Work 677
Work 677
No 677
Macroeconomic
implications of financial
imperfections: a survey
by Stijn Claessens and M Ayhan Kose
November 2017
JEL classification: D53, E21, E32, E44, E51, F36, F44, F65,
G01, G10, G12, G14, G15, G21
Bank for International Settlements 2017. All rights reserved. Brief excerpts may be
reproduced or translated provided the source is stated.
Abstract
This paper surveys the theoretical and empirical literature on the macroeconomic
implications of financial imperfections. It focuses on two major channels through
which financial imperfections can affect macroeconomic outcomes. The first channel,
which operates through the demand side of finance and is captured by financial
accelerator-type mechanisms, describes how changes in borrowers balance sheets
can affect their access to finance and thereby amplify and propagate economic and
financial shocks. The second channel, which is associated with the supply side of
finance, emphasises the implications of changes in financial intermediaries balance
sheets for the supply of credit, liquidity and asset prices, and, consequently, for
macroeconomic outcomes. These channels have been shown to be important in
explaining the linkages between the real economy and the financial sector. That said,
many questions remain.
JEL classifications: D53, E21, E32, E44, E51, F36, F44, F65, G01, G10, G12, G14, G15,
G21
Keywords: asset prices, balance sheets, credit, financial accelerator, financial
intermediation, financial linkages, international linkages, leverage, liquidity,
macrofinancial linkages, output, real-financial linkages
Claessens (Monetary and Economic Department, Bank for International Settlements; CEPR;
Stijn.Claessens@bis.org); Kose (Development Prospects Group, World Bank; Brookings Institution;
CEPR; CAMA; akose@worldbank.org). We are grateful to Olivier Blanchard for his helpful comments
at the early stages of this project of surveying the literature on macrofinancial linkages when we were
with the Research Department of the International Monetary Fund. This paper has a companion,
Asset prices and macroeconomic outcomes: a survey, BIS Working Paper no. 676. We would like to
thank Boragan Aruoba, Claudio Borio, Menzie Chinn, Ergys Islamaj, Jaime de Jesus Filho, Raju
Huidrom, Atsushi Kawamoto, Seong Tae Kim, Grace Li, Raoul Minetti, Ezgi Ozturk, Eswar Prasad, Lucio
Sarno, Hyun Song Shin, Kenneth Singleton, Naotaka Sugawara, Hui Tong, Kamil Yilmaz, Kei-Mu Yi,
Boyang Zhang, Tianli Zhao, and many colleagues and participants at seminars and conferences for
useful comments and inputs. Miyoko Asai, Ishita Dugar and Xinghao Gong provided excellent
research assistance. Mark Felsenthal, Sonja Fritz, Krista Hughes, Serge Jeanneau, Graeme Littler,
Tracey Lookadoo, and Rosalie Singson provided outstanding editorial help. The views expressed in
this paper are those of the authors and do not necessarily represent those of the institutions they are
affiliated with or have been affiliated with.
Hello, Paul, where have you been for the last 30 years? Pretty much all we have
been doing for 30 years is introducing flaws, frictions and new behaviors... The long
literature on financial crises and banking has also been doing exactly the same.
John H. Cochrane (2011)
I believe that during the last financial crisis, macroeconomists (and I include myself
among them) failed the country, and indeed the world. In September 2008, central
bankers were in desperate need of a playbook that offered a systematic plan of attack
to deal with fast evolving circumstances. Macroeconomics should have been able to
provide that playbook. It could not
Narayana Kocherlakota (2010)
What does concern me of my discipline, however, is that its current core by which
I mainly mean the so-called dynamic stochastic general equilibrium (DSGE) approach
has become so mesmerized with its own internal logic This is dangerous for both
methodological and policy reasons To be fair to our field, an enormous amount of
work at the intersection of macroeconomics and corporate finance has been chasing
many of the issues that played a central role during the current crisis However, much
of this literature belongs to the periphery of macroeconomics rather than to its core
Ricardo Caballero (2010)
One can safely argue that there is a hole in our knowledge of macro financial
interactions; one might also argue more controversially that economists have filled this
hole with rocks as opposed to diamonds; but it is harder to argue that the hole is empty.
Ricardo Reis (2017)
The financial crisis made it clear that the basic model, and even its DSGE cousins,
had other serious problems, that the financial sector was much more central to
macroeconomics than had been assumed...
Olivier Blanchard (2017a)
1. Introduction ....................................................................................................................................... 1
8. Conclusions ...................................................................................................................................... 38
A summary ...................................................................................................................................... 38
What next? ...................................................................................................................................... 39
References ................................................................................................................................................ 79
The Great Financial Crisis (GFC) of 200709 confirmed the vital importance of
advancing our understanding of macrofinancial linkages. The GFC was a bitter
reminder of how sharp fluctuations in asset prices, credit and capital flows can have
a dramatic impact on the financial position of households, corporations and
sovereign nations.1 These fluctuations were amplified by macrofinancial linkages,
bringing the global financial system to the brink of collapse and leading to the
deepest contraction in world output in more than half a century. Moreover, these
linkages have resulted in unprecedented challenges for fiscal, monetary and financial
sector policies.
Macrofinancial linkages centre on the two-way interactions between the real
economy and the financial sector. Shocks arising in the real economy can be
propagated through financial markets, thereby amplifying business cycles.
Conversely, financial markets can be the source of shocks, which, in turn, can lead to
more pronounced macroeconomic fluctuations. The global dimensions of these
linkages can result in cross-border spillovers through both real and financial channels.
The crisis has led to a lively debate over the state of research on the role of
financial market imperfections in explaining macroeconomic fluctuations. Some
argue that the crisis showed that the profession did not pay sufficient attention to
these linkages. Others, by contrast, claim that they have been recognised for a long
time and that substantial progress has been made in understanding them. But most
acknowledge that financial market imperfections can often intensify fluctuations in
the financial and real sectors. Yet, the absence of a unifying framework to study the
two-way interactions between the financial sector and the real economy has limited
the practical applications of existing knowledge and impeded the formulation of
policies.2
This debate can be seen as a natural extension of the long-standing discussion
about the importance of financial market developments for the real economy (as
described in detail in Appendix I).3 The diverging paths followed by the fields of
1
A large literature documents the various macrofinancial linkages that have contributed to the
devastating impact of the GFC. Some of the important books on the topic include Krugman (2009b),
Sorkin (2009), Wessel (2009), Lewis (2010), Kose and Prasad (2010), Paulson (2010), Gorton (2012),
Turner (2012), Bernanke (2013), Blinder (2013), Claessens et al (2014), Geithner (2014), Mian and Sufi
(2014a), Wolf (2014), Farmer (2016), King (2016) and Taylor (2016). Lo (2012) reviews a set of 21 books
on the GFC.
2
We presented some quotes reflecting the flavour of this debate at the beginning of the survey.
Krugman (2009a) criticises the macroeconomics literature for its failure to recognise the strong
relationship between the financial sector and the real economy, while Cochrane (2011, 2017)
responds critically to Krugmans views. Caballero (2010), Kocherlakota (2010), Taylor (2011), Romer
(2016) and Reis (2017) provide varying assessments of research on macroeconomics. Blanchard
(2017a) stresses the need for a broader class of macroeconomic models.
3
Early surveys of the literature on macrofinancial linkages include Gertler (1988), Bernanke (1993),
Lowe and Rohling (1993) and Bernanke et al (1996). Mankiw (2006) and Blanchard (2000, 2009) offer
more general reviews of the state of macroeconomics before the crisis. Recent (but more selective)
updates on macrofinancial linkages include Gilchrist and Zakrajek (2008), Matsuyama (2008),
Solimano (2010), BCBS (2011, 2012), Caprio (2011), Gertler and Kiyotaki (2011), Quadrini (2011), Borio
(2014) and Morley (2016), as well as papers in Friedman and Woodford (2011). Work related to
macrofinancial linkages includes: Cochrane (2006) on financial markets and the real economy;
Brunnermeier (2001) and Cochrane (2005) on asset pricing; and Tirole (2006) on the role and impact
of financial imperfections on corporate finance and other economic variables. Crowe et al (2010) and
Nowotny et al (2014) present collections of papers on macrofinancial linkages. Brunnermeier et al
(2013) provide an analytical review of the literature on macro models with financial frictions. On the
supply side, Adrian and Shin (2010b) survey the literature on the changing role of financial institutions
and the growing importance of the shadow banking system. Gorton and Metrick (2013) and Pozsar
et al (2013) review the role of securitisation and shadow banking; Brunnermeier and Oehmke (2013)
and Scherbina and Schlusche (2014) review the literature on asset price bubbles; and Forbes (2012)
reviews the literature on asset price contagion. Blanchard (2017a) looks at the state of
macroeconomics, focusing on the need to include distortions other than nominal price rigidities,
including financial frictions. Kocherlakota (2016) shows how the predictions of real business cycle
models significantly change in the presence of small nominal rigidities and argues that these types
of models are not useful tools for analysis of business cycles. For a recent discussion of the need to
incorporate financial frictions, labour market frictions and household heterogeneity in benchmark
macroeconomic models, see Ghironi (2017). Obstfeld and Taylor (2017) consider the importance of
finance in the context of the international monetary system. The literature on law and finance also
relates to the broad theme of macrofinancial linkages (see La Porta et al (2013) for a recent review)
but more from a longer-run developmental perspective.
4
Claessens and Kose (2017) review research on the interactions between asset prices and
macroeconomic outcomes in models without financial market imperfections. In these studies,
changes in financial variables, such as asset prices, are associated with individual consumption and
investment decisions but there are no aggregate feedback mechanisms from financial to real
variables and little scope for macrofinancial linkages.
5
Some of the pioneering papers on financial frictions include Akerlof (1970), Jensen and Meckling
(1976), Townsend (1979), Stiglitz and Weiss (1981), Bulow and Rogoff (1989) and Hart and Moore
(1994).
6
Throughout this paper, the terms financial market imperfections and financial frictions are used
interchangeably. Financial frictions in conjunction with a countrys legal, regulatory and tax system
influence the design and evolution of financial contracts, markets and intermediaries.
7
Important contributions on the supply side include Diamond and Dybvig (1983), Gale and Hellwig
(1985), Calomiris and Kahn (1991), Holmstrm and Tirole (1997), Allen and Gale (1998) and Diamond
and Rajan (2001). More recent work includes Adrian and Shin (2008), Geanakoplos (2008) and
Brunnermeier and Pedersen (2009), Danielsson et al (2011), Gertler and Karadi (2011), Gertler and
Kiyotaki (2011), Goodhart et al (2012), He and Krishnamurthy (2012), Brunnermeier and Sannikov
(2014, 2016), Begenau (2016), Gertler et al (2016), Begenau and Landvoigt (2017) and Piazzesi and
Schneider (2017). Many recent models combine demand and supply side considerations (including
banks, shadow banks and payment and collateral services).
This section reviews the basic mechanisms through which financial market
imperfections on the demand side can lead to macrofinancial linkages. It starts with
a conceptual discussion of how the state of borrowers balance sheets affects their
access to external financing at the microeconomic level, ie at the levels of households
and corporations. This is followed by a summary of microeconomic and sectoral
evidence supporting the importance of imperfections on the demand side. Although
most demand side studies support the role of imperfections, their quantitative
importance is still under debate.
A. Basic mechanisms
8
As Quadrini (2011) notes in his review, the presence of financial frictions implies the absence of
complete trade in certain risks, that is, models with financial frictions feature missing markets, thus
limiting a full sharing of risk. Models (implicitly or explicitly) also assume heterogeneous agents since
there would otherwise be no reason to trade claims inter-temporally or intra-temporally. The fact
that markets are missing may be exogenously assumed or can arise from financial frictions, that is,
forms of market incompleteness due to information asymmetries and limited enforcement. There has
been an extensive theoretical literature on how such imperfections can lead to missing markets or
incomplete contracts, including in financial markets. Seminal corporate finance papers include
Akerlof (1970), Jaffee and Russell (1976), Jensen and Meckling (1976), Rothschild and Stiglitz (1976),
Stiglitz and Weiss (1981) and Myers and Majluf (1984). Tirole (2006) provides a textbook treatment
of the role of imperfections in corporate finance and discusses some of their macroeconomic
implications. Meyers (2015) reviews capital structure theories and how they have been tested.
Samphantharak and Townsend (2009) discuss how finance affects household behaviour. Dewatripont
and Tirole (1994) and Freixas and Rochet (1997) present early analytical overviews of imperfections
as they affect financial institutions (see also Greenbaum et al (2016)).
9
These imperfections also affect the formation of firms since a firm (as an organiser of activities) can
be a mechanism to internalise some of those constraints (Coase (1937)). Aghion and Holden (2011)
present a review of research on incomplete contracts and the theory of the firm.
B. Empirical evidence
10
Other studies include Hubbard (1998), Hoshi et al (1991), Whited (1992), Calomiris et al (1995), Gross
(1995) and Hubbard et al (1995). Stein (2003) presents a survey of theoretical and empirical work
regarding the determinants of firm-level investment dynamics.
11
For example, Kaplan and Zingales (1997) criticise the approach taken by Fazzari et al (1988) because
they do not control for the endogeneity of financing constraints. See further Kaplan and Zingales
(2000) on why investment-cash flow sensitivities are not good indicators of financing constraints.
12
See for additional work, Cantor (1990), Blinder and Maccini (1991), Oliner and Rudebusch (1993),
Carpenter et al (1994) and Guiso et al (2013).
This section describes how financial imperfections can lead to more pronounced
macroeconomic fluctuations. It first introduces the financial accelerator mechanism
through which financial imperfections amplify and propagate aggregate cyclical
fluctuations. In general, equilibrium models featuring this mechanism display that
real, monetary and financial shocks can have a magnified effect on the real economy
because borrowers adjust their investment in response to changes in external
financing. The section then presents a summary discussion of the empirical evidence
How does the financial accelerator work? The relationships described above, between
the amount of external financing and its cost (premium), on one hand, and the
strength of borrowers balance sheets and cash flow positions, on the other, lead to
an amplification mechanism where small shocks can result in large economy-wide
adjustments to investment and consumption. Since wealth is a state (or given)
variable and economic agents cannot quickly (or optimally) adjust their
investment/saving plans (as they face costs of doing so), this mechanism persists over
time, causing short-lived shocks to real or financial variables to have longer-lasting
effects on the real economy. This propagation mechanism can also have general
equilibrium effects as individual agents actions affect other agents behaviour in a
mutually reinforcing fashion (see Figures 4A and 4B).13
13
For surveys describing variants of the financial accelerator mechanism, see Antony and Broer (2010),
BCBS (2011), Coric (2011) and Quadrini (2011).
14
Earlier general equilibrium models often featured incomplete financial markets (rather than market
imperfections). Farmer (1984), for example, presents a setting where a complete set of futures
markets does not exist because traders have finite lives.
15
Each model has its specific advantages and limitations. Some models assume that agents are short-
lived, as in the overlapping generations model of Bernanke and Gertler (1990) and Suarez and
Sussman (1997). The model by Kiyotaki and Moore (1997) is dynamic with long-lived agents but, as
Suarez and Sussman (1997) highlight, it rules out price indexation as a way of insuring against
unanticipated shocks.
16
See also Bernanke and Gertler (2000) and Cecchetti et al (2000) for a discussion of different
mechanisms, including those working through asset prices.
17
Other papers employing general equilibrium models (with the financial accelerator) discuss how
financial institutions fit into broader real activity, including Christiano et al (2003), Christensen and
Dib (2008) and De Graeve (2008).
18
Guerrieri and Iacoviello (2013) present a model with collateral constraints that displays asymmetric
responses to house price changes. In their model, collateral constraints become muted when housing
wealth is high (shocks to house prices lead to small and positive changes in consumption and hours
worked). However, collateral constraints become tight when housing wealth is low (shocks to house
prices translate into negative and large changes in consumption and hours worked). Kannan et al
(2012) consider the importance of credit constraints in driving the linkages between house prices and
macroeconomic fluctuations. Iacoviello and Pavan (2013) and Iacoviello (2004, 2015) also consider
the role of housing markets in explaining business cycles. Wachter et al (2014) present a collection
of papers on the role of housing markets during the GFC.
19
For a discussion of these other channels, see Claessens and Kose (2017). For a review of the relevance
of the corporate finance literature for monetary policy, see Trichet (2006).
20
See, for example, Faia and Monacelli (2007), Iacoviello and Minetti (2008), Christiano et al (2008),
Carlstrom et al (2009), De Fiore and Tristani (2011), Eggertsson and Krugman (2012) and Crdia and
Woodford (2016). Woodford (2011) reviews this abundant literature. Considering balance sheet
effects, Taylor (2008) proposes a modified Taylor rule that adjusts the short-term interest rate to
observed increases in credit spreads. Kannan et al (2012) provide evidence of how the inclusion of
house price movements in the conduct of monetary policy can help stabilise the economy in the face
of pressures in the housing market.
21
Christiano and Davis (2006) claim that the result by Chari et al (2007) is not warranted if spillovers
across wedges are taken into consideration (see also Justiniano et al (2010, 2011)). Ajello (2016) finds
evidence that a significant fraction of US output and investment volatility is driven by shocks to
financial intermediation spreads.
22
Using a standard New Keynesian model, Eggertsson (2012) argues that the New Deal policies of the
Great Depression were helpful in promoting the recovery. These policies were expansionary because
they changed expectations (from deflationary to inflationary), thus eliminating the deflationary spiral
of 192933. This made lending cheaper and stimulated demand.
23
Since some countries respond to these risks by building up foreign exchange reserves, such
precautionary holdings of foreign, liquid assets could turn sudden stops into low-probability events
nested within normal cycles, as observed in the data (Mendoza et al (2009), Borio and Disyatat (2011)).
Bianchi (2011) studies the implications of credit constraints for overborrowing in a small open
economy DSGE model and concludes that raising the cost of borrowing during tranquil times restores
constrained efficiency and significantly reduces the incidence and severity of financial crises.
Brunnermeier and Sannikov (2015) study a model in which short-term capital flows could be excessive
and be a source of financial stress. Kalantzis (2015) study a two-sector model in which large capital
inflows lead to financial crises.
24
Devereux and Yetman (2010) and Dedola and Lombardo (2012) also examine how credit market
shocks in DSGE models with financial market imperfections can generate international business cycle
spillovers. Kollmann et al (2011) introduce a banking sector in an international business cycle model
and study how shocks to this sector can generate global spillovers. Bacchetta and van Wincoop (2016)
show that national business cycles can become highly synchronised when the world economy is hit
by a global panic shock. Rose and Spiegel (2010, 2011) and Kamin and DeMarco (2012) examine this
issue using empirical approaches.
25
Some recent studies emphasise the importance of various imperfections associated with financial
shocks, trade credit, and working capital in explaining the sharp decline of trade relative to output
during the GFC (Amiti and Weinstein (2011), Chor and Manova (2012) and Bems et al (2013)).
The bank lending channel, also referred to as the narrow credit channel, arises from
the special role played by banks in credit extension. As explained in the previous
section, certain asymmetric information problems are more likely to be prevalent
among households and small firms. This can limit their access to financial services,
even when households have adequate income or when firms have projects with
reasonably high risk-adjusted returns. Banks invest in information acquisition and
monitoring and can thereby (partially) overcome the problems arising from
information asymmetry (and other contracting problems).26 However, during this
process, some households and smaller firms may become bank-dependent in that
they are unable to substitute with ease other forms of finance for bank loans (or
whatever financial services they obtain from a bank). Larger firms, by contrast, may
26
Many studies examine the special roles of banks (Freixas and Rochet (2008) and BCBS (2016) review
the literature). For example, banks can screen potential borrowers, acquire information on firms
collateral (Rajan and Winton (1995) and Diamond and Rajan (2001)) or directly monitor borrowers
actions in order to prevent problems associated with moral hazard (Repullo and Suarez (2000) and
Holmstrm and Tirole (1997)). Earlier contributions include Brunner and Meltzer (1963) and Bernanke
(1983), and Rajan (1998) provides a comprehensive review of the functions of banks.
27
Bernanke and Blinder (1988) provide a stylised discussion of the lending channel using an IS/LM type
framework. Stein (1998) provides a micro-founded adverse selection model of bank asset and
liability management that generates a lending channel. For an early overview of the theory and
empirical evidence relating to the bank lending channel, see Kashyap and Stein (1994).
28
Holmstrm and Tirole (1998) represents the pioneering study of the special role of banks in a general
equilibrium environment. For models that endogenise the superior skills of banks in collecting
entrepreneurs collateral, see Habib and Johnsen (1999) and Araujo and Minetti (2007).
29
Gorton and Huang (2004) show how, in an environment in which private investors make inefficient
project choices (eg as they cannot accumulate the liquidity needed to buy the assets of distressed
financial institutions), the government can provide liquidity and help mitigate such inefficient choices.
Lorenzoni (2007) shows that competitive financial contracts can result in excessive borrowing ex ante
and excessive volatility ex post in an economy with financial frictions and hit by aggregate shocks.
30
Early surveys of the literature on the bank lending channel include Bernanke (1993), Bernanke and
Gertler (1995), Cecchetti (1995), Hubbard (1995) and Peek and Rosengren (1995a). As discussed later,
there are also studies highlighting the importance of risk-taking by banks in their lending decisions
(eg Disyatat (2011) and Borio and Zhu (2012)).
31
A number of studies (Calomiris and Kahn (1991), Kashyap et al (2002), Diamond and Rajan (2001) and
Huberman and Repullo (2014)) explain why banks fund themselves with short maturities given those
risks. These models generally rely on the disciplining features of short-term debt and the beneficial
tension between making illiquid loans to borrowers and providing liquidity on demand to depositors.
While other intermediaries, such as money market funds, are also vulnerable to runs, as seen during
the GFC (Schmidt et al (2016) and Covitz et al (2013)), they generally do not lend and take short-term
on-demand deposits at the same time. They are also thought to be less special, in that their
intermediation functions can be more easily replaced, although the GFC raised questions about such
an assumption.
32
Some of these issues are discussed further in Claessens and Kose (2014). Acharya et al (2011a, 2011b)
model freezes in the market for bank assets. In their models, depending on the information
environment and the nature of liquidation costs, small shocks can lead to sudden interruptions in
financial institutions ability to roll over their liabilities. Bank liquidity may also be countercyclical, that
is, inefficiently high during booms but excessively low during crises, making interventions to resolve
banking crises desirable ex post but not ex ante (Farhi and Tirole (2012)). See further Tirole (2011) for
a review of various aspects of illiquidity and Holmstrm and Tirole (2011) for an extensive analysis of
(private and public) forms of liquidity.
The health of a financial intermediarys balance sheet can influence its lending and
other intermediation activities. Balance sheet positions, especially net worth, matter
for financial intermediaries just as is the case for non-financial corporations. Net
worth has an impact on financial intermediaries access to funds and their liquidity
positions and thereby affects their lending activities. Banks also need to satisfy capital
adequacy requirements (whether market- or regulation-driven). Given the costs
associated with raising new capital quickly on the open market, a banks net worth
depends over the short run on changes in the quality of its loan portfolio and the
value of its other assets, including securities.
Consequently, changes in the value of a banks assets will affect its access to and
cost of funding and its ability to make new loans. A decline in loan quality, for
example, or a fall in the value of tradable assets, can lead to a drop in a banks capital.
This can make its funding more costly or make its capital adequacy requirements
binding, forcing the bank to shrink its loan book. When these shocks take place
simultaneously at many banks, they can lead to systemic consequences, especially
when alternative sources of external financing are limited.
These effects can be a source of aggregate cyclical fluctuations through what has
been called the bank capital channel (Greenwald and Stiglitz (1993); see Borio and
Zhu (2012) for further references). When many banks are affected by the same capital
shock, aggregate effects can occur. For example, during a recession, the quality of
bank loan portfolios tends to weaken, adversely impacting banks balance sheets. In
order to shore up their relative capital positions (as desired by the market or to satisfy
regulatory requirements) but unlikely to be able to raise capital quickly banks may
need to tighten their lending standards and reduce the volume of risky credit they
provide.33 Since borrowers who rely on banks for their external funding needs have a
limited set of alternatives, this can lead to a slowdown in economic activity, or even a
recession, with a higher proportion of non-performing loans and deteriorating bank
balance sheets. The decline in bank lending induced by such a capital crunch can
affect (and interact with) economic activity through various channels (see Bernanke
and Lown (1991), Holmstrm and Tirole (1997), Repullo and Suarez (2000) and Van
den Heuvel (2006, 2008)). With this mechanism, a strong link can arise between bank
capital, the supply of bank financing and macroeconomic outcomes.
The interaction between bank capital and firm liquidity matters in various ways,
especially when firms are locked into credit relationships with banks. Den Haan et al
(2003) and Minetti (2007) show that a capital crunch can induce firms to abandon
high quality projects or break up credit relationships. Thus, a capital crunch depresses
not only the volume of investment but also its average productivity. Chen (2001)
shows that a capital crunch can cause a drop in asset prices (eg real estate), which
can, in turn, have feedback effects, including a contraction in bank capital. Minetti and
Peng (2013) investigate the mechanics of the bank capital channel in an open
economy model (calibrated for Argentina) and show that real interest rate shocks
generate large fluctuations in output and real estate prices.
Some recent studies employing DSGE models help gauge the quantitative
importance of the bank capital channel. Gertler and Kiyotaki (2011, 2015) and Gertler
33
Repullo and Suarez (2013) provide a model in which banks are subject to regulatory capital
requirements and have limited access to equity markets. Gorton and Winton (2017) present a general
equilibrium model to study the private and social costs of bank capital.
Leverage, defined as the ratio of total assets to shareholder equity, has received much
attention recently because of its role in the GFC. Fluctuations in the leverage of
financial institutions (and other agents) relate to changes in asset prices through both
simple accounting and the behaviour of agents. The basic accounting relationship
between movements in asset values and changes in leverage is negative, ie rising
asset prices boosts net worth and thereby makes measured leverage drop, ie leverage
is countercyclical. Similar to the financial accelerator mechanism, this means that
financial institutions and other agents can fund themselves easier in times of rising
asset prices, and consequently lend more to others, even without raising their
leverage.
In practice, as Adrian and Shin (2008) show, leverage is not countercyclical (or
even acyclical), but procyclical, at least for broker-dealers; it increases when asset
prices rise and falls when asset prices decline. And, in financial markets, as
Geanakoplos (2010) shows, margins (or haircuts), which dictate the share of financing
available for a unit of collateral, are procyclical too, ie lower during booms and higher
in busts. While the exact mechanisms remain unclear, many attribute this
procyclicality to perverse incentive structures, incomplete corporate governance
arrangements, herding behaviour and other agency issues (see Borio et al (2001);
Claessens and Kodres (2017)).
As leverage fluctuates, it affects the supply of financing. When leverage is high,
supply can be expected to be more ample since it means that intermediaries face
34
There is a large literature on the effectiveness of UMPs that uses various approaches, see Eggertsson
and Woodford (2003), Krishnamurthy and Vissing-Jorgensen (2011), Farmer (2012), Woodford (2012),
Bauer and Rudebusch (2014), Baumeister and Benati (2013), Swanson and Williams (2014), Arteta et
al (2015, 2016), Farmer and Zabczyk (2016), and Borio and Zabai (2016).
35
The leverage cycle can relate to the presence of asset price bubbles, which can be rational or
irrational. Either type can be welfare enhancing or reducing (see further Claessens and Kose (2017)
on asset bubbles). Nuo and Thomas (2017) document that leverage has contributed more than
equity to fluctuations in total assets and that it is positively correlated with assets and GDP but
negatively correlated with equity (see also Halling et al (2016)). He and Krishnamurthy (2013) develop
a model of financial intermediaries to study the linkage between risk premia and leverage.
36
See Adrian and Ashcraft (2016) for an overview of the shadow banking system, its growth and
functioning; Claessens et al (2012b) for a review of the functions performed by shadow banking
systems; Gennaioli et al (2013), Gertler et al (2016) and Begenau and Landvoigt (2017) for analytical
models of shadow banking; and Gorton and Ordoez (2013) on how an economy can become fragile
if it relies extensively on privately-produced safe assets, such as those generated by shadow banking.
See further Gorton (2016) and Golec and Perotti (2017) for reviews of the literature on safe assets
with a domestic focus and Gourinchas and Rey (2016) for an analysis of the role and effects of safe
assets globally.
The bank lending channel has been extensively studied empirically, at least as regards
the effect of changes in bank liquidity conditions. There is intense debate about
whether this channel can be identified with macroeconomic data given the difficulty
of separating factors driving demand from those driving supply. Some studies look
at credit market indicators, such as the ratio between bank lending and commercial
paper, showing that tighter monetary policy leads to a decline of the ratio (Kashyap
et al (1993) and Ludvigson (1998)). Oliner and Rudebusch (1996) argue that a change
in the mix of finance can capture the bigger decline in the amount of credit granted
to small firms compared with large firms. Others question the potency of the bank
lending channel in relation to monetary policy, especially for the United States. Some
studies argue that since banks can accumulate deposits by issuing money market
37
Bruno and Shin (2015) study the dynamics linking monetary policy and bank leverage. They construct
a model of the risk-taking channel of monetary policy in an international context. The model rests on
a feedback loop between global banks increased leverage and capital flows amid currency
appreciation for capital recipient economies. It shows that adjustments to leverage act as a linchpin
between fluctuations in risk-taking and monetary transmission.
38
Other research aimed at understanding how externalities can carry the seeds of ensuing busts
includes DellAriccia and Marquez (2006) who show how lending standards tend to weaken during
booms as adverse selection is less severe and lenders find it optimal to weaken screening and lending
standards (with the objective of trading quality for market share). This leads to deteriorating
portfolios, lower profits and a higher probability of a downward correction.
39
See Gertler and Gilchrist (1993, 1994), Friedman and Kuttner (1993), Kashyap and Stein (1995), Peek
and Rosengren (1995b) and Kakes and Sturm (2002).
A number of studies find empirical support for the bank capital channel, especially
during periods of substantial capital shortage, leading to so-called credit crunches.
For example, Lown and Morgan (2006) show for the United States that surveys of
senior loan officers, which partly reflect supply conditions, provide significant
explanatory power for US real activity. De Bondt et al (2010) show how similar lending
surveys, especially those relating to enterprises, are a significant leading indicator of
bank credit and real GDP growth in the euro area. Some studies report that credit
losses at commercial banks had some, albeit not large, regional effects on the US
recovery from the 199091 recession (Bernanke and Lown (1991), Hancock and
Wilcox (1993, 1994), Berger and Udell (1994) and Peek and Rosengren (1994)).40 These
studies found multipliers, that is the effect of a 1% change in bank capital on the
percent change in lending, ranged from 1.5 to 2.7. Ashcraft (2006) also found some
small effects of variations in commercial bank loans on real activity in normal times.
Gambacorta and Marques-Ibanez (2011) show that weaker banks in the United States
and Europe restricted loan supply more strongly during the GFC than other banks.
Other studies document varying effects. Bayoumi and Melander (2008) employ a
VAR model and report that an exogenous fall in the bank capital/asset ratio of one
percentage point reduces real US GDP by some 1.5% through weaker credit
availability. Moreover, an exogenous fall in demand of 1% of GDP is gradually
magnified to about 2% through financial feedback effects. Greenlaw et al (2008)
regress the log difference of GDP on the lagged four quarter (log) change of domestic
non-financial debt, using as instruments TED spreads and lending standards. They
find a change in credit growth of 1% to affect real GDP growth by about 0.34% in the
short run and 0.47% in the long run. By contrast, Berrospide and Edge (2010), who
update studies from the early 1990s using panel regression techniques, find a modest
effect on lending: capital shortfalls affect the extension of new loans with a range of
0.7% to 1.2% and not significant changes in GDP growth (however, in a VAR setting
they do find bank capital shocks to affect GDP growth up to 2.75%). Francis and
Osborne (2010) also report a smaller effect for UK banks.
Effects can vary by bank capitalisation and by the state of the business/financial
cycles. Bernanke (1992) and Meh and Moran (2010) find that the health of banks plays
an important role during recessions and subsequent recoveries because bank capital
can (or cannot) cushion shocks (see also Berger and Bouwman (2013)). Other studies
also find a greater role for bank capital during periods of significant credit losses or
outright shortages of bank capital. Some report that during such periods, weakly
capitalised banks limit their lending much more than highly capitalised banks (Peek
and Rosengren (1995b) and Woo (2003)). In a related paper, Ashcraft (2005) finds,
however, that it is the closure of commercial banks rather than shocks to their capital
base that leads to large and persistent negative effects on real output. He reports a
decline in real income growth of about 3% to 6% in counties where subsidiaries of
failed US banks are located.
Sectoral and event-based studies provide more direct (and sometimes clearer
causal) evidence linking bank capital to economic fluctuations. Calomiris and Mason
(2004) find that bank loan supply shocks have an impact on local area income over
40
Hancock and Wilcox (1994) investigate the impact of the bank capital channel on the US housing
market and find significant effects of the early 1990s capital crunch on commercial and residential
real estate activity.
41
The potency of the bank capital channel also depends on accounting standards and the recognition
of capital losses. In particular, the speed at which loan losses are recognised in banks balance sheets
and, consequently, their capital positions, determines in part the pace at which banks may amplify
negative aggregate shocks by cutting back on lending. At the same time, the lack of prompt
recognition of loan losses may distort banks incentives, inducing them to direct funds to inefficient
borrowers. Caballero et al (2008), for example, find evidence that the limited recognition of capital
depletion in Japanese banks slowed Japans recovery from the early 1990s recession.
42
See also Cerutti et al (2017a) on the role of global factors in driving capital flows and Cerutti et al
(2015) on how dependence on specific lenders and investors can affect countries exposure to such
global factors. See also Cerutti et al (2017b) for a recent empirical assessment. Landau (2013) and
Hartmann (2017) provide general reviews of global and international liquidity, as provided by the
private sector or by central banks, and how it relates to the designs of the international monetary
and financial systems.
43
A fire sale spiral, as first pointed out by Stiglitz (1982) and Geanakoplos and Polemarchakis (1986),
creates a negative externality and a possible rationale for regulation. Because each institution does
not bear the full cost of its own actions, it will not fully take into account the price impact of its own
fire sales on asset prices. See further Brunnermeier and Pedersen (2009), Gorton (2010) and Choi and
Cook (2012) for models of fire sales. Brunnermeier and Oehmke (2013) review the related literature
on bubbles.
44
More generally, heterogeneity among firms (and households) and specific patterns of external
financing are likely to be important factors in explaining why financial frictions can lead to relatively
large effects on business cycles (Zetlin-Jones and Shourideh (2017)).
45
See further Adrian et al (2016) on the relevance of leverage for macroeconomic modelling (and
macrofinancial linkages). They show that a parsimonious model using detrended dealer leverage as
a price-of-risk variable performs well in time series and cross-sectional tests for a wide variety of
equity and bond portfolios (at least better than models that use intermediary net worth as a state
variable) and in comparison to benchmark asset pricing models.
The previous sections documented that imperfections on the demand and supply
sides of finance could be associated with pronounced fluctuations in the real
economy. Complementary to this literature has been long standing research that
provides important insights into the general patterns of aggregate macrofinancial
linkages (see the overview of this literature in Appendix IV). Since it is hard to identify
the direction of causality between changes in financial markets and fluctuations in
real activity, and whether demand or supply channels are the main factors, many of
these studies have taken a largely agnostic approach.
Until recently, this research programme did not present a comprehensive
perspective on business and financial cycles. This was for at least two major reasons.
First, most studies consider only selected aspects of business and financial cycles. For
example, many examined the implications of booms in asset prices or credit only
rather than considering the full financial cycle. Second, research tended to focus on
case studies or used small country samples. Although the literature on financial crises
has employed broader samples, the identification of crises has often suffered from
46
For the effects of monetary policy on risk-taking, including when interest rates are particularly low,
see further DellAriccia and Marquez (2013), DellAriccia et al (2014), Jimenez et al (2014), Valencia
(2014), Ioannidou et al (2015), and Borio and Hofmann (2017). Posen (2009), Bean et al (2010) and
Bernanke (2010) argue otherwise. Adrian and Shin (2008, 2011a) and Adrian et al (2010a) analyse how
the risk-taking channel works in the United States. Some other recent studies providing further
microeconomic evidence for the presence of the risk-taking channel domestically and internationally
include Maddaloni and Peydro (2011), Altunbas et al (2014), Bruno and Shin (2015, 2017), Morais et
al (2015), DellAriccia et al (2017), and Domanski, Shin and Sushko (2017). See also Rajan (2005) and
Hanson and Stein (2015) for arguments and models linking low interest rates to search-for-yield
motives for investors other than banks. For a review of studies of the effects of low interest rates on
financial institutions' profitability and capitalization and risk-taking, see European Systemic Risk
Board (2016).
47
Claessens and Kose (2014) review a number of studies that focus specifically on periods of financial
stress and crisis and the behaviour of real and financial variables during such events. Reinhart and
Rogoff (2008, 2009a, 2009b, 2014) review the characteristics of various types of financial crisis for
many countries over a long period. Boissay et al (2016) analyse the links between credit booms,
(interbank) liquidity and banking crises (see also Allen et al (2011)).
48
The notion of financial cycles was empirically documented in early studies for smaller samples of
countries and periods of time, such as Borio et al (1994) and Borio and Lowe (2002), and refined in
subsequent work, including Drehmann et al (2012), Aikman et al (2015) and Juselius and Drehmann
(2015). This section draws on Claessens et al (2009, 2011, 2012a) which provide detailed empirical
analyses of the interactions between business and financial cycles.
49
An alternative methodology would be to consider how economic activity fluctuates around a trend
and then to identify a growth cycle as a deviation from this trend (Stock and Watson (1999)). While
several studies have use detrended series (and their second moments, such as volatility and
correlation) to study the various aspects of cycles, it is well known that the results of these studies
have depended on the choice of detrending methodology (Canova (1998)). The advantage of turning
points identified by using the classical methodology is that they are robust to the inclusion of newly
available data. In other methodologies, the addition of new data can affect the estimated trend and
thus the identification of a growth cycle. Fatas and Mihov (2013) analyse different approaches for the
dating of recoveries using US data. See also Ng and Wright (2013) for a survey of business cycles
facts and methodologies.
Business cycles
Recessions can be long, deep and costly. A typical recession lasts close to four
quarters while a recovery last about five quarters (Figure 11).51 The typical decline in
output from peak to trough, the recessions amplitude, is about 3% for the full sample
and the typical cumulative output loss amounts to about 5%. The amplitude of a
recovery, defined as the increase in the first four quarters following the trough, is
typically about 4%. Although most recessions (recoveries) are associated with
moderate declines (increases) in output, there can be much larger changes in activity
as well.
Business cycles in EMEs are more pronounced than those in advanced
economies. In particular, the median decline in output during recessions is much
larger in EMEs (4.9%) than in advanced economies (2.2%) and recoveries in EMEs are
twice as strong as those in advanced countries. In terms of cumulative loss, recessions
in EMEs are almost three times more costly than those in advanced economies. These
findings suggest that macroeconomic developments, policy factors and institutional
characteristics, including possibly the degree of financial frictions, potentially affect
the evolution of business cycles in different countries.
Business cycles are highly synchronised across countries. For some observers, the
global nature of the GFC, during which many economies experienced a recession at
the same time, was surprising. However, this is not so unusual because recessions and
50
The results reported in this section are based on a large database that comprises a total of 44
countries: 21 advanced OECD economies and 23 EMEs. For the former group, the data coverage
ranges from Q1 1960 to Q4 2010 while for the latter it ranges from Q1 1978 to Q4 2010 (because
quarterly data series are less consistently available prior to 1978). In order to study business cycles,
GDP is used because that variable is the best available measure of economic activity. Financial cycles
are studied considering three distinct but interdependent market segments: credit, housing and
equity. See further Claessens et al (2012a).
51
Claessens et al (2012a) identify 243 recessions and 245 recoveries. The number of recessions and
recoveries differs slightly because of the timing of events. There are 804 complete financial cycles
over the period Q1 1960 to Q4 2010. The sample features 253 downturns in credit, 183 in house
prices and 443 in equity prices; and 220, 155 and 429 upturns in credit, house and equity prices,
respectively. Since equity prices are more volatile than credit and house prices, they feature naturally
more often in downturns and upturns than the other financial variables.
Financial cycles
Financial cycles are often longer and more pronounced than business cycles, with
financial downturns particularly deeper and longer than recessions. Downturns
(upturns) of financial cycles tend to be longer than recessions (recoveries) (Figures 13
and 14). Episodes of house price downturns, in contrast, persist for about eight
quarters and other financial downturns last around six quarters. A typical financial
downturn corresponds to about a 6% decline in credit, 6%7% fall in house prices
and 30% decline in equity prices. Upturns are often longer than downturns, with the
strength of upturns to differ across financial markets. Equity price upturns are the
sharpest, some 26%. Financial cycles are also more intense than business cycles, ie
financial variables adjust much more quickly than real ones, as shown by their slope
coefficient. These findings are consistent with various studies documenting that asset
prices are more volatile than economic fundamentals (see Shiller (1981, 2003) and
Campbell (2003)).
The main features of financial downturns vary across EMEs and advanced
economies. While not necessarily longer, financial downturns in EMEs are much
sharper than in advanced countries. Credit contractions last about the same in both
groups but are one-third deeper in EMEs. Equity downturns last as long in both
groups but upturns are much shorter in EMEs. Comparisons between mean and
medians show that the distributions of duration and amplitude of the phases of
financial cycles are also more skewed to the right in EMEs than in advanced
economies. These differences indicate that factors possibly related to the presence of
financial frictions could affect financial cycles.
Financial cycles also tend to feed off of each other and become amplified. The
likelihood of a credit downturn (or upturn) taking place goes up substantially if there
is also a disruption (or boom) in house prices. There are also strong interactions
between financial cycles. Credit downturns that overlap with house price busts tend
to be longer and deeper than other credit downturns. Similarly, a typical credit upturn
becomes 30% longer and twice as large when it coincides with a housing boom. This
suggests that feedback effects play a role as disruptions in one market aggravate the
problems in another, probably because of collateral constraints and
complementarities between credit and housing finance. Moreover, globally
synchronised financial downturns are often longer and deeper, especially for credit
and equity markets. During highly synchronised equity market downturns, for
example, prices drop by about 30% compared with some 18% for other downturns.
Recent research using cross-country data has revealed important links between
business and financial cycles. Claessens et al (2012a) use a comprehensive database
for a large sample of advanced economies and EMEs over a long period of time to
provide a broad empirical characterisation of macrofinancial linkages. They report
three main results. First, business cycles are more closely synchronised with credit and
house price cycles than with equity price cycles. Second, financial cycles appear to
52
The concordance statistic provides a measure of the fraction of time that the two series are in the
same phase of their respective cycles. The series are perfectly procyclical (countercyclical) if the
concordance index is equal to unity (zero).
53
For a deeper perspective, it is useful to consider additional measures of credit and asset prices. For
example, some papers (Chari et al (2008) and Cohen-Cole et al (2008)) highlight the importance of
going beyond aggregate measures (for example, by differentiating credit to corporations from credit
to households) to study the dynamics of credit markets. Unfortunately, such disaggregated series are
not available for a large number of countries over long periods.
54
Analytical models also support this notion. Using the financial accelerator mechanism presented in
Section 5, for example, some studies (Aoki et al (2004) and Iacoviello (2005)) use DSGE models to
show specifically how endogenous developments in housing markets can magnify and transmit
various types of shock to the real economy and find quantitatively large effects. Mian and Sufi (2010,
2014) provide empirical evidence at the regional level for the real economy effects of mortgage credit
expansion in the United States.
The GFC was a painful reminder of the importance of macrofinancial linkages. These
linkages centre on the two-way interactions between the real economy and the
financial sector. Imperfections in financial markets can intensify these linkages and
lead to gyrations in the financial sector and the real economy. Global dimensions of
these linkages can result in spillovers across borders through both real and financial
channels.
Research on macrofinancial linkages has a long tradition, but has become a
central topic only over the past three decades. This paper reviews this rich literature
and presents a broad perspective on theoretical and empirical work. The survey
considers the two channels the demand and supply sides through which financial
imperfections can affect macroeconomic outcomes. The demand side channel, largely
captured by financial accelerator-type mechanisms, describes how, through changes
in the balance sheets of borrowers, financial markets can amplify macroeconomic
fluctuations. The supply side channel emphasises the importance of changes in
financial intermediaries balance sheets for the availability of external financing and
liquidity, the role of financial markets in the determination of asset prices and the
implications of those factors for the real economy.
A summary
The literature has made significant progress in understanding the demand side of
macrofinancial linkages. Many models now feature amplification mechanisms
operating through the demand side. These models show how a financial accelerator-
type mechanism can lead to the propagation and amplification of small (real or
financial) shocks across the real economy (through their impact on access to finance).
A number of models that incorporate financial accelerator mechanisms show the
importance of changes in asset prices and other financial shocks in driving
movements in borrowers net worth and access to finance, leading to fluctuations in
aggregate activity.
These analytical findings are also supported by empirical studies. In particular,
extensive evidence documents how the state of borrowers balance sheets affects
their access to external finance. Demand side imperfections in financial markets have
been shown to lead to an amplification of shocks (monetary, real and financial)
because changes in the net worth of borrowers affect their access to finance and,
therefore, to consumption, investment and output. Empirical studies confirm that
these imperfections tend to affect small firms and households particularly strongly,
especially during periods of financial stress. Although most findings support the roles
played by financial imperfections, there is nevertheless an ongoing debate about the
quantitative importance of the financial accelerator.
The GFC has shifted attention to the critical role played by amplification channels
operating through the supply side of finance. Earlier theoretical work on the bank
lending channel analysed the possible general equilibrium effects arising from the
special role of banks in financial intermediation. Empirical studies documented how
the dependence of some firms on bank financing influences the transmission of
monetary policy to the real economy. Since the GFC, there has been a broader
recognition that the supply side of finance (beyond the specific role played by banks
for some firms) can be a source of shocks, amplification and propagation.
What next?
55
Several surveys discuss advances in the modelling of heterogeneous agents, news shocks, financial
crises, bubbles and systemic risk (Heathcote et al (2009), Lorenzoni (2011, 2014), and Brunnermeier
and Oehmke (2013)). Another strand of the literature uses ad hoc borrowing constraints to model
financial imperfections in environments with a continuum of households (Huggett (1993), Aiyagari
(1994), Krusell and Smith (1998)), which is particularly useful in studying distributional issues. Moll
(2014) studies an environment in which financial frictions lead to misallocation of resources.
56
See Acemoglu et al (2015) and Boissay et al (2016) for analysis focusing on effects of networks and
links among financial institutions, including through the interbank market. See Freixas et al (2011) for
analysis on monetary and prudential policies in the interbank markets. See Dou et al (2017) for a
review of the role of macrofinancial interactions in dynamics models used at central banks.
57
Bernanke (2010), Blanchard et al (2010, 2013), Caballero (2010), Kocherlakota (2010), Woodford
(2010), Turner (2012), Claessens et al (2014), Kose and Terrones (2015), Blanchard and Summers
(2017) and several papers in the Fall 2010 issue of the Journal of Economic Perspectives look at how
the GFC may have influenced research, including the literature on the intersection between
macroeconomics and finance. Blanchard et al (2012, 2016) and Akerlof et al (2014) discuss how
economic policies have been reassessed by economists since the GFC. Gopinath (2017) provides a
review of the recent macroeconomic policy-related work in international economics.
58
See, for example, Dewatripont and Tirole (2012), Stein (2012), Goodhart et al (2013), Admati and
Hellwig (2014), Fender and Lewrick (2016), Kara and Ozsoy (2016) and Kashyap et al (2017).
59
See Eggertsson and Krugman (2012) and Eggertsson et al (2016); see also Borio (2017) on the possible
role of monetary policy in affecting the real interest rate.
60
See the early contributions by Crockett (2000), Borio (2003) and Knight (2006) (see also Clement
(2010)). Subsequently, the BIS and other related entities have conducted much research on various
aspects of macroprudential policy. The Committee on the Global Financial System (CGFS), for
example, reported on operational aspects in 2010, 2011 and 2012 (CGFS (2010, 2012) and CGSG
(2011)). An influential, post-GFC take was Brunnermeier et al (2009). Recent work on macroprudential
policy includes Bianchi et al (2012, 2016), Farhi and Werning (2016) and Bianchi and Mendoza
(forthcoming).
61
See Jeanne (2014, 2016), Pereira da Silva (2016) and Agnor et al (2017).
Research on macrofinancial linkages has a long history. The Great Depression created
much interest in such linkages but this interest faded away over the next few decades.
There has been a resurgence of interest since the early 1980s, with the introduction
of rigorous general equilibrium models that have provided rich theoretical insights.
Such insights have been complemented by the results of empirical studies at the
macroeconomic and microeconomic levels. This appendix provides a brief review of
the evolution of this literature (see Figure A1 for a schematic presentation).
The study of credit cycles, which precedes that of business cycles, goes back at
least to Mills (1867) but, as just mentioned, the Great Depression was the primary
motivation for the early qualitative work on the role of financial factors in shaping
macroeconomic outcomes. Fisher (1933) provided a descriptive account of the
relationship between the high leverage of borrowers and the severity of the downturn
during the Great Depression. His debt-deflation mechanism was the first elegant
narrative showing how a decline in net worth induced by a drop in prices, ie deflation,
could lead borrowers to reduce their spending and investment, which, in turn, could
cause activity to contract and result in a vicious cycle of falling output and deflation.
Haberler (1937) reviewed early studies of business cycle fluctuations, focusing on the
so-called monetary, over-investment, under-consumption, and psychological
theories.
The literature that followed, however, turned its attention to the role of money,
rather than credit, as the critical financial variable. While Keynes (1936) also brought
out financial developments, eg as he discussed how the confidence of borrowers and
lenders could changes in ways nor easily explained with economic models (animal
spirits). He focussed more on the importance of money for the real economy. Armed
with insights from the liquidity preference hypothesis, the early Keynesian models
paid special attention to the mechanisms linking money to real activity, including the
multiplier mechanism and the role of fiscal policy (Hicks (1937), Modigliani (1944) and
Tobin (1958)). The monetarist school, on the other hand, insisted on the importance
of monetary rather than real factors (Friedman (1956), Friedman and Schwartz (1963)
and Tobin (1969)).
Later studies documented the critical role of financial intermediation in economic
development and macroeconomic fluctuations but these did not lead to a
fundamental shift in thinking. Gurley and Shaw (1955) showed that economic
development and financial sophistication go hand in hand. Others, including
Cameron (1961), Shaw (1973) and McKinnon (1973), also highlighted the importance
of finance for development, contrasting among others the East Asian and Latin
American experiences. Their main argument was that a countrys overall financial
capacity, ie its financial systems ability to provide credit, was more relevant to the
real economy than money. In the early stages of financial development, money could
be important but it becomes less relevant in more developed systems, particularly as
a measure of credit availability. Instead, banks increasingly use non-deposit sources
of funding and non-bank intermediaries provide alternative sources of financing. This
view, while also advanced by many financial historians (eg Goldsmith (1969)), did not
prevail and research on finance and macroeconomics followed separate paths.
New analytical insights were gained during the 1950s with the applications of
portfolio theory. A major breakthrough was the introduction of portfolio theory by
Macroeconomics Finance
Studies Approach Approach Studies
Mills (1867)
Fisher (`33) Qualitative discussions of macrofinancial linkages
Haberler (`37)
Keynes (`36) Narratives of debt-deflation
Marschak (`38)
WP677 Macroeconomic implications of financial imperfections: a survey
Gurley-Shaw (`55)
Hicks (`37) Financial
Goldsmith (`69)
Modigliani (`44) Keynesian approach intermediation
Shaw (`73)
Tobin (`58) Financial repression
Cameron (`67)
Portfolio theory
Markowitz (`52)
Friedman-Schwartz Capital structure
Monetarist approach Modigliani-Miller (`58)
(`63) irrelevance
Tobin (`69)
Q ratio
Kahneman-Tversky (`79)
Brunner-Meltzer (`88) Bubbles, crises Shiller (`81)
New Keynesian DSGE
Bernanke-Gertler-Gilchrist (`99) Market imperfections Minsky (`86)
models
Smets-Wouters (`03) Limits of arbitrage Fazzari-Hubbard-Petersen (`88)
Inflation targeting Shleifer-Vishny (`97)
Woodford (`03) Behavioural finance
Thaler (`05)
Notes: This diagram presents a rough representation of the evolution of research on macrofinancial linkages. Early interest in such linkages was triggered by the Great Depression. However, it
remained largely qualitative. Although research became more analytical over time, it moved away from the analysis of financial market imperfections, especially over the period 195080. Since then,
the literature has made significant progress in capturing the effects of microeconomic imperfections. More recently, it has combined theoretical insights from macroeconomics and finance in
general equilibrium models. It has also produced many empirical studies at the micro- and macroeconomic levels. This diagram includes only a select set of studies because of space constraints.
We summarize many other studies in the main text.
49
Appendix II. Financial accelerator mechanisms
Although there are many variants of the financial accelerator mechanism, they tend
to describe similar channels of transmission and propagation. A number of theoretical
studies show that the mechanism can play an important role in accentuating
macroeconomic fluctuations. This appendix presents a summary of the various
mechanisms found in the literature (see Quadrini (2011) for a systematic analytical
review of the causes of financial frictions and the types of accelerator model; Gerke
et al (2013) also compare the various models that feature a financial accelerator
mechanism and collateral constraints).
One mechanism works through changes in cash flows that depend on the overall
state of the economy. Greenwald and Stiglitz (1993) develop a dynamic model in
which changes in corporate cash flows play a critical role in propagating financial
market disturbances to real sector variables, including employment and inventories.
Another mechanism acts through changes in the intensity of adverse selection in
credit markets. Azariadis and Smith (1998) show that the presence of adverse
selection can create indeterminacy, with the economy fluctuating between Walrasian
and credit rationing regimes, and with cyclical downturns exhibiting declines in real
interest rates and increases in credit rationing (see also Mankiw (1986)).
Changes in default probabilities over the cycle can also amplify cyclical
fluctuations. During the upswing of a cycle, default probabilities decline, allowing
investors to lend greater sums of money, as predicted by the Stiglitz-Weiss model
(1981) of lending under asymmetric information and moral hazard. This fuels the
upswing. On the downswing, the mechanism works in reverse, leading to sharp
declines in available external financing. The general equilibrium implication of this
mechanism is modelled by Suarez and Sussman (1997) using an overlapping
generations model. In particular, during booms old entrepreneurs sell larger
quantities and prices fall, implying that young entrepreneurs must rely to a greater
extent on external sources of financing. Since external financing can generate
excessive risk-taking, booms are often followed by higher failure rates. Fire sales by
bankrupt corporations during such periods then lead to asset price declines, which,
in turn, generate macroeconomic fluctuations.
The allocation of capital across heterogeneous firms can, in the presence of
financial imperfections, also create procyclicality. Bernanke and Gertler (1989) and
Kiyotaki and Moore (1997) focus on the impact of financial imperfections on
aggregate investment and employment but neglect the effects of the reallocation of
inputs across heterogeneous firms. Recent studies try to fill this gap. Eisfeldt and
Rampini (2006) find that the costs of capital reallocation, such as those induced by
financial frictions, need to be countercyclical to be consistent with the cyclical
dynamics of capital reallocation and productivity dispersion. Moll (2014), in a model
where entrepreneurs are subject to borrowing constraints and idiosyncratic
productivity shocks, shows with plant-level panel data that financial frictions can
explain aggregate productivity losses in two EMEs that are 20% larger than in the
United States. In related research, Khan and Thomas (2013), using a DSGE model in
which capital reallocation across firms is distorted by frictions, show that a shock can
be amplified and propagated through disruptions to the distribution of capital across
firms. Herrera et al (2011), in their study of US firms and the dynamic properties of
gross credit flows relative to macroeconomic variables, find that financial frictions can
impact aggregate productivity by hindering the inter-firm reallocation of credit.
Using various methodologies and measures to proxy cycles, a number of studies have
examined the features of business and financial cycles and the aggregate linkages
between such cycles. They have pointed out the procyclical nature of financial markets
and provided the broad patterns describing the linkages between business and
financial cycles. This Appendix reviews these studies for the three most important
market segments: credit, equity and housing.62
The study of credit cycles has a history that goes back to Mills (1867) at least. Most
of the early work in this area employed qualitative approaches and considered the
extreme versions of these cycles: booms and busts (or crunches) (see Keynes (1936),
Galbraith (1954), Shiller (1989, 2000) and Sinai (1993); Niehans (1992) reviews very
early work on credit cycles (Juglar (1862)). A number of studies also consider specific
credit crunches in the United States and other countries (see Wojnilower (1980, 1985),
Owens and Schreft (1995), Cantor and Wenninger (1993) and Helbling et al (2011)).
Using US data going back to 1875, Bordo and Haubrich (2010) document that credit
disruptions appear to exacerbate cyclical downturns. A number of studies also
consider the important role played by credit in driving business cycles. Using VAR
models, Meeks (2012) examines the role of credit shocks in explaining US business
cycles and finds that such shocks play an important role during financial crises but a
somewhat smaller role during normal business cycles.
Recent studies apply a variety of quantitative approaches to cross-country data
to analyse episodes of credit booms and crunches. Mendoza and Terrones (2008), for
example, use a thresholds method to identify credit booms in 48 countries over the
period 19602006. They find that booms generally coincide with above-trend growth
in output, consumption and investment during the build-up phase and below-trend
growth of those variables in the unwinding phase. During the build-up phase, a surge
in private capital inflows is accompanied by a deterioration of current account
positions (see also Gourinchas et al (2001), Schularick and Taylor (2012)). Other
researchers (such as Castro and Kubota (2013) and DellAriccia et al (2016)) also study
the determinants of credit booms length.
Booms and busts in asset prices have also been a major area of research. Borio and
Lowe (2002), using an aggregate index of asset price (equities, and residential and
commercial property), define booms as periods during which asset prices deviate
from their trends by specified amounts. They also consider the interaction between
developments in asset prices and credit. They report that there are substantial
declines in house prices and residential investment during housing busts (after
62
For additional references on the literature covering the linkages between business and financial cycles
see Rebelo (2005), Claessens et al (2009, 2011, 2012a), Siregar and Lim (2011), Guarda and Jeanfils
(2012), Hubrich et al (2013), Borio (2014), De Rezende (2014), Groe Steffen (2015), Hubrich and
Tetlow (2015), Kose and Terrones (2015), Hartmann et al (2015), Abbate (2016), Abildgren (2016),
Jord et al (2016, 2017), Bluwstein (2017) and Gandr (2017).
Note: Each cell represents the concordance statistic for the corresponding two cycles. Concordance
is calculated as the fraction of time that two cycles are in the same phase. *** and ** imply
significance at the 1% and 5% levels, respectively. Significance refers to the difference between
advanced economy and emerging markets (means and medians only).
-3
-6
-9
-12
Large Small
-15
Sales Inventories Short-term Debt
Source: Kudlyak and Snchez (2017).
Note: The figure presents the difference between the minimum value of the detrended series in an interval of
12 quarters following an episode of tight money and the peak value of the series. Tight money periods are:
Q2 1966, Q4 1968, Q2 1974, Q3 1978, Q4 1979, Q4 1988 and Q2 1994 based on the historical record analysed
in Romer and Romer (1989). Small firms are defined as those at or below the 30th percentile of assets and
large firms as those above the 30th percentile.
Note: The chart depicts the financial accelerator mechanism by which shocks to the economy may be amplified
by changes in access to external financing, which then translates into changes in economic agents investment
and consumption spending. In turn, these changes are propagated and reinforced as asset prices and
economic activity fluctuate, which then affects the demand and availability of external financing. This creates
further feedback loops that are propagated through financial markets and the real economy.
Note: The chart depicts both the demand (as in Figure 4A) and the supply sides of the financial accelerator
mechanism. As financial institutions balance sheets and profitability increase and asset prices rise, the
assessment of economic prospects is viewed more positively and the supply of external financing expands.
This then translates into changes in investment that enhance the feedback loops between financial markets
and the real economy.
-5
-10
-15
-20
Actual Counterfactual
-25
2007:3 2007:4 2008:1 2008:2 2008:3 2008:4 2009:1 2009:2 2009:3 2009:4
2.0
Credit
1.5 Productivity
1.0
0.5
0.0
-0.5
2007:3 2007:4 2008:1 2008:2 2008:3 2008:4 2009:1 2009:2 2009:3 2009:4
Source: Helbling et al (2011).
Note: Panel A compares the results of counterfactual simulations for the global GDP factor during the GFC.
The solid line represents the actual global GDP factor and the dashed line represents the counterfactual when
the global credit shock is set to zero during the period considered. Panel B compares the contributions of
credit and productivity shocks to cumulative global GDP growth based on the counterfactual simulations. The
bars represent the median difference. A positive (negative) bar captures how the decrease in the global GDP
factor would have been smaller (greater) in the absence of the respective shocks. Credit is defined as the
aggregate claims on the private sector by deposit money banks and is obtained from the IFS. Labour
productivity is defined as real GDP per hours worked and is obtained from the OECD.
5
6
4
4
3
2 2
1
0
0
-2
-1
-4 -2
-1 -0.5 0 0.5 1 1.5 -2.5 -1.5 -0.5 0.5 1.5 2.5
Leverage Growth (Percent Quarterly) Leverage Growth (Percent Quarterly)
5
Total Asset Growth (Percent Quarterly)
30
4
20
3
10
2
0
1
-10
0
-1 -20
-2 -30
-50 -30 -10 10 30 50 -60 -40 -20 0 20 40
Leverage Growth (Percent Quarterly) Leverage Growth (Percent Quarterly)
6 Recoveries
**
4
0
Full Sample Advanced Economies Emerging Market Economies
Amplitude
(in percent)
8
Recessions
6
Recoveries
4 ***
2
0
-2
-4 ***
-6
Full Sample Advanced Economies Emerging Market Economies
-2
-4
***
-6
-8
-10
Full Sample Advanced Economies Emerging Market Economies
Source: Claessens et al (2012a).
Note: All the statistics except for those relating to duration correspond to sample medians. For duration, the
means are shown. The duration of a recession is the number of quarters that have elapsed between the peak
and the trough. The duration of a recovery is the time taken to attain the level of output reached at the previous
peak. The amplitude of a recession is the decline in output from peak to trough. The amplitude of a recovery
is the one-year change in output after the trough. The cumulative loss combines information about the
duration and the amplitude to measure the overall cost of a recession. ***, ** and * imply significance at the
1%, 5% and 10% levels, respectively. Significance refers to the difference between advanced economies and
EMEs.
80
60
40
20
0
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Source: Claessens et al (2012a).
Note: The share of countries experiencing a recession is presented. The figure includes completed as well as
ongoing episodes. The sample contains quarterly data for 21 advanced economies over the period Q2 1960
to Q4 2010.
Amplitude
(in percent)
0
-10
-20
-30
*** ***
-40
-50 Downturns
-60 Crunches/Busts
***
-70
Credit House Price Equity Price
Slope
(in percent)
0
-2
***
-4 ***
-6
***
Downturns
-8
Crunches/Busts
-10
Credit House Price Equity Price
Source: Claessens et al (2012a).
Note: the amplitude and slope statistics correspond to sample medians. For duration, the means are shown.
Duration is the number of quarters between peak and trough. Amplitude is based on the decline in each
variable during the downturn. Slope is the amplitude divided by the duration. Busts (crunches) are the worst
25% of downturns calculated by the amplitude. ***, ** and * imply significance at the 1%, 5% and 10% levels,
respectively. Significance refers to the difference between busts (crunches) and other financial downturns.
6
3
0
Credit House Price Equity Price
Amplitude
(in percent)
70
***
60
Upturns Booms
50
40
30
20 ***
***
10
0
Credit House Price Equity Price
Slope
(in percent)
12
***
10
Upturns Booms
8
6
4 ***
***
2
0
Credit House Price Equity Price
Source: Claessens et al (2012a).
Note: The amplitude and slope correspond to sample medians. For duration, the means are shown. Duration
is the time it takes to attain the level of the previous peak. Amplitude is the change in one year after the trough
of each variable. Slope is the amplitude from the trough to the period where the financial variable reaches its
last peak, divided by duration. Booms are the top 25% of upturns calculated by the amplitude. ***, ** and *
imply significance at the 1%, 5%, and 10% levels, respectively. Significance refers to the difference between
financial booms and other financial upturns.
Amplitude
(in percent)
0
-1
-2
-3 **
-4 * ***
Recessions associated with Recessions associated without
-5
Credit Crunches House Price Busts Equity Price Busts
Cumulative loss
(in percent)
0
-2
-4
-6
-8 *** ***
**
Recessions associated with Recessions associated without
-10
Credit Crunches House Price Busts Equity Price Busts
Source: Claessens et al (2012a).
Note: The amplitude and cumulative loss statistics correspond to sample medians. Duration corresponds to
the sample mean. Disruptions are the worst 25% of downturns as represented by the amplitude. ***, ** and *
imply significance at the 1%, 5% and 10% levels, respectively. Significance refers to the difference between
recessions with and without financial disruptions. For other definitions, see the notes to Figures 13 and 14.
Amplitude
(in percent)
10
*** Recoveries associated with
8
Recoveries associated without
***
6
0
Credit Booms House Price Booms Equity Price Booms
Slope
(in percent)
3
Recoveries associated with
*** Recoveries associated without
2
***
1
0
Credit Booms House Price Booms Equity Price Booms
Source: Claessens et al (2012a).
Note: The amplitude and slope correspond to the sample medians. Duration corresponds to the sample means.
Booms are the highest 25% of upturns by amplitude. ***, ** and * imply significance at the 1%, 5% and 10%
levels, respectively. Significance refers to the difference between recoveries with and without booms. For other
definitions, see the notes to Figures 13 and 14.
No Title Author
676 Asset prices and macroeconomic outcomes: a Stijn Claessens and M Ayhan Kose
November 2017 survey
675 Macroprudential Policies in Peru: The effects Elias Minaya, Jos Lup and Miguel
November 2017 of Dynamic Provisioning and Conditional Cabello
Reserve Requirements
673 Loan-to-value policy and housing finance: Douglas Kiarelly Godoy de Araujo,
November 2017 effects on constrained borrowers Joao Barata Ribeiro Blanco Barroso
and Rodrigo Barbone Gonzalez
669 Credit misallocation during the European Fabiano Schivardi, Enrico Sette and
November 2017 financial crisis Guido Tabellini
668 Financial and real shocks and the Javier Garcia-Cicco, Markus Kirchner,
October 2017 effectiveness of monetary and Julio Carrillo, Diego Rodriguez,
macroprudential policies in Latin American Fernando Perez, Rocio Gondo,
countries Carlos Montoro and Roberto Chang
666 Bank capital allocation under multiple Tirupam Goel, Ulf Lewrick and
October 2017 constraints Nikola Tarashev
665 Interest rates and house prices in the United Gregory D Sutton, Dubravko
October 2017 States and around the world Mihaljek and Agne Subelyte
664 Is the price right? Swing pricing and investor Ulf Lewrick and Jochen Schanz
October 2017 redemptions