Macroeconomics and Monetary Economics
Summary and Keywords
In recent decades, macroeconomic researchers have looked to incorporate financial intermediaries explicitly into business-cycle models. These modeling developments have helped us to understand the role of the financial sector in the transmission of policy and external shocks into macroeconomic dynamics. They also have helped us to understand better the consequences of financial instability for the macroeconomy. Large gaps remain in our knowledge of the interactions between the financial sector and macroeconomic outcomes. Specifically, the effects of financial stability and macroprudential policies are not well understood.
Following the recent financial crisis, there has been an intensification of efforts by economists and economic historians to understand better the role of financial factors in business cycles, especially around periods of severe economic disruption. There is still no clear consensus on what those lessons are. The role played by many factors, such as financial innovation, financial regulation, and other government policies, remains far from clear. This article selectively reviews those research efforts. The focus is particularly, although not exclusively, on how insights from the long-run historical record of advanced economies are being developed and reflected in frictions in financial intermediation incorporated into closed-economy dynamic, stochastic general equilibrium (DSGE) models.
The paper’s principal aims, therefore, are (i) to review the core workhorse models of financial frictions that existed before the recent financial crisis; (ii) to identify the key facts emerging from the historical record as to what features ought to be built into DSGE models with financial frictions; (iii) to summarize the key modeling developments around credit intermediation since the crisis; and (iv) to identify gaps in the literature that are especially important for policymakers and modelers.
Models of financial frictions are not necessarily models of financial crises. The precrisis workhorse macroeconomic models incorporating financial frictions, as discussed in more detail later in this article, focused on balance-sheet constraints facing nonfinancial firms. Recent research efforts, in part reflecting new insights from long-run historical analyses, have sought to incorporate leveraged financial institutions into DSGE models. One result of the new models has been to extend the financial accelerator mechanism to new sets of agents. These more recent models, by introducing frictions and asymmetric information problems between banks, as well as by working in explicitly nonlinear environments, have introduced financial crises into DSGE models. These nonlinear models capture the possibility that balance-sheet constraints may not always bind. Financial crises then are associated with periods during which the constraints do in fact bind. In this way, sharp changes in the level of aggregate economic activity can be modeled.
Incorporating some of the insights of Diamond and Dybvig (1983) is another way to introduce nonlinearity into macro models. These authors pointed out that banks may be structurally vulnerable to runs, and that these runs could have a self-fulfilling element; in the absence of the run, banks are solvent, while in the presence of the run, they are insolvent. That is not to say, of course, that banks may not become insolvent in the absence of a run. These extensions to the workhorse models reflect, in part, a substantial body of recent work that seeks to understand better the historical record of financial crises—especially banking crises—coupled with the role of secured credit growth and asset price inflation in indicating the probability of a crisis.
The next section outlines, relatively briefly, the workhorse models that existed before the recent financial crisis, as well as some of the criticisms that have been leveled against those models. Then, some key themes emerging from the historical analyses of financial crises are summarized.1 Our aim is to try to uncover features of these studies that may help inform the construction of new models useful for policy analysis. Next, some of the areas where the lessons from historical analyses are disputed or remain unclear, or where potentially significant gaps remain, are discussed. These two sections indicate that the two core models, despite their insights, lacked a financial intermediaries sector that could connect with some key stylized facts emerging from historical analyses, as well as those from the 2008 crisis. The paper then summarizes some of the key DSGE models that have been developed since the crisis, emphasizing the incorporation of financial intermediaries. The intention is particularly to emphasize the key ideas underlying these extensions to the core models, rather than attempting a detailed technical exposition of these models. Finally, important questions for future research are outlined.
Core Macro Models With Financial Frictions: A Brief Overview
The dominant perspective on financial frictions in macroeconomic models before the 2007/2008 crisis derived from two related approaches, which highlighted credit market imperfections facing nonfinancial borrowers.2 The first of these was developed by Bernanke and Gertler (1989), Carlstrom and Fuerst (1997), and Bernanke, Gertler, and Gilchrist (1999), who situated it in a fully specified, sticky-price DSGE model. These contributions drew on the costly state verification literature pioneered by Robert Townsend (1979). The key idea was that a debt contract between a borrower and a lender was the best way to overcome asymmetrical information between the parties to the contract. Moreover, the more capital the borrower could sink into the project, the lower the debt service costs would be.3 The second approach comes from Kiyotaki and Moore (1997), drawing on the idea of the inalienability of human capital emphasized by Hart and Moore (1994). Borrowers may walk away from a project, taking their valuable skills with them and jeopardizing a lender’s investment. Therefore, the lender requires that the value of capital to which they have recourse in a bankruptcy at least matches the value of the outstanding debt.
Different in a number of details, these two approaches nevertheless share a number of common features.
First, leverage and exposure to aggregate shocks are key to both, as are feedback effects from asset markets (so-called fire sale effects). Specifically, firms are unable to purchase insurance against productivity and other aggregate shocks, so following a negative productivity shock, firms sell capital, which compounds the initial negative shock. Second, and related, both are intended to capture the situation where credit frictions amplify other shocks, notably productivity shocks (and monetary policy shocks in the case of Bernanke et al., 1999).
That amplification has both intratemporal and, more important, intertemporal elements. Hence, both approaches emphasize the impact of financial frictions via the production of investment goods. In the Bernanke et al. (1999) model, the entrepreneur’s net worth affects the production of new capital goods, which in turn affects the next period’s production. In the Kiyotaki and Moore (1997) model, due to the presence of collateral constraints, higher net worth permits entrepreneurs to purchase more capital and increase production. In both models, the financial friction amplifies the impact of productivity and other shocks and lays the foundation for more volatile and protracted economic fluctuations. The top panel of Figure 1 portrays what happens when there are no financial frictions. In that case, changes in returns on investment and asset prices work to mitigate the initial negative shock, helping to stabilize aggregate economic activity more quickly. The bottom panel portrays what happens when a financial accelerator is present. Now, the equilibrating role of asset price changes is disrupted as the value of collateral, not just the expected return on investment, is altered.
Core Models: Extensions
These workhorse models led to a large body of literature examining these frictions in a variety of model environments. Some models have focused on the interaction between mortgage lending, house prices, and the wider economy.4 An important contribution is Iacoviello (2005), which builds on the approach of Kiyotaki and Moore (1997). He constructs a New Keynesian (sticky-price) model in which a subset of agents borrow as much as they can in nominal terms to purchase housing services. Firms also are constrained in their borrowing capacities as the aggregate value of loans is tied to the value of the housing stock. The rich structure of the model yields substantive insights relative to Bernanke et al. (1999) and Kiyotaki and Moore (1997). Shocks that induce a negative correlation between prices and output (such as a negative productivity shock) are ameliorated (as debtors’ net worth is boosted by rising prices). However, following a demand shock, a financial accelerator mechanism kicks in as prices in the economy, including asset (house) prices, rise. That increases the borrowing capacity of debtors while reducing the real value of their outstanding debt. This process boosts debtors’ net wealth and increases aggregate demand (because debtors’ marginal propensity to spend is larger than lenders’). Together, these effects explain why Iacoviello finds that nominal debt improves the model economy’s inflation-output variance trade-off (compared to an economy with indexed debt). Iacoviello (2005) also shows that the combined role of collateral and nominal debt helps the model fit the US data’s implications for house price shocks. The idea that the housing market may have wider implications for the economy has been a theme of much subsequent research; see for example, Iacoviello and Neri (2010).5
Building on Bernanke et al. (1999), other researchers have sought to examine whether these models provided a justification for monetary policy to “lean against the wind” (e.g., Gilchrist & Leahy, 2002); that is, to raise interest rates in response to rising asset prices. The consensus appeared to be that such a policy added little to the efficacy of monetary stabilization policy.6 Another branch of the literature has attempted to assess the wider empirical properties of the workhorse models. Some researchers argued that while financial friction made investments more volatile, its impact on capital stock was much less substantial. As a result, the labor input (a complement with capital) was not variable enough to match the US data. The consensus that emerged was that these frictions did not appear to be quantitatively substantial.7
That led some researchers to add shocks to the friction. Nolan and Thoenissen (2009) do just that, arguing that the financial sector may also be a source of shocks. They suggest that extension helped the model to capture the US data better, and that financial frictions shocks tended to be quite long-lasting and quantitatively significant. Specifically, they found that these shocks (i) were linked tightly with the onset of recessions, more so than total factor productivity (TFP) or monetary shocks; (ii) remain contractionary after recessions have ended; (iii) account for a large part of the variance of gross domestic product (GDP); (iv) are generally much more important than money shocks; and (v) are strongly negatively correlated with the external finance premium. Gilchrist, Yankov, and Zakrajšek (2009) via a very different empirical strategy to Nolan and Thoenissen (2009) also argued that credit market shocks contributed significantly to US economic fluctuations over the period 1990–2008. Other researchers found similar results using a variety of estimation procedures.8 Hence, some argued that financial frictions appeared to be significant once one allowed for a stochastic, time-varying component to the friction. However, while a useful result (since the extension need not have helped these models fit the data), it suggested that more work was needed to understand that time-varying feature of the friction.
So, for some (see, e.g., Hall, 2010), these workhorse models provide a good basis for understanding at least the initial contraction in durable expenditure and aggregate economic activity following the onset of the financial crisis.9 For others, these models lack important features of modern financial economies (such as financial intermediaries and interbank lending) or sufficiently robust microfoundations.
Core Models: Assessment
The core models provide a mechanism, grounded in microeconomics, for incorporating financial frictions into the main body of macroeconomic theory. These models have proved useful for studying a variety of theoretical and applied questions and have been shown to connect in meaningful ways with actual data. Nevertheless, the workhorse models have a number of important shortcomings.
First, neither the Bernanke et al. (1999) nor the Kiyotaki and Moore (1997) approach develops a detailed role for financial intermediaries. Second, the microfoundations of the contracts/lending arrangements are indistinct, meaning that policy recommendations, especially macroprudential recommendations, are unclear. For instance, the Townsend (1979) model of external finance does not typically promote debt contracts as optimal, at least not without considerable modification (see Krasa & Villamil, 2000; Duncan & Nolan, 2017a). As this microfoundation is not fully understood, the effects of policy changes on external finance contracts (and, ultimately, on financial stability) remain unclear. Third, and related to the previous point, both approaches assume that financial markets, or other mechanisms to hedge macro risks (and hence the impact of financial frictions), are closed, despite there being no private information problem. For example, Carlstrom, Fuerst, and Paustian (2016) show that agents in the Bernanke et al. (1999) model can write better contracts than Bernanke et al. (1999) assumed, essentially neutralizing the effect of the financial accelerator.10 Fourth, these models do not feature default, arguably a central issue during financial crises. Fifth, excluded from the core models are many institutions (e.g., financial intermediary external capital and other financial regulations, deposit protection, or tax shields for debt contracts) that reasonably may be expected to have a substantial impact on the vulnerability of the macroeconomy to certain shocks and the incentives underlying the credit intermediation process.
It is argued later in this article that progress has been made in making the core models descriptively richer, but that the microfoundations remain unclear. Moreover, the role of financial regulations is largely absent in recent contributions. The upshot is that the guidance that the literature is able to provide to microprudential (and perhaps especially macroprudential) policymakers remains highly tentative.
Before turning to the recent theoretical extensions to the core models, the next section discusses a mostly recent body of literature that seeks to uncover the key facts of financial crises in advanced economies.
Facts and Features of Financial Crises
The principal focus here will be on the literature on banking crises.11 Highly leveraged financial intermediaries—traditionally banks—appear to have played a central role in many so-called crisis events. And there is evidence that banks are, in some sense, special.12 For example, Bernanke (1983) is an influential paper suggesting that bank failures and the damage done to the efficiency of credit intermediation are important for understanding the persistence of the Great Depression. More recently, Giesecke, Longstaff, Schaefer, and Strebulaev (2014) contrast bond default crises with banking/credit crises (the latter taken from Schularick & Taylor, 2012, as discussed later in this article). The authors constructs a data set to study corporate bond default crises in the United States from 1866 to 2010, a sample that includes examples of widespread corporate default.13 Typically, banks are more important overall as a source of debt finance, although both sources are significant in absolute terms. The ratio of bank loans to GDP averages 33.2% over the sample, while the ratio of corporate bonds to GDP averages 19.2%.14 Moreover, bonds are typically unsecured.
Following World War II, bank lending grew more rapidly than corporate bonds (and currently is about twice as large). The authors find that banking crises and corporate default crises were largely distinct events; that is, there appeared to be a low correlation between the timing of corporate default and banking crises. Moreover, bank lending increased significantly shortly after a corporate default crisis, but the corporate debt market did not pick up after a banking crisis.
Large corporate bond issuers seemed able to substitute sources of credit after a corporate default crisis, thereby mitigating the impact of the credit channel mechanism. These findings, the authors argue, seem broadly consistent with experience in the recent crisis, insofar as big firms suffered less restricted access to credit than did small and medium-sized enterprises.
Meanwhile, declines in housing and stock market values led to notable declines in subsequent bank lending, they had no apparent effect on the subsequent amount of corporate bond issuance. That appears to support the Kiyotaki and Moore (1997) insight that collateral values and credit availability are important and that they may be central to understanding the macroeconomic impact of banking crises. Moreover, Giesecