risks shared more equally among the various parties? And, when investors finance a boom,
why is it exclusively through this contractual form? The answers to these very important
questions are probably related to information asymmetries and tax treatment.
6
But rather
than getting bogged down, we simply note that the basic form of debt has remained
remarkably constant both over history and across countries, empires and legal systems.
As for its uses, borrowing allows individuals to smooth their consumption in the face of a
variable income. It allows corporations to smooth investment and production in the face of
variable sales. It allows governments to smooth taxes in the face of variable expenditures.
7

And it improves the efficiency of capital allocation across its various possible uses in the
economy. At least in principle, it should also shift risk to those most able to bear it.
And public debt, in particular, can help smooth consumption not only through the lifetime of
individuals who are currently alive, but also across generations. To the extent that future
generations will be richer than the current ones because they will have a combination of
more human capital and more productive technology a transfer from future to current
generations can raise societys intertemporal welfare.
8
Since part of the tax rise needed to
fund higher current consumption is postponed, public debt may rise, at least up to a point,
without growth necessarily slowing. Furthermore, government debt also provides liquidity
services, which can contribute to easing the credit conditions faced by firms and households,
thus crowding in private investment.
9

For all these reasons, financial deepening and rising debt go hand in hand with
improvements in economic well-being.
10
Without debt, economies cannot grow and
macroeconomic volatility would also be greater than desirable.
11

5
A prominent recent example is Woodford (2011).
6
For a discussion of the basics of information asymmetries, see Cecchetti and Schoenholtz (2011). On taxes,
see Myers (2001).
7
See Barro (1979).
8
See Cukierman and Meltzer (1989) for a formal model in which agents cannot leave negative bequests to their
children on their own, so they vote to raise public debt. The argument in favour of a backward
intergenerational transfer is strengthened if part of government debt is financing investment that will benefit
future generations. However, it is important to note that the model in Cukierman and Meltzer (1989) is
deterministic: agents who maximise their as well as their offsprings welfare know with certainty what their
future income will be. In reality, the risk that future generations income might turn out less than expected
should play an important role in restraining the rise in government debt.
9
See Woodford (1990).
10
Arguably, an increase in government debt may not necessarily be welfare-improving. Part of the observed
increase in public debt in industrial countries can also be ascribed to the common revenue pool problem:
those members of society that benefit from additional spending are not the same as those bearing the extra

4/34

But financial development is not some magic potion. The accumulation of debt involves risk.
As debt levels increase, borrowers ability to repay becomes progressively more sensitive to
drops in income and sales as well as increases in interest rates. For a given shock, the
higher debt, the higher is the probability of defaulting. Even for a mild shock, highly indebted
borrowers may suddenly no longer be regarded as creditworthy. And when lenders stop
lending, consumption and investment fall. If the downturn is bad enough, defaults, deficient
demand and high unemployment might be the grim result. The higher the level of debt, the
bigger the drop for a given size of shock to the economy. And the bigger the drop in
aggregate activity, the higher the probability that borrowers will not be able to make
payments on their non-state-contingent debt. In other words, higher nominal debt raises real
volatility, increases financial fragility and reduces average growth.
12

Hence, instead of high, stable growth with low, stable inflation, debt can mean disruptive
financial cycles in which economies alternate between credit-fuelled booms and default-
driven busts. And, when the busts are deep enough, the financial system collapses, taking
the real economy with it.
In principle, as highly indebted borrowers stop spending, less indebted borrowers or lenders
could take up the slack. For example, wealthy households could purchase goods at reduced
prices and cash-rich firms could invest at improved expected return. But they need not. As
Eggertson and Krugman (2011) point out, it is the asymmetry between those who are highly
indebted and those who are not that leads to a decline in aggregate demand. Those authors
suggest that, in order to avoid high unemployment and deflation, the public sector should
borrow to fill the spending gap left by private sector borrowers as the latter repair their
balance sheets.
13

But, while the argument put forward by Eggertson and Krugman (2011) is correct in principle,
even the capacity of the public sector to borrow is not unlimited. When a crisis strikes, the
ability of the government to intervene depends on the amount of debt that it has already
accumulated as well as what its creditors perceive to be its fiscal capacity that is, the
capacity to raise tax revenues to service and repay the debt. Fiscal authorities may become
constrained both in their attempt to engage in traditional countercyclical stabilisation policies
and in their role as lender of last resort during a financial crisis.
14
That is, high levels of public
debt can limit essential government functions.
15

cost of funding it. Even so, to the extent that costly tax increases are postponed, the increase in government
debt may, up to a point, not have an immediate negative impact on growth. (For an overview of the common
revenue pool problem, see eg Eichengreen et al (2011).)
11
See the survey in Levine (2005).
12
See Bernanke and Gertler (1990) for an early example of a full general equilibrium model based on this
intuition.
13
Despite the lack of satisfactory formal models, central banks have been aware for some time of the
importance of the distribution of debt and wealth across the economy, both for the conduct of monetary policy
and for its financial stability implications. For an examination of the role of the distribution of household debt in
the United States, see eg Dynan and Kohn (2007); and for a discussion of the role of household debt in the
United Kingdom, see eg Benito et al (2007) and Waldron and Zampolli (2010b).
14
Aghion et al (2011) find evidence that industries that rely more heavily on external finance or hold less tangible
assets tend to grow faster in OECD countries that implement more countercyclical fiscal policies.
15
When examining the effects of the stock of debt on growth, it is also important to consider the potential
interaction between the stock and flow of credit. The burden of debt and the risks associated with it depend on
the stock of accumulated debt. Knowing this, both lenders and borrowers may begin to restrain the future flow
of credit after the stock of debt has passed some critical point. A diminished flow of credit may, in turn, hamper
growth.
5/34

It is important to note that there is a clear interaction between public and private debt. As we
have seen during the recent crisis, when private borrowing has fiscal backing, default
increases public debt. And the ability of the public sector to sustain a given level of debt
depends on its ability to raise revenue or its fiscal capacity something that could become
compromised if the private sector is already highly indebted.
Our conclusion is that, at low levels, debt is good. It is a source of economic growth and
stability. But, at high levels, private and public debt are bad, increasing volatility and retarding
growth. It is in this sense that borrowing can first be beneficial, so long as it is modest. But
beyond a certain point, debt becomes dangerous and excessive. We now turn to a
description of trends in industrial country debt before moving to the empirical analysis, in
which we look for the turning point.
3. Rising debt: a preliminary examination
The past three decades have witnessed a remarkable rise in advanced country
indebtedness. In this section, we briefly review trends in household, non-financial corporate
and government debt. In what follows, we will refer to the total combined debt of these three
sectors as the total non-financial debt.
16

Trends in aggregate non-financial sector debt
Graph 1 shows the aggregate non-financial sector debt of advanced economies and its
composition since 1980.
17
Two facts stand out: first, total non-financial debt as a percentage
of GDP, as well as its sectoral components, have been rising steadily for much of the past
three decades (left-hand panel). Starting at a relatively modest 167% of GDP three decades
ago, total non-financial debt has reached 314% of GDP. Of this increase, governments
account for 49 percentage points, corporates for 42 percentage points, and households for
the remaining 56 percentage points.
The right-hand panel of Graph 1 shows an index of debt adjusted for inflation, offering a
slightly different perspective of debt evolution. Even adjusting for inflation does not change
the message: real corporate debt has risen by a factor of roughly 3 (an average annual
compounded growth rate of 3.8%); government debt by about 4 times (5.1% annual rate);
and household debt by 6 times (6.2% annual rate). Overall, real debt of the non-financial
sector in advanced economies has been growing steadily at a rate of slightly less than
4% for the past 30 years.
What these panels show is that the surge in non-financial debt preceding the recent crisis is
not a new phenomenon. It is merely the continuation of a trend that was ongoing over the
entire period for which we have been able to assemble comprehensive data a trend that
was also accompanied by significant changes in composition.

16
Ideally, we would prefer to measure either a stock relative to a stock or a flow divided by a flow. The former, a
conventional measure of leverage like the ratio of total assets to debt, would require data on assets. And the
latter, like a measure of debt burden to income, would require us to have data on debt servicing.
Unfortunately, the limits of available data precluded both of these approaches.
17
In what follows, we use the public sector to refer to the general government sector, and the private sector
corresponds to non-financial corporations and households. As a consequence, non-financial corporate
sector debt includes the debt of public non-financial corporations, which are controlled by governments but are
market producers. Household debt includes debt of non-profit institutions serving households. See Appendix 1
for more details.
6/34

Graph 1
Non-financial sector debt
As a percentage of GDP
1
Real levels, deflated by consumer prices
2

30
80
130
180
230
280
330
1985 1990 1995 2000 2005 2009
Total
Government
Private sector
Households
Corporate
100
200
300
400
500
600
700
1985 1990 1995 2000 2005 2009
1
Simple averages for 18 OECD economies.
2
Rebased to 1980 = 100; simple average of 16 OECD economies, including the
United States.
As shown in Graph 1, from 1995 to the middle of the last decade, public debt had been
relatively stable as a percentage of GDP. But this period of relative public sector restraint
was accompanied by a rapid rise in household and non-financial corporate debt. The right-
hand panel of Graph 1 makes the rather stark point that real household debt tripled between
1995 and 2010, dwarfing the accumulation of debt in other sectors of the economy.
Table 1 provides a country breakdown of the simple and GDP-weighted averages plotted in
Graph 1.
18
These data show that the build-up of total non-financial debt is not a development
confined to a few large economies. Instead, it is a common feature across the 18 countries in
our sample. Total non-financial debt now exceeds 450% of GDP in Japan, 350% in Belgium,
Portugal and Spain, and 300% in two thirds of the countries.
Although countries share a similar upward trend in total debt, there are differences. In many
European ones, corporate debt makes up more than 40% of the total. (In Belgium, Finland,
Norway, Spain and Sweden, corporate debt is more than half of the total non-financial debt.)
In others, households account for the largest share. Australia, Denmark and the Netherlands
are examples. While in Japan, Italy and Greece, it is public debt that is dominant. Our sense
is that both the level and composition of debt should matter for growth, something we
examine below.
One clear limitation of our dataset is that it starts in 1980. It is sufficient, however, to look
back at the history of the United States (for which long back data are easily available) to
understand how extraordinary the developments over the last 30 years have been. As
Graph 2 shows, the US non-financial debt-to-GDP ratio was steady at around 150% from the
early 1950s until the mid-1980s. In some periods, public debt was high, but then private debt
was low; while in others it was the reverse.
19
Since the mid-1980s, however, both public and
private debt have been moving up together.

18
Appendix 2 contains tables reporting a breakdown of total non-financial debt into its components: government,
non-financial corporate and household.
19
See Friedman (1981 and 1986) for a discussion.
7/34

Table 1
Household, corporate and government debt
as a percentage of nominal GDP
Levels Changes
2

1980 1990 2000 2010
1
198090
1990
2000 200010
United States 151 200 198 268 49 2 70
Japan 290 364 410 456 75 46 46
Germany 136 137 226 241 1 89 15
United Kingdom 160 203 223 322 43 20 99
France 160 198 243 321 37 45 78
Italy 109 180 252 310 71 72 58
Canada 236 278 293 313 42 15 20
Australia 128 174 185 235 46 11 49
Austria 162 178 205 238 16 27 32
Belgium 170 264 298 356 94 34 58
Denmark 259 336 77
Finland 146 173 222 270 26 49 48
Greece 92 139 195 262 47 55 67
Netherlands 205 265 294 327 60 29 33
Norway 256 334 78
Portugal 144 141 251 366 2 110 115
Spain 172 187 258 355 15 70 97
Sweden 219 289 320 340 70 31 21
Total of above
Median 160 192 251 322 45 40 58
Weighted average
3
172 218 246 306 47 28 61
Simple average 168 211 255 314 43 44 59
G7 177 223 264 303 45 41 55
Other advanced 160 201 249 321 41 46 61
Memo: Std deviation 50 64 54 43
1
Some figures refer to 2009.
2
In percentage points of GDP.
3
Based on 2005 GDP and PPP exchange rates.
Sources: OECD; national data, authors estimates.

Graph 2
Outstanding debt of US non-financial borrowers
0
50
100
150
200
250
300
1960 1970 1980 1990 2000 2010
Total
Public
1
Private
2
1
Sum of federal government and state and local governments.
2
Sum of household and non-financial business.
Sources: Board of Governors of the Federal Reserve System; Friedman (1981).

Why has debt been rising so steadily?
What can then explain the trend in non-financial debt in advanced economies? The
relentless accumulation of non-financial debt has coincided with some important institutional
8/34

and market developments. First, from the late 1970s onwards, restrictions on financial
market activity and lending had been progressively and systematically removed, increasing
opportunities to borrow. Combined with improvements in financial theory and information
technology, this liberalisation has led to an intensification of financial innovation. The ability
to price complex financial products is indeed a prerequisite for fabricating and selling them.
Second, starting in the mid-1980s and continuing until the start of the recent crisis, the
macroeconomic environment had grown more stable. The Great Moderation brought lower
unemployment rates, lower inflation rates and less uncertainty. Believing the world to be a
safer place, borrowers borrowed more, lenders lent more and inflation remained low. There
was also a likely feedback here: as financial innovation improved the stability of credit supply
and allowed risk to flow to those best able to bear it, it improved general economic stability.
20

Third, since the mid-1990s, the substantial decline in real interest rates has made it easier to
support ever higher levels of debt.
21
The reasons behind such reduction are controversial.
The most prominent hypothesis is that low long-term interest rates are a consequence of a
high preference for saving in emerging markets a preference that arose for a variety of
reasons, including a poor social safety net, ageing populations retirement needs, and a
desire for insurance after the East Asian financial crisis of the late 1990s.
22

Finally, tax policies may have played a role, if not in explaining the rapid rise in debt, at least
in making the level of debt higher than it would have been otherwise.
23
For instance, the
preferential treatment of interest payments encourages firms to issue debt a factor that
could be behind the rising corporate indebtedness we see in some countries. It may also play
a role in the rise in household debt, where generous tax relief for mortgage interest
payments, along with explicit subsidies and implicit guarantees, could have played a role in
expanding home ownership in some places.
24

Regardless of the cause, the consequences are clear. Over the past 30 years, debt has risen
relentlessly across the industrial world. Looking at the simple average, total non-financial
debt rose by 147 percentage points of GDP from 1980 to 2010. Of this, 38% (56 percentage
points) was accounted for by households, 29% (42 percentage points) was a consequence of

20
See Cecchetti et al (2005) for a summary of factors that accounted for the Great Moderation. Dynan et al
(2006) focus on the role of financial innovation.
21
Waldron and Zampolli (2010a) investigate the effect of demographic changes and of a reduction in the real
interest rate on house prices and household debt in a quantitative overlapping-generations model calibrated to
the UK economy.
22
This is also known as the global saving glut hypothesis (Bernanke (2005), Bernanke et al (2011)). In its latest
incarnation, the hypothesis includes not only high saving rates but also a dearth of investment opportunities as
a possible explanation for low global real interest rates. This is why researchers normally refer to excess or
net savings, to emphasise the gap between saving and investment that opened up in a number of countries
(see eg Obstfeld and Rogoff (2009)). For a critical view of the global saving glut hypothesis, see eg Borio and
Disyatat (2011) and Shin (2009), both of which favour an explanation based on the procyclical behaviour of
financial institutions leverage. See also Modigliani and Cao (2004), who argue that the high saving rate
experienced by China is not due to a different preference for saving but is mainly a phenomenon driven by
that countrys high growth rate.
23
See eg Keen et al (2010).
24
Demographic changes in some advanced countries may also have contributed in the last few decades to
boosting household debt. Typically, the demand for housing is higher among middle-aged households. Thus
housing demand, and hence household debt, may have risen with the baby boom generation going through
middle age (as well as higher immigration rates and the continued rise in the number of households due to
higher divorce rates, etc). Yet demographic explanations in support of house prices are not new and they are
suspicious. They have been used in the past to justify unjustifiable housing booms. See eg Akerlof and Shiller
(2010).
9/34

additional corporate borrowing, and a third (49 percentage points) represents increases in
public debt.
4. The impact of debt on economic growth
Debt has been rising for decades, and economies have been growing. And, with high levels
of debt, policymakers are counting on robust growth to ensure sustainability. Without rising
GDP, there will be no way to raise the revenues governments need to reduce their exploding
debts. But now, debt is rising to points that are above anything we have seen, except during
major wars. Have we come to the point where debt levels are so high that they are harming
medium- and long-term growth?
We now turn to an empirical investigation of this question, starting with some simple statistics
concerning the macroeconomic link between debt and growth, and then run some more
sophisticated panel regressions in an effort to detect the impact of debt on growth.
Preliminary evidence
We examine annual data on GDP per capita and the stock of non-financial sector debt for a
group of 18 OECD countries over the period 19802006.
25
The novelty of our dataset is the
inclusion of private debt for a large number of industrial countries as well as its breakdown
into non-financial corporate and household debt. Since we are interested in trend growth, we
choose to end our sample in 2006, the year prior to the beginning of the latest financial crisis.
Table 2
Correlation of debt with annual per capita GDP growth
Total non-financial debt 0.0199***
(0.000)
Government debt 0.0026
(0.594)
Private debt 0.0197***
(0.000)
Corporate debt 0.0204***
(0.000)
Household debt 0.0254***

(0.004)
The table reports simple correlations, computed using ordinary least squares, of the annual per capita growth
rate with various definitions of debt. Asymptotic p-values for the test that the correlation is equal to zero are in
parentheses. -/--/--- indicate correlations significantly different from zero at the 10/5/1% level.
Source: Authors calculations.

We begin by looking at the relationship between debt and growth, in terms of both level and
volatility. Table 2 reports correlations based on bivariate least squares regressions of annual
per capita GDP growth on various measures of aggregate non-financial debt. Table 3 reports

25
The countries in our dataset are: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany,
Greece, Italy, Japan, the Netherlands, Norway, Portugal, Spain, Sweden, the United Kingdom and the United
States. All debt series begin in 1980 with the exception of the household debt series of Denmark and
Norway. Appendix 1 describes in detail how the dataset has been constructed.
10/34

equivalent results for the (overlapping) standard deviation of the five-year-ahead growth rate
of per capita GDP, a measure of aggregate volatility.
26
In both cases, we include country and
time-period fixed effects.
Starting from Table 2, we see that there is a negative within-country (or time-series)
correlation between growth and total non-financial debt. Looking at the details, we see that
both non-financial corporate and household debt display a statistically and economically
significant negative correlation with growth. For corporate debt, a 1 percentage point
increase is associated with an approximately 2 basis point reduction in per capita GDP
growth. For household debt, the impact is even larger: a 1 percentage point rise in household
debt-to-GDP is associated with a 2 basis point reduction in growth.
27

Turning to Table 3, we see that public debt has no statistically (or economically) significant
relationship with the future volatility of aggregate growth. But, for corporate plus household
debt combined, a 10 percentage point increase in debt leads to an increase in the standard
deviation of future growth of about 0.10 percentage points. Corporate debt, rather than
household debt, appears to be driving this result.
Table 3
Correlations of per capita GDP volatility with debt
Total non-financial debt 0.0042
(0.331)
Government debt 0.0057
(0.301)
Private debt 0.0103***
(0.007)
Corporate debt 0.0102**
(0.036)
Household debt 0.0156

(0.293)
The regressions are panel regressions with both country and period fixed effects. The dependent variable is
the standard deviation of future annual growth rates of per capita GDP over the following five years.
Observations are overlapping, so robust standard errors are reported. Debt variables are shares of GDP.
Numbers in parentheses are asymptotic p-values for the test that the coefficient estimate is equal to zero
computed using standard errors estimated using the Huber-White sandwich estimator. */**/*** indicate
coefficient estimates significantly different from zero at the 10/5/1% level.

These results are somewhat surprising, as we do not believe that debt in the generally low
range found in our sample is uniformly bad for growth. There are a variety of explanations for
this simple correlation. The most obvious is that simple correlations are masking the effects
of other common influences. With that in mind, we turn to an analysis based on somewhat
more sophisticated growth regressions.

26
That is, for each period t we compute the standard deviation of the annual growth rates of per capita GDP for
period t+1, t+2, and t+5.
27
We also ran these regressions using an estimator that is robust to the presence of outliers. In all cases, the
association with growth is slightly less strong but remains both statistically and economically significant.
Dropping fixed country and time effects from the regressions generally leads to a non-significant relationship
between growth and measures of debt. Using five-year-ahead averages of per capita growth instead of annual
observations leaves these conclusions largely unchanged.
11/34

Growth regressions: the model
Our empirical strategy proceeds in two steps. First, we specify and estimate a growth
equation based on the empirical growth literature (eg Barro and Sala-i-Martin (2004)). Then
we add various measures of non-financial debt to see whether they have an impact on
growth over and above other determinants.
The empirical specification is derived from the neoclassical growth model of Solow, in which
per capita income growth depends on the initial level of physical and human capital,
converging to its steady state rate slowly over time. In turn, the steady state depends
positively on the saving rate and negatively on the growth rate of the labour force, in addition
to a number of parameters describing the technology and the preferences of the country.
Earlier empirical studies of growth focused on regional or cross-country differences. Here,
instead, we follow Islam (1995) and others in estimating panel data regressions with country-
specific fixed effects (as well as time-specific fixed effects). This allows us to measure the
impact of a change in one factor on growth within a country.
28

Another important aspect of our empirical analysis is that we use overlapping five-year
forward averages of the per capita income growth rate. The use of five-year averages,
common in the growth literature, reduces the potential effects of cyclical movements and
allows us to focus on the medium-term growth rate.
29

Turning to the details, we model the growth rate of per capita income for country i as:
k t t i t i t i t i k t t i
X y g
+ + +
+ + + + =
, , , , , 1 ,
' c | | , (1)
where
( )
t i k t i
k t
t j
j i k t t i
y y
k
g
k
g
, ,
1
, , 1 ,
1 1
~ =
+
+
+ =
+ +

is the k-year forward average of annual growth rates between year t+1 and t+k and y is the
log of real per capita GDP. In our analysis, we set k=5. The regressors in equation (1)
include: the log of real per capita GDP at time t (to capture the catch-up effect or conditional
convergence of the economy to its steady state); a set of other regressors, X, thought to
explain growth; country-specific dummies,
i
; time-specific dummies meant to capture
common effects across time,
t
(eg global business cycle conditions that will affect all
countries and so on); and residuals c.
It is important to note that, in order to minimise the potential for the endogeneity bias (and the
problem of reverse causation), all regressors (with the exception of the population growth
rate) on the right-hand side of (1) are predetermined with respect to the five-year forward
average growth rate. Furthermore, the overlapping nature of the data imparts a moving-

28
Our choice of looking at within-country or time effects is essentially dictated by the relatively small number of
countries in our database (N=18), which precludes a sensible analysis of cross-country differences. But using
panel data with fixed effects also has considerable advantages. Indeed, it is possible to control for
unobservable differences between countries using simple dummies, provided such heterogeneity is assumed
to be constant over time. For example, countries may differ in terms of their legal and institutional system,
culture, religions, etc. While these characteristics are difficult to measure, to the extent that they do not change
over time they will be captured by country-specific dummies.
29
We use overlapping observations both to increase efficiency and to avoid what would be the arbitrary
construction of five-year non-overlapping blocks. Indeed, one common practice is to take five-year averages
over 198084, 198589, 199094, and so forth. However, there is no reason why one cannot choose different
periods, eg 198286, 198791, etc. Given that we use overlapping averages of five years forward and that our
sample ends in 2006, our last observation is 2001. We note that overlapping data are also used by Bekaert et
al (2001, 2005) in their study of the effects of financial liberalisation on economic growth.
12/34

average process to the errors, so that we use a robust procedure to compute the standard
errors of our coefficient estimates.
30

The list of regressors in our specification includes:
- gross saving (public and private) as a share of GDP;
- population growth;
- the number of years spent in secondary education, a proxy for the level of human
capital;
31

- the (total) dependency ratio as a measure of population structure and ageing;
- openness to trade, measured by the absolute sum of exports and imports over GDP;
- CPI inflation, a measure of macroeconomic stability;
- the ratio of liquid liabilities to GDP, as a measure of financial development;
32
and
- a control for banking crises taking the value of zero if in the subsequent five years
(as identified by Reinhart and Rogoff (2008)) there is no banking crisis, and the
value of 1/5, 2/5, and so forth, if a banking crisis occurs in one, two, etc, of the
subsequent five years.
Before turning to the results, two additional points are worth emphasising. First, given the
difficulty of measuring the physical capital stock and the link that should exist between capital
and income, one can also interpret the log of real per capita GDP on the right-hand side of
(1) as a proxy for physical capital. And second, related to the crisis variable, we note that
high levels of debt for a country as a whole or for one of its sectors may be a reason why a
country may end up facing a banking crisis. But it may also be the reason why a given
downturn, originating from events outside the country or the indebted sector, may turn out to

30
The presence of the log real GDP among the regressors makes (1) a dynamic fixed-effects panel data model.
This means that, in principle, the estimates from using the least squares dummy variable (LSDV) estimator
may be biased. Starting with Nickell (1981), several researchers have shown that the coefficient on the lagged
dependent variable is biased downwards even if N (the number of units in the panel) goes to infinity. The
estimates are, however, consistent in T (the time dimension of the panel), with the bias of the order O(1/T). A
number of alternative estimators to the LSDV have been proposed in the literature to overcome the finite
sample bias. These are based on the instrumental variable estimation (Anderson and Hsiao (1981)) or
generalised method of moments (GMM) (Arellano and Bond (1991), Arellano and Bover (1995)). Yet
instrumental-variable (IV) and GMM techniques normally work well when the number of panel units is large
and the time dimension small. These techniques are also subject to drawbacks which do not make them
suitable for application to the typical macroeconomic panel data with moderate N and T. First, IV estimators
are generally less efficient than ordinary least squares (OLS). Second, instrumental variables may only be
weakly correlated with the instrumented variables. As a result of this, estimates obtained through IV
techniques may also be biased and tests of hypothesis may have low power (Stock et al (2002)). Most
importantly, Monte Carlo simulations show that even for moderate sample size of T=30 and N=20 the typical
size of a macroeconomic panel the LSDV estimator performs as well as or better than the instrumental
variable estimators (Judson and Owen (1999)). In our case, N=18 and T=25. We therefore follow the
suggestion of Judson and Owen (1999) and use the LSDV estimator.
31
We also tried to include life expectancy at birth and fertility rates as potential proxies for human capital, but
these variables turn out to be largely irrelevant in our sample of advanced economies when we control for the
other determinants of growth.
32
We also experimented with other measures of financial depth, such as stock market capitalisation (as
constructed by the World Bank; see Beck et al (2000) and Beck and Demirg-Kunt (2009)). Stock market
capitalisation has some explanatory power, but unfortunately it is available only from 1989 for most countries.
We also experimented with the flow of private credit to GDP and the flow of public borrowing to GDP
(constructed as the change in the stock of debt divided by GDP) for further details see below.
13/34

be worse than it could have been otherwise. Using this variable thus allows us to check
whether or not the effects of debt on growth are related to periods of financial stress.
33

Growth regressions: the estimates
Table 4 reports our basic growth regression, along with summary statistics for the variables
included. As noted above, the presence of overlapping data imparts serial dependence in the
country-specific residuals, so we use a robust procedure to estimate the standard errors of
the estimated coefficients in (1).
34

All coefficients have the expected sign and are in most cases statistically significantly
different from zero at conventional levels. The size of the coefficients also conforms to those
reported in earlier studies. In particular, the rate of convergence is approximately 15% a
year, consistent with what is found in prior panel data studies. The saving rate has a positive
effect on growth, although imprecisely estimated. (The precision increases when we add the
debt variables.
35
) Furthermore, consistent with the theory, the change in the population has a
negative (albeit imprecisely estimated) impact on growth. And, as we expect, banking crises
are associated with lower growth. The estimates imply that, for each additional year spent in
crisis, average growth falls by an average of 27 basis points over the proceeding five years.
36

Before moving on, we note that the dependency ratio has a strongly negative and statistically
significant impact on subsequent growth. Our estimates imply that a one standard deviation
increase in the total dependency ratio an increase of roughly 3.5 percentage points is
associated with a 0.6 percentage point reduction of future average annual growth.
Growth regressions: the impact of debt
We now add the debt variables to these regressions. The results are in Table 5. The top
panel reports the results of the model that controls for banking crisis, while the middle and
bottom panels examine two important variants on this. Starting at the top, the first five
columns show the consequences of adding one variable at a time, while the remaining five
columns report the results for various combinations. The tables in Appendix 3 report the full
details of the regressions, including estimated coefficients for the control variables.

33
We also attempted to control for the size of the country, as proxied by the log of its population, and for the size
of government, as proxied by government consumption as a share of GDP. These variables generally turn out
to be statistically insignificant when most of the other regressors are included. Moreover, their inclusion or
exclusion has no noticeable effects on the other regression coefficients.
34
Specifically, we employ the Huber (1967)-White (1980) sandwich estimator. In the computation of the robust
standard errors, observations are clustered by country (Rogers (1993)). We find that the standard errors
computed using this procedure are much larger than the ordinary ones, indirectly confirming the presence of
heteroskedasticity and time dependence in the estimated residuals. We also computed Newey-West standard
errors, but these turn out to be smaller even when specifying an autocorrelation lag as large as 10 (which
should take care of the moving-average structure imparted by the overlapping nature of the observations plus
any other additional autocorrelation inherent in the choice of our empirical specification).
35
Table A3.1 in Appendix 3 shows how the estimates of the coefficients change as debt variables are added to
the basic growth model. Tables A3.2 and A3.3 show how the regressions change as financial flow variables
are added to the model or as the crisis variable is dropped, respectively.
36
The variable in the regression takes the value of 1/5, 2/5, and so forth, if a banking crisis occurs in one, two,
etc, of the subsequent five years. So, if there is a crisis in one year, then we multiply the coefficient 0.0134 by
0.2, which gives 0.0027, or 27 basis points.
14/34

Table 4
Basic growth regression
Variables Summary statistics

Coefficient Mean Standard
deviation
Dependent variable:

Five-year forward mean growth rate of per capita
GDP

0.0210 0.0119

Independent variables:

National gross saving (as a share of GDP) 0.0409 0.2193 0.0447

(0.131)
Change in population 0.4482 0.0050 0.0038

(0.233)
Schooling 0.0051*** 8.8262 1.8540

(0.001)
Log of real per capita GDP 0.1565*** 10.1601 0.2521

(0.000)
Trade openness 0.0311** 0.6203 0.2922

(0.019)
Inflation rate 0.0049 0.0471 0.0476

(0.787)
Total dependency ratio 0.1955*** 0.5015 0.0342

(0.000)
Liquid liabilities (as a share of GDP) 0.9935 0.0075 0.0035

(0.170)
Banking crisis 0.0134*** 0.1616 0.2936

(0.000)
Reported coefficients are for the marginal impact of debt on the five-year forward average per capita growth
rate from estimating text equation (1). Numbers in parentheses are asymptotic p-values for the test that the
coefficient estimate is equal to zero computed using standard errors estimated using the Huber-White
sandwich estimator. */**/*** indicate coefficient estimates significantly different from zero at the 10/5/1% level.
Looking at these estimates, two facts stand out. First, total non-financial debt has a negative
impact but the p-value is a relatively large 0.177. Second, when we disaggregate debt, we
see that public debt has a consistently significant negative impact on future growth. And, the
impact is big: a 10 percentage point increase in the ratio of public debt to GDP is associated
with a 1718 basis point reduction in subsequent average annual growth. For corporate and
household debt, estimates are very imprecise, so we are unable to come to any real
conclusions.
To see whether there are specific factors driving the results in the top panel of Table 5, we
examine two alternatives. First, in the middle panel of Table 5 we drop the banking crisis
variable in an effort to see whether the effects of debt depend on periods of financial stress.
Government debt continues to be negatively associated with subsequent growth, although
the effect is generally slightly smaller. But, the remarkable fact from the middle panel of
Table 5 is that corporate debt is now negatively related to future growth. Specifically, when
corporate debt is added on its own (column 4), the effect of corporate debt on growth is
negative and economically large (an 8 basis point decline per 10 percentage point increase),
15/34

Table 5
Effect of debt on future growth
Regressions (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Controlling for
banking crises

Total 0.0078
(0.177)
Government 0.0167*** 0.0180** 0.0174*** 0.0175*** 0.0180**
(0.007) (0.011) (0.009) (0.008) (0.010)
Private sector 0.0016 0.0023
(0.824) (0.701)
Corporate 0.0006 0.0030 0.0023 0.0008

(0.938) (0.629) (0.745) (0.91
6)
Household 0.0050 0.0033 0.0027 0.0047

(0.716) (0.789) (0.821) (0.71
7)

Not controlling
for banking
crises
Total 0.0116**
(0.025)
Government 0.0164** 0.0169** 0.0191*** 0.0136* 0.0165**
(0.025) (0.032) (0.006) (0.093) (0.030)
Private sector 0.0054 0.0093**
(0.279) (0.046)
Corporate 0.0082 0.0117** 0.0109*
0.0078

(0.163) (0.028) (0.058) (0.19
4)
Household 0.0023 0.0043 0.0013

0.0055

(0.870) (0.756) (0.923) (0.70
9)

Including
financial flow
variables
Total 0.0103*
(0.051)
Government 0.0208*** 0.0240*** 0.0226*** 0.0218*** 0.0240***
(0.000) (0.000) (0.001) (0.000) (0.000)
Private sector 0.0030 0.0051
(0.597) (0.300)
Corporate 0.0027 0.0043 0.0054 0.0023

(0.689) (0.459) (0.377) (0.75
6)
Household 0.0065 0.0047 0.0041 0.0057

(0.554) (0.632) (0.675) (0.61
0)
Reported coefficients are for the marginal impact of debt on the five-year forward average per capita growth rate from estimating
text equation (1). Numbers in parentheses are asymptotic p-values for the test that the coefficient estimate is equal to zero
computed using standard errors estimated using the Huber-White sandwich estimator. */**/*** indicate coefficient estimates
significantly different from zero at the 10/5/1% level.
but fairly imprecise (with a p-value of 0.16). However, when we control for government debt
(column 7), the impact is strongly large and precisely estimated: a 10 percentage point
increase in corporate debt is associated with a reduction in subsequent average growth of
16/34

1112 basis points (and the p-value is 0.028). These findings suggest that corporate debt
makes growth worse during periods of financial stress. The fact that the impact is stronger
when we control for government debt means that high levels of private debt, in the presence
of large government debts, make the economy more vulnerable to shocks.
37

Finally, we check to see if the outstanding levels of public and private debt are capturing the
effects that high debt might have on the future flow of credit and, through this channel, on
growth. To do this, we add the average flow of private credit and government borrowing
(always as a share of GDP) to the regressions. The results are reported in the bottom panel
of Table 5. The existence of what might be termed the crowding-out effect, whereby higher
debt reduces the future availability of credit, is confirmed by separate regressions, in which
controlling for both country and period fixed effects we find (not reported here) that higher
levels of both public and private debt are negatively associated with future credit flows.
As for the consequences for the level of debt itself, the object of our primary interest,
controlling for the credit flow variables makes the estimated effects of total non-financial debt
and government debt larger (in absolute value) and more precise.
Threshold effects
In our preliminary discussions, we noted the possibility that, as it increases, indebtedness
can turn from good to bad from initially growth-enhancing (or neutral) to eventually growth-
reducing. Our interest is in looking for this effect in the data. We do this first by looking at a
simple picture and then at some statistical results.
Graph 3
Non-financial sector debt and output growth for 18 OECD countries
0
100
200
300
400
500
1.25
1.50
1.75
2.00
2.25
2.50
1
st
quartile 2
nd
quartile 3
rd
quartile 4
th
quartile
Non-financial debt, as a percentage of GDP
1
(lhs)
GDP, growth
2
(rhs)
1
The average of the quartiles.
2
Average growth within the quartiles.
Sources: OECD; Penn World Tables 7.0.

The picture is in Graph 3. We have split the sample of observations on per capita GDP
growth based on the distribution of the debt-to-GDP ratio. Mean per capita GDP growth rises

37
We also added government consumption as a share of GDP to the regressions in order to check whether the
effects of public debt would change. If government consumption has a negative impact on growth, then
omitting it may lead us to incorrectly conclude that public debt has a negative impact over and above the
effects of distortionary taxation and other disincentives imposed on the private sector, which might be proxied
by government consumption. It turns out that government consumption is always statistically insignificant in
our growth regressions. We also dropped the saving rate variable to check whether debt has a negative
impact on growth through capital accumulation. The estimates are little changed.
17/34

as we move from the first to the third quartiles and then falls back in the fourth quartile.
Moreover, the difference between growth in the bottom and the top quartiles is small, so
there is no evidence that debt is necessarily bad for growth. Indeed, for sufficiently low
levels, debt may help foster capital deepening and allocative efficiency, thus boosting growth.
Yet the graph also suggests that, as debt approaches high levels, the effect of further
increases in debt on growth may begin to subside.
38
Although the graph suggests that there
is no big difference between high- and low-debt economies in terms of mean growth, it might
be that more sophisticated statistical techniques are needed to bring out any larger negative
effect on growth from high debt.

Turning to the regression, we ask whether the relationship between growth and debt is non-
linear. We look for this using the following empirical model, which incorporates threshold
effects but is otherwise identical to equation (1):
( ) ( )
k t t i t i t i t i t i t i t i t i k t t i
d I d d I d X y g
+ + + +
+ + + > + < + + =
, , , , , , , , , 1 ,
' c t t | | , (2)
where I(.) is an indicator variable that takes the value of 1 if debt is below a given threshold,
t; and zero otherwise. The indicator variable has the effect of splitting the debt variable into
two, allowing for the impact to differ above and below the threshold.
We look for threshold effects by including one debt variable at a time in equation (2). To
estimate the threshold, we estimate equation (2) for a series of values of debt-to-GDP and
then select the one that minimises the sum of squared residuals. To examine the statistical
significance of the estimated threshold, we can then use a likelihood ratio (LR) statistic,
computed as the difference between the sum of squared residuals of the model for a generic
value of the threshold and the sum of squared residuals corresponding to the estimated
threshold (scaled by the variance of the sample residuals).
39

We can illustrate how this procedure works for the case of public debt in the model with the
crisis variable. (In the absence of the threshold, this is the case reported in column 2 of the
top panel of Table 5.) Graph 4 reports the results for the LR statistic in this case. By
definition, this statistic equals zero at the estimated threshold level. The graph shows that
96% of GDP is the point estimate of the threshold level. At the 1% confidence level, the
threshold level lies between 92 and 99% of GDP that is, the level at which we estimate that
public debt starts to be harmful to growth may be as low as 92% of GDP and as high as 99%
(using 5% or 10% confidence levels would not change the interval much).

38
The results are mostly unaltered when we use the median instead of the mean debt and growth values. We
also split the annual observations using the debt brackets employed by Reinhart and Rogoff (2010) in their
recent analysis of public debt and growth. Using a different dataset, which covers 20 advanced economies
over 19462009, they find that over the 90% threshold public debt tends to be associated with much lower
GDP growth than at lower levels: GDP growth is about 1 percentage point lower at the median and almost
4 percentage points lower at the mean compared with the lowest debt burden group (debt ratios less than
30% of GDP) (see their Figure 2). By contrast, in our sample and using the same brackets for the debt-to-GDP
ratio as Reinhart and Rogoff (2010), we do not find any significant differences in either the mean or median
growth. Our sample, however, is much more limited than theirs: it starts from a much later date and excludes
the latest financial crisis; it covers two countries fewer; and it uses GDP per capita growth rather than GDP
growth, and general government debt rather than central government debt.
39
Hansen (1999) has developed threshold regression methods for non-dynamic panel data models with fixed
effects. Since we are not aware of well established methods to estimate threshold effects in dynamic panel
data models, we follow his suggested inference methods.
18/34

Graph 4
Likelihood ratio statistic
0
10
20
30
40
50 57 64 71 78 85 92 99 106 113 120
Government debt threshold
L
R
LR Statistic 1% Critical value
5% Critical value
10% Critical value
The LR statistic is constructed as ( ) ( ) ( ) ( )
2
/ o t t t SSR SSR LR =
, where ( ) t t SSR min arg = ; SSR is the sum of squared
residuals obtained by estimating text equation (2) for different values of the threshold variable.

Table 6 reports results from estimating this threshold model for government, corporate and
household debt separately, with and without the crisis variable. Focusing on the results
where we do not control for crises, we estimate the threshold for government debt at 84% of
GDP. And, when government debt rises to this level, an additional 10 percentage points of
GDP drives trend growth down by some 1015 basis points.
Before continuing, it is worth noting that the impact of public debt on growth could in part
reflect the quality of government. That is, poor governments do a number of things that slow
their economies, and debt is a consequence. We note, however, that because we include
country fixed effects, it would have to be a deterioration in the quality of governance that was
responsible. Unfortunately, we have found no straightforward way of controlling for bad
government.
40

Turning to corporate debt, and again focusing on the results where we do not control for
banking crises, we find two thresholds. One is around 75% and the second is close to 90%.
41

Our estimate is that there is a range between the two thresholds over which
accumulation of corporate debt is relatively neutral. But once that debt reaches the higher of
the thresholds, there is a negative impact on growth. The coefficients suggest, however, that
the economic impact is in the order of half that for government debt.
Finally, there is household debt. The results suggest that we have pushed the data to the
limit. While we find a threshold of 84%, and that the impact of household debt on growth is
first positive and then negative, our estimates lack statistical precision. In fact, the p-values
for the test of whether the coefficients are zero is nearly one half. So, while we may believe
that there is a point beyond which household debt is bad for growth, we are unable to reliably
estimate that point using the historical record available to us.

40
One possibility would be to use the size of government itself, on the reasonable assumption that bad
government leads to bloated public expenditure. However, well run societies may well opt for larger
governments. Corruption measures seem more promising. For example, Kaufmann (2010) provides intriguing
evidence that industrial countries budget deficits over the period 200609 were negatively associated with
measures of perceived corruption. We leave further examination of this issue for future research.
41
We compute the threshold point estimates sequentially. We first look for multiple minima in the sum of
squared residuals of the estimated model. If we find more than one minimum, we fix the first point and repeat
the search for a new point that minimises the sum of squared residuals, and so forth.
19/34

Table 6
Threshold effects

Threshold estimate Coefficients
Government debt

Controlling for crises 96% <96% >=96%

0.0065 0.0138***

(0.232) (0.004)
Not controlling for crises 84% <84% >=84%

0.0074 0.0133*

(0.382) (0.057)

Corporate debt


0.0119 0.0047

(0.156) (0.474)
Controlling for crises
(2 threshold points)
73%; 99% <73% >=73% &
<99%
>=99%

0.0055 0.0019 0.0038

(0.151) (0.399) (0.208)
Not controlling for crises
(2 threshold points)
73%; 88% <73% >=73% &
<88%
>=88%

0.0041 0.0044 0.0059**

(0.221) (0.260) (0.041)
Household debt


0.0069 0.0065

(0.618) (0.658)
Not controlling for crises 84% <84% >=84%

0.0049 0.0115

(0.733) (0.458)
Reported threshold estimates are obtained by minimising the sum of squared residuals in text equation (2). Reported
coefficients are for the marginal impact of debt on the five-year forward average per capita growth rate from estimating text
equation (2). Numbers in parentheses are asymptotic p-values for the test that the coefficient estimate is equal to zero
computed using standard errors estimated using the Huber-White sandwich estimator. */**/*** indicate coefficient estimates
significantly different from zero at the 10/5/1% level.

5. Prospects and challenges
Several industrial countries already have debt levels that, according to the empirical
evidence presented in the previous section, might be growth-damaging. Or, they soon will
be.
As we noted in a previous paper (Cecchetti et al (2011)), public debt ratios are currently on
an explosive path in a number of industrial countries. To prevent further deterioration, these
countries will need to implement drastic policy changes that reduce current deficits, as well
as future contingent and implicit liabilities. Yet stabilisation might not be enough, especially if
it is at a level high enough to damage growth.
20/34

Unfortunately, the unprecedented acceleration of population ageing that many industrial
countries now face may make this task even more difficult. First, ageing drives government
expenditure up and revenue down, directly worsening debt. But, as our results in the
previous section suggest, there is an additional effect: rising dependency ratios put further
downward pressure on trend growth, over and above the negative effects of debt.
Graph 5 reports dependency ratios, measured as the young and old in society (the non-
working-age population) as a percentage of the working-age population, for advanced and
emerging market economies. The left-hand panel of the graph shows that a majority of
industrial countries are now close to a turning point similar to the one experienced by Japan
in the early 1990s. After having declined and remained relatively stable, total dependency
ratios will increase rapidly in these countries over the next few decades.
Emerging market economies are also ageing. But, with the exception of Central and Eastern
Europe, they lag advanced economies by at least two to three decades. This means that
these economies will continue to enjoy a demographic dividend: as they catch up with richer
economies, their young workforces should continue to support strong growth and saving.

Graph 5
Total dependency ratio
1

Advanced economies Emerging economies
0
20
40
60
80
100
1975 2000 2025 2050 2075 2100
More developed regions
Europe
Japan
United States
0
20
40
60
80
100
1975 2000 2025 2050 2075 2100
Less developed regions
Central and Eastern Europe
Latin America and the Caribbean
Africa
Asia
1
Non-working-age population as a percentage of working-age population.
Source: United Nations, World Population Prospects, 2010 revision.

Recent studies have combined the implications of current fiscal deficits with the estimates of
future increases in health and pension spending in an effort to project public debt to
2040. While they differ in their optimism, these studies all show that, under unchanged fiscal
policy, debt-to-GDP ratios will explode in all but a few countries.
42

The consequences are striking. Debt quickly rises to more than 100% of GDP a level
clearly consistent with negative consequences for growth. And, in a number of countries
Japan, the United Kingdom and the United States the projections rise much further. In the
euro area, the public debt ratio will also rise, albeit less rapidly than in the UK or US,
reflecting the fact that many countries face only a modest rise in the future costs of ageing.
In addition to putting further pressures on public finances, ageing itself might also reduce per
capita growth, making it potentially even more difficult for a country to sustain a given level of

42
See eg Auerbach (2011), Cecchetti et al (2011), Gagnon (2011) and IMF (2011).
21/34

debt. With unchanged public policies, the ever greater amount of resources that will be
channelled to the elderly through pension and health care spending will increase.
Furthermore, older people save less than people in younger age groups. The exact timing at
which saving might be reduced and the impact on real interest rates are controversial,
depending on public policies and saving in the emerging world, among other things.
43,

44

That said, the fact that ageing is asynchronous around the world may help more advanced
and highly indebted countries to smooth the consequences. There are at least three reasons
for that. First, immigration can partly slow the shrinking of labour forces in advanced
economies. Second, as incomes and wealth rise, emerging economy savings may continue
to flow to countries with more advanced financial markets and lower-risk assets, keeping
interest rates down and permitting their capital stocks to grow.
45
And finally, trade may also
reduce the need for more radical changes in the composition of demand that ageing might
otherwise bring with it.
Such benefits of globalisation should help countries adopt the necessary reforms needed to
reduce their public debt while at the same time helping the private sector through the
abundance of the supply of savings and the continuous low real interest rates globally to do
the necessary post-crisis balance sheet repair.
6. Conclusions
While the attention of policymakers following the recent crisis has been on reducing systemic
risk stemming from a highly leveraged financial system, the challenges extend beyond that.
Our examination of debt and economic activity in industrial countries leads us to conclude
that there is a clear linkage: high debt is bad for growth. When public debt is in a range of
85% of GDP, further increases in debt may begin to have a significant impact on growth:
specifically, a further 10 percentage point increase reduces trend growth by more than one
tenth of 1 percentage point. For corporate debt, the threshold is slightly lower, closer to
90%, and the impact is roughly half as big. Meanwhile for household debt, our best guess is
that there is a threshold at something like 85% of GDP, but the estimate of the impact is
extremely imprecise.
A clear implication of these results is that the debt problems facing advanced economies are
even worse than we thought. Given the benefits that governments have promised to their
populations, ageing will sharply raise public debt to much higher levels in the next few
decades. At the same time, ageing may reduce future growth and may also raise interest
rates, further undermining debt sustainability. So, as public debt rises and populations age,
growth will fall. As growth falls, debt rises even more, reinforcing the downward impact on an
already low growth rate. The only possible conclusion is that advanced countries with high

43
In theory, ageing has an ambiguous effect on capital intensity. The reduction of labour forces might increase
capital-to-labour ratios. Indeed, some studies suggest that ageing at the global level will continue to put
downward pressure on global real interest rates for many years to come (see eg Attanasio et al (2007) and
Krueger and Ludwig (2007)). On the other hand, with unchanged policies, the need to finance ever larger age-
related spending may lead to a shortage of capital, which would put upward pressure on real interest rates
(see eg Fehr et al (2005)).
44
Ageing may also adversely affect asset prices. For example, a recent study by Takts (2010) finds that ageing
may have a negative impact on house prices.
45
On emerging markets demand for safe assets, see Caballero et al (2008) and Caballero (2010). More
generally on the ex ante excess saving in emerging markets and its implications for global real interest rates,
see Bernanke (2005) and Bernanke et al (2011).
22/34

debt must act quickly and decisively to address their looming fiscal problems. The longer
they wait, the bigger the negative impact will be on growth, and the harder it will be to adjust.
It is important to note that our finding of a threshold for the effects of public debt on growth
does not imply that authorities should aim at stabilising their debt at this level. On the
contrary, since governments never know when an extraordinary shock will hit, it is wise to
aim at keeping debt at levels well below this threshold.
As with government debt, we have known for some time that when the private sector
becomes highly indebted, the real economy can suffer.
46
But, what should we do about it?
Current efforts focus on raising the cost of credit and making funding less readily available to
would-be borrowers. Maybe we should go further, reducing both direct government subsidies
and the preferential treatment debt receives. In the end, the only way out is to increase
saving.

46
See eg Tang and Upper (2010) for a study of debt deleveraging following systemic banking crises.
23/34

Appendix 1: Description and sources of data used in the paper
A. Non-financial sector debt
The time series constructed are generally taken from national balance sheet statistics (flow
of funds) as available either from the OECD website or from national sources/databases
(eg Federal Reserve Flow of Funds, Bank of Japan Flow of Funds). The target dataset is
annual frequency beginning in 1980, for 18 OECD countries. The countries included in the
sample are the following: Australia, Austria, Belgium, Canada, Denmark, Finland, France,
Germany, Greece, Italy, Japan, the Netherlands, Norway, Portugal, Spain, Sweden, the
United Kingdom and the United States.
The sectors covered are (i) households and non-profit institutions serving households
(NPISHs) (S14 + S15); (ii) non-financial corporations (S11); and (iii) general government
(S13), as defined in the System of National Accounts (SNA 2008) or its previous version
(SNA 1993). For simplicity, debt is defined as the following (for all apart from the United
States): gross liabilities for households and general government, and total liabilities less
shares and other equities for non-financial corporations. For the United States, the item
credit market instruments on the liabilities side is taken for each sector.
The debt series are mostly at market value and on a non-consolidated basis, so may differ
from other sources. For many countries (all except Canada, Japan, Spain and the United
States), the data under SNA 1993 are not available from 1980. Hence, these are
extended/backdated
47
using data from old compilations. For example, for Italy and Sweden,
old flow of funds data can be retrieved from their national websites. For France, Germany
and the United Kingdom, old data are taken from old national sources.
For some countries, flow of funds data do not go back far enough, so we use other
sources/proxy series. For non-financial corporations, an old OECD publication entitled Non-
financial enterprises financial statements and bank credit data have been used to backdate
the series.
48
For the household sector, bank credit/loans to households have been used. For
general government, the IMFs historical public debt database (2010) has been used and the
data gaps interpolated.
49, 50

B. Other data used in the panel model
Gross national savings sourced from IMF, World Economic Outlook. Data on population, real
GDP per capita and openness in current prices from Penn World Tables. Average number of
years spent in school of population aged 15 and over taken from Barro and Lee
(2000) (available only up to 2000 and every five years, interpolated by repeating the last
available value). Data on consumer prices, overall dependency ratio and liquid liabilities as a
share of GDP sourced from World Bank, World Development Indicators. Crisis dates taken
from Carmen M Reinhart.
51

47
Backdating is based on the first common period link method, where the level of the new series and growth
rates of old series are reflected in the final time series. For Denmark and Norway, no backdates of household
credit are available.
48
Using bank credit may have limitations, as credit from capital markets is excluded.
49
See Abbas et al (2010), http://www.imf.org/external/pubs/cat/longres.cfm?sk=24332.0. In general, the levels
from the Flow of Funds total liabilities (which are the final series used) are higher than those from IMF public
debt data, as there are more items included in the former. The data gaps/missing values are normally one to
two years.
50
Our debt database reflects data availability up to early 2011.
51
http://terpconnect.umd.edu/~creinhar/Courses.html.
24/34

Appendix 2: Sectoral composition of non-financial debt
Table A2.1
Household debt
Levels Changes
2

1980 1990 2000 2010
1
198090 19902000 200010
United States 52 64 74 95 12 10 21
Japan 60 82 87 82 22 5 5
Germany 59 61 73 64 2 13 9
United Kingdom 37 73 75 106 36 2 31
France 27 46 47 69 18 2 22
Italy 6 21 30 53 15 9 23
Canada 56 63 67 94 7 4 27
Australia 42 46 74 113 5 27 39
Austria 41 41 47 57 0 6 10
Belgium 35 38 41 56 3 3 15
Denmark 95 152 57
Finland 29 48 35 67 19 14 33
Greece 8 9 20 65 1 11 45
Netherlands 43 49 87 130 6 38 43
Norway 64 94 31
Portugal 15 23 75 106 7 52 31
Spain 24 41 54 91 17 13 37
Sweden 53 61 51 87 8 10 36
Total of above
Median 39 47 65 94 8 8 31
Weighted average
3
46 60 69 90 14 9 18
Simple average 37 48 61 93 11 11 27
G7 43 59 65 87 16 6 16
Other advanced 32 39 58 97 7 14 34
1
2
3
Based on 2005 GDP and PPP exchange
rates.

25/34

Table A2.2
Non-financial corporate debt
Levels Changes
2

1980 1990 2000 2010
1
198090 19902000 200010
United States 53 65 66 76 12 1 9
Japan 176 215 178 161 39 37 17
Germany 46 35 91 100 11 56 9
United Kingdom 64 88 93 126 23 6 33
France 99 106 123 155 7 17 32
Italy 48 66 96 128 17 30 32
Canada 109 106 111 107 4 5 4
Australia 44 82 74 80 38 8 6
Austria 86 78 82 99 8 4 17
Belgium 73 86 136 185 12 51 49
Denmark 72 78 90 119 6 13 28
Finland 101 102 121 145 1 19 25
Greece 59 47 51 65 12 3 15
Netherlands 98 119 140 121 21 21 19
Norway 84 105 148 174 21 43 26
Portugal 93 50 114 153 42 63 39
Spain 120 97 133 193 23 36 60
Sweden 109 174 191 196 66 17 4
Total of above
Median 85 87 112 126 9 17 21
Weighted average
3
79 92 99 113 13 7 12
Simple average 85 94 113 128 9 19 19
G7 85 97 108 109 12 11 13
Other advanced 85 93 116 138 7 24 23

1
2
3
rates.

26/34

Table A2.3
Government debt
Levels Changes
2

1980 1990 2000 2010
1
198090 19902000 200010
United States 46 71 58 97 25 13 39
Japan 53 66 145 213 13 78 68
Germany 31 42 61 77 10 20 16
United Kingdom 58 42 54 89 16 12 35
France 34 46 73 97 12 27 24
Italy 54 93 126 129 39 33 4
Canada 71 109 115 113 39 6 3
Australia 43 46 37 41 3 8 4
Austria 36 59 76 82 23 17 6
Belgium 61 140 121 115 79 19 6
Denmark 36 77 73 65 41 5 8
Finland 16 23 67 57 7 44 9
Greece 26 83 124 132 57 42 7
Netherlands 65 97 67 76 33 30 9
Norway 43 38 44 65 6 6 21
Portugal 36 68 63 107 33 6 45
Spain 27 49 71 72 21 22 1
Sweden 58 54 77 58 4 24 20
Total of above
Median 43 63 72 97 22 15 7
Weighted average
3
46 66 78 104 20 12 31
Simple average 44 67 81 93 23 14 13
G7 50 67 90 107 17 23 26
Other advanced 41 67 75 85 26 8 5

1
2
3
rates.

27/34

Appendix 3: Detailed results for growth regressions
Table A3.1 Growth regressions with debt and crisis variables
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
VARIABLES growth05 growth05 growth05 growth05 growth05 growth05 growth05 growth05 growth05 growth05 growth05
National gross saving to GDP 0.0409 0.0550* 0.0430 0.0648** 0.0408 0.0689* 0.0635* 0.0437 0.0622* 0.0630* 0.0685*
(0.026) (0.030) (0.027) (0.030) (0.025) (0.035) (0.030) (0.027) (0.032) (0.031) (0.033)
Change in population 0.4482 0.2540 0.4183 0.3689 0.4531 0.3593 0.2684 0.3907 0.2883 0.2687 0.3658
(0.362) (0.349) (0.385) (0.342) (0.371) (0.344) (0.362) (0.401) (0.349) (0.362) (0.346)
Schooling 0.0051*** 0.0033 0.0054*** 0.0034 0.0051*** 0.0035* 0.0042* 0.0054*** 0.0042* 0.0042** 0.0035*
(0.001) (0.002) (0.001) (0.002) (0.001) (0.002) (0.002) (0.001) (0.002) (0.002) (0.002)
Log of real per capita GDP 0.1565*** 0.1635*** 0.1865*** 0.1568*** 0.1566*** 0.1591*** 0.1870*** 0.1872*** 0.1857*** 0.1868*** 0.1590***
(0.015) (0.018) (0.014) (0.015) (0.014) (0.017) (0.015) (0.015) (0.015) (0.016) (0.017)
Trade openness 0.0311** 0.0326** 0.0341** 0.0305** 0.0311** 0.0306** 0.0312** 0.0343** 0.0308** 0.0312** 0.0304**
(0.012) (0.012) (0.012) (0.013) (0.012) (0.013) (0.012) (0.012) (0.012) (0.012) (0.013)
Inflation rate 0.0049 0.0239 0.0176 0.0150 0.0049 0.0139 0.0268 0.0181 0.0267 0.0270 0.0139
(0.018) (0.021) (0.016) (0.017) (0.018) (0.018) (0.018) (0.017) (0.019) (0.019) (0.018)
Total dependency ratio 0.1955*** 0.1814*** 0.2195*** 0.1972*** 0.1971*** 0.1925*** 0.2142*** 0.2117*** 0.2204*** 0.2145*** 0.1950***
(0.044) (0.045) (0.041) (0.045) (0.045) (0.043) (0.041) (0.040) (0.042) (0.039) (0.043)
Liquid liabilities to GDP 0.9935 1.1207 1.2387*** 1.0038 0.9995 0.9530 1.2702*** 1.2166*** 1.2968*** 1.2753*** 0.9643
(0.693) (0.662) (0.387) (0.683) (0.647) (0.783) (0.368) (0.373) (0.445) (0.432) (0.724)
Banking crisis 0.0134*** 0.0116*** 0.0135*** 0.0132*** 0.0135*** 0.0129*** 0.0136*** 0.0130*** 0.0141*** 0.0136*** 0.0131***
(0.003) (0.004) (0.003) (0.004) (0.003) (0.003) (0.004) (0.003) (0.004) (0.004) (0.005)
Total non-financial debt 0.0078
(0.006)
Government debt 0.0167*** 0.0180** 0.0174*** 0.0175*** 0.0180**
(0.005) (0.006) (0.006) (0.006) (0.006)
Private debt 0.0016 0.0023
(0.007) (0.006)
Corporate debt 0.0006 0.0030 0.0023 0.0008
(0.007) (0.006) (0.007) (0.008)
Household debt 0.0050 0.0033 0.0027 0.0047
(0.014) (0.012) (0.012) (0.013)
Constant 1.6429*** 1.6854*** 1.9714*** 1.6102*** 1.6441*** 1.6717*** 1.9243*** 1.9781*** 1.9136*** 1.9225*** 1.6709***
(0.157) (0.195) (0.155) (0.153) (0.153) (0.179) (0.153) (0.161) (0.158) (0.169) (0.177)
Observations 383 354 383 354 383 354 354 383 354 354 354
R-squared 0.749 0.766 0.772 0.757 0.749 0.757 0.780 0.773 0.779 0.780 0.757
Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1.

28/34

Table A3.2 Growth regressions adding financial flow variables
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
National gross saving to GDP 0.0240 0.0038 0.0193 0.0269 0.0241 0.0301 0.0071 0.0186 0.0098 0.0083 0.0298
(0.020) (0.021) (0.016) (0.025) (0.020) (0.027) (0.018) (0.017) (0.021) (0.020) (0.028)
Change in population 0.8203* 0.5437 0.6365* 0.8679* 0.8600* 0.8272* 0.5065 0.5574 0.6071 0.5045 0.8615*
(0.444) (0.490) (0.363) (0.464) (0.456) (0.458) (0.406) (0.392) (0.370) (0.409) (0.470)
Schooling 0.0021 0.0021 0.0033 0.0022 0.0021 0.0023 0.0032 0.0035 0.0031 0.0032 0.0023
(0.002) (0.002) (0.003) (0.002) (0.002) (0.002) (0.003) (0.002) (0.003) (0.002) (0.002)
(0.015) (0.019) (0.014) (0.014) (0.014) (0.018) (0.015) (0.016) (0.015) (0.019) (0.017)
Trade openness 0.0184 0.0149 0.0118 0.0181 0.0185 0.0178 0.0103 0.0111 0.0112 0.0103 0.0179
(0.013) (0.012) (0.012) (0.013) (0.012) (0.014) (0.012) (0.012) (0.012) (0.012) (0.013)
Inflation rate 0.0035 0.0019 0.0009 0.0049 0.0032 0.0067 0.0016 0.0012 0.0010 0.0021 0.0060
(0.023) (0.025) (0.019) (0.023) (0.023) (0.024) (0.021) (0.020) (0.022) (0.023) (0.024)
Total dependency ratio 0.1888*** 0.1849*** 0.2320*** 0.1965*** 0.1958*** 0.1890*** 0.2270*** 0.2244*** 0.2350*** 0.2263*** 0.1948***
(0.047) (0.046) (0.042) (0.047) (0.046) (0.047) (0.042) (0.039) (0.041) (0.038) (0.045)
Liquid liabilities to GDP 1.1373* 1.2306** 1.3937*** 1.1389* 1.1720* 1.0664 1.4142*** 1.3595*** 1.4461*** 1.4002*** 1.1031
(0.626) (0.575) (0.375) (0.622) (0.594) (0.692) (0.354) (0.351) (0.442) (0.419) (0.670)
Banking crisis 0.0079*** 0.0059** 0.0089*** 0.0086*** 0.0085*** 0.0079** 0.0077*** 0.0081*** 0.0087*** 0.0076*** 0.0085**
(0.003) (0.003) (0.002) (0.003) (0.003) (0.003) (0.002) (0.002) (0.002) (0.002) (0.003)
Five-year forward average of
private borrowing to GDP 0.0437* 0.0259 0.0288* 0.0461* 0.0458* 0.0437* 0.0206 0.0242 0.0267 0.0204 0.0456*
(0.023) (0.027) (0.016) (0.025) (0.024) (0.024) (0.020) (0.019) (0.017) (0.021) (0.025)
Five-year forward average of
government borrowing to GDP 0.0567 0.0950*** 0.0939*** 0.0546 0.0537 0.0588 0.1138*** 0.1019*** 0.1018*** 0.1140*** 0.0555
(0.034) (0.031) (0.028) (0.040) (0.034) (0.040) (0.034) (0.031) (0.032) (0.035) (0.042)
Total non-financial debt 0.0103*
(0.005)
Government debt 0.0208*** 0.0240*** 0.0226*** 0.0218*** 0.0240***
(0.004) (0.006) (0.005) (0.004) (0.005)
Private debt 0.0030 0.0051
(0.006) (0.005)
Corporate debt 0.0027 0.0043 0.0054 0.0023
(0.007) (0.006) (0.006) (0.007)
Household debt 0.0065 0.0047 0.0041 0.0057
(0.011) (0.010) (0.010) (0.011)
Constant 1.7571*** 1.8342*** 2.1818*** 1.7318*** 1.7630*** 1.7976*** 2.1482*** 2.2080*** 2.1131*** 2.1538*** 1.7958***
(0.170) (0.208) (0.154) (0.154) (0.157) (0.198) (0.169) (0.183) (0.165) (0.203) (0.193)
Observations 362 354 362 354 362 354 354 362 354 354 354
R-squared 0.798 0.803 0.825 0.793 0.799 0.793 0.823 0.827 0.821 0.823 0.793

29/34

Table A3.3 Growth regressions without crisis variable
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
National gross saving to GDP 0.0395 0.0431 0.0414 0.0479 0.0414 0.0562 0.0462 0.0446 0.0502 0.0558 0.0611*
(0.026) (0.033) (0.029) (0.029) (0.026) (0.035) (0.033) (0.030) (0.035) (0.037) (0.035)
Change in population 0.2109 0.0966 0.1800 0.1896 0.1650 0.2296 0.0905 0.1094 0.1654 0.0897 0.1854
(0.306) (0.303) (0.281) (0.321) (0.296) (0.330) (0.298) (0.283) (0.306) (0.301) (0.324)
Schooling 0.0051** 0.0019 0.0054*** 0.0015 0.0051** 0.0017 0.0022 0.0054*** 0.0021 0.0025 0.0020
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
(0.017) (0.020) (0.020) (0.019) (0.019) (0.021) (0.021) (0.019) (0.022) (0.022) (0.022)
Trade openness 0.0407** 0.0452** 0.0437*** 0.0445** 0.0401** 0.0444** 0.0455** 0.0433** 0.0456** 0.0451** 0.0439**
(0.015) (0.016) (0.015) (0.017) (0.016) (0.017) (0.016) (0.015) (0.016) (0.017) (0.018)
Inflation rate 0.0157 0.0401 0.0282 0.0313 0.0146 0.0292 0.0429 0.0287 0.0402 0.0397 0.0272
(0.026) (0.026) (0.024) (0.026) (0.026) (0.024) (0.025) (0.024) (0.024) (0.024) (0.025)
Total dependency ratio 0.1637*** 0.1533*** 0.1870*** 0.1521*** 0.1431** 0.1658*** 0.1670*** 0.1614*** 0.1856*** 0.1624*** 0.1464***
(0.054) (0.051) (0.052) (0.049) (0.056) (0.049) (0.049) (0.052) (0.049) (0.048) (0.049)
Liquid liabilities to GDP 0.3023 0.7008 0.5382 0.5013 0.2907 0.4208 0.7387 0.5598 0.6512 0.6486 0.3842
(1.028) (0.880) (0.809) (1.050) (1.013) (1.095) (0.783) (0.711) (0.917) (0.828) (1.083)
Total non-financial debt 0.0116**
(0.005)
Government debt 0.0164** 0.0169** 0.0191*** 0.0136* 0.0165**
(0.007) (0.007) (0.006) (0.008) (0.007)
Private debt 0.0054 0.0093**
(0.005) (0.004)
Corporate debt 0.0082 0.0117** 0.0109* 0.0078
(0.006) (0.005) (0.005) (0.006)
Household debt 0.0023 0.0043 0.0013 0.0055
(0.014) (0.014) (0.014) (0.014)
Constant 1.6835*** 1.7789*** 2.0059*** 1.6251*** 1.6615*** 1.6877*** 1.9234*** 2.0268*** 1.9139*** 1.9616*** 1.6876***
(0.193) (0.212) (0.212) (0.201) (0.211) (0.208) (0.217) (0.211) (0.218) (0.232) (0.224)
Observations 383 354 383 354 383 354 354 383 354 354 354
R-squared 0.683 0.726 0.705 0.710 0.690 0.706 0.730 0.719 0.720 0.731 0.712

30/34
References
Abbas, A, N Belhocine, A ElGanainy and M Horton (2010): A historical public debt
database, IMF Working Papers, no. 10/245.
Aghion, P, D Hemous and E Kharroubi (2011): Cyclical fiscal policy, credit constraints, and
industry growth, BIS Working Papers, no 340, February.
Akerlof, G and R Shiller (2010): Animal spirits: how human psychology drives the economy,
and why it matters, Princeton University Press.
Anderson, T and C Hsiao (1981): Estimation of dynamic models with error components,
Journal of the American Statistical Association, no 76, pp 598606.
Arellano, M and S Bond (1991): Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations, Review of Economic Studies, no 58,
pp 27797.
Arellano, M and O Bover (1995): Another look at the instrumental variables estimation of
error component models, Journal of Econometrics, no 68, pp 2952.
Attanasio, O, S Kitao and G Violante (2007): Global demographic trends and social security
reform, Journal of Monetary Economics, no 54, pp 14498.
Auerbach, A (2011): Long-term fiscal sustainability in major economies, paper presented at
the 10th BIS Annual Conference, Lucerne, June.
Barro, R (1979): On the determination of the public debt, Journal of Political Economy,
no 87, pp 94071.
Barro, R and J Lee (2000): International data on educational attainment: updates and
implications, CID Working Papers, no 42, April.
Barro, R and X Sala-i-Martin (2004): Economic growth, second edition, MIT Press.
Beck, T and A Demirg-Kunt (2009): Financial institutions and markets across countries
and over time: data and analysis, World Bank Policy Research Working Papers, no 4943.
Beck, T, A Demirg-Kunt and R Levine (2000): A new database on financial development
and structure, World Bank Economic Review, no 14, pp 597605.
Bekaert, G, C Harvey and C Lundblad (2001): Emerging equity markets and economic
development, Journal of Development Economics, no 66, pp 465504.
(2005): Does financial liberalization spur growth?, Journal of Financial Economics,
no 77, pp 355.
Benito, A, M Waldron, G Young and F Zampolli (2007): The role of household debt and
balance sheets in the monetary transmission mechanism, Bank of England Quarterly
Bulletin, 47(1), pp 708.
Bernanke, B (2005): The global saving glut and the US current account deficit, speech at
the Sandridge Lecture, Richmond, 10 March.
Bernanke, B, C Bertaut, L DeMarco and S Kamin (2011), International capital flows and the
returns to safe assets in the United States, 20032007, International Finance Discussion
Papers, no 1014.
Bernanke, B and M Gertler (1990): Financial fragility and economic performance, Quarterly
Journal of Economics, no 105, pp 87114.
Bernanke, B, M Gertler and S Gilchrist (1999): The financial accelerator in a quantitative
business cycle framework, in J Taylor and M Woodford (eds), Handbook of
Macroeconomics, chapter 21, Elsevier, pp 134193.
Cecchetti, Mohanty and Zampolli
31/34

Borio, C and P Disyatat (2011): Global imbalances and the financial crisis: link or no link?,
BIS Working Papers, no 346, May.
Borio, C, C Furfine and P Lowe (2001): Procyclicality of the financial system and financial
stability: issues and policy options, in Marrying the macro and microprudential dimensions of
financial stability, BIS Papers, no 1, March.
Caballero, R (2010): The other imbalance and the financial crisis, NBER Working Papers,
no 15636.
Caballero, R, E Farhi and P Gourinchas (2008): An equilibrium model of global imbalances
and low interest rates, American Economic Review, vol 98, no 1, pp 35893.
Cecchetti, S, A Flores-Lagunes and S Krause (2005): Assessing the sources of changes in
the volatility of real growth, in C Kent and D Norman (eds), The changing nature of the
business cycle, proceedings of the Research Conference of the Reserve Bank of Australia,
November, pp 11538.
Cecchetti, S, M Mohanty and F Zampolli (2011): The future of public debt, in S Gokarn (ed),
Challenges to central banking in the context of the financial crisis, New Delhi: Academic
Foundation, pp 183217.
Cecchetti, S and K Schoenholtz (2011): Money, banking and financial markets, third edition,
New York, McGraw-Hill Irwin.
Checherita, C and P Rother (2010): The impact of high and growing government debt on
economic growth. An empirical investigation for the euro area, European Central Bank
Working Papers, no 1237, August.
Cukierman, A and A Meltzer (1989): A political theory of government debt and deficits in a
neo-Ricardian framework, American Economic Review, vol 79, no 4, pp 71332.
Dynan, K, D Elmendorf and D Sichel (2006): Can financial innovation help to explain the
reduced volatility of economic activity?, Journal of Monetary Economics, no 53, pp 12350.
Dynan, K and D Kohn (2007): The role of US household indebtedness: causes and
consequences, Federal Reserve Board Finance and Economics Discussion Series,
2007-37, August.
Eggertson, G and P Krugman (2011): Debt, deleveraging, and the liquidity trap: a Fisher-
Minsky-Koo approach, Federal Reserve Bank of New York, unpublished manuscript,
February.
Eichengreen, B, R Feldman, J Liebman, J von Hagen and C Wyplosz (2011): Public debts:
nuts, bolts and worries, Geneva Report on the World Economy, no 13, CEPR, August.
Fehr, H, S Jokisch and L Kotlikoff (2005): The developed worlds demographic transition
the roles of capital flows, immigration, and policy, in R Brooks and A Razin (eds), Social
security reform financial and political issues in international perspective, Cambridge
University Press.
Friedman, M (1981): Debt and economic activity in the United States, in M Friedman (ed),
The changing roles of debt and equity in financing US capital formation, University of
Chicago Press.
(1986): Increasing indebtedness and financial stability in the United States, in Debt,
Financial Stability and Public Policy, a symposium sponsored by the Federal Reserve Bank
of Kansas City, pp 2752.
(1987): New directions in the relationship between public and private debt, NBER
Working Papers, no 2186, March.
32/34

Gagnon, J (2011): The global outlook for government debt over the next 25 years:
implications for the economy and public policy, Peterson Institute for International
Economics, Washington DC, June.
Hansen, B (1999): Threshold effects in non-dynamic panels: estimation, testing, and
inference, Journal of Econometrics, no 93, pp 34568.
Huber, P (1967): The behavior of maximum likelihood estimates under non-standard
conditions, in Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and
Probability, Berkeley, University of California Press, vol 1, pp 22133.
International Monetary Fund (2011): Shifting gears: tackling challenges on the road to fiscal
adjustment, Fiscal Monitor, April.
Islam, N (1995): Growth empirics: a panel data approach, Quarterly Journal of Economics,
no 110, pp 112770.
Judson, R and A Owen (1999): Estimating dynamic panel data models: a guide for
macroeconomists, Economic Letters, no 65, pp 915.
Kaufmann, D (2010): Can corruption adversely affect public finances in industrialized
countries?, Brookings Institution, April.
Keen, M, A Klemm and V Perry (2010): Tax and the crisis, Fiscal Studies, no 31, pp 4379.
Kiyotaki, N and J Moore (1997): Credit cycles, Journal of Political Economy, no 105,
pp 21148.
Krueger, D and A Ludwig (2007): On the consequences of demographic change for rates of
returns to capital, and the distribution of wealth and welfare, Journal of Monetary
Economics, no 54, pp 4987.
Kumar, M and J Woo (2010): Public debt and growth, IMF Working Papers, no 10/174,
July.
Levine, R (2005): Finance and growth: theory, evidence and mechanisms, in P Aghion and
S Durlauf (eds), Handbook of economic growth, vol 1, part 1, Amsterdam: North Holland,
pp 865934.
Modigliani, F and S Cao (2004): The Chinese saving puzzle and the life-cycle hypothesis,
Journal of Economic Literature, no 42, pp 14570.
Myers, S (2010): Capital structure, Journal of Economic Perspectives, no 15, pp 81102.
Nickell, S (1981): Biases in dynamic models with fixed effects, Econometrica, no 49,
pp 141726.
Obstfeld, M and K Rogoff (2009): Global imbalance and the financial crisis: product of
common causes, paper prepared for the Federal Reserve Bank of San Franciscos
conference on Asia and the Global Financial Crisis.
Reinhart, C and K Rogoff (2008): Banking crises: an equal opportunity menace, NBER
Working Papers, no 14587, December.
(2009): This time is different. Eight centuries of financial folly, Princeton University
Press.
(2010): Growth in time of debt, American Economic Review Papers & Proceedings,
no 100, pp 5738.
Rogers, W (1993): Regression standard errors in clustered samples, Stata Technical
Bulletin, no 13, pp 1923; reprinted in Stata Technical Bulletin Reprints, no 3, pp 8894.
33/34

Shin, H (2009): Global imbalances, twin crises and the financial stability role of monetary
policy, paper prepared for the KIEP/CEPR conference on The World Economy with the G20,
Seoul, 20 November.
Stock, J, J Wright and M Yogo (2002): A survey of weak instruments and weak identification
in generalized method of moments, Journal of Business and Economic Statistics, no 20,
pp 51829.
Takts, E (2010): Ageing and asset prices, BIS Working Papers, no 318, August.
Tang, G and C Upper (2010): Debt reduction after crises, BIS Quarterly Review,
September.
Waldron, M and F Zampolli (2010a): Household debt, house prices and consumption in the
United Kingdom: a quantitative theoretical analysis, Bank of England Working Papers,
no 379.
(2010b): The rise in home prices and household debt in the UK: potential causes and
implications, chapter 5 in S Smith and B Searle (eds), The Blackwell companion to the
economics of housing. The housing wealth of nations, Wiley-Blackwell.
White, H (1980): A heteroscedasticity-consistent covariance matrix estimator and a direct
test for heteroscedasticity, Econometrica, no 48, pp 81730.
Woodford, M (1990): Public debt as private liquidity, American Economic Review, vol 80,
no 2, pp 38288.
(2003): Interest and prices: foundations of a theory of monetary policy, Princeton
University Press.
(2011): Financial stability and monetary policy, presentation at the NBER Summer
Institute, July.

BIS - The Real Effects of Debt

Uploaded by

Copyright:

Available Formats

BIS - The Real Effects of Debt

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

BIS - The Real Effects of Debt

Uploaded by

Copyright:

Available Formats

Cecchetti, Mohanty and Zampolli The real effects of debt

The real effects of debt

You might also like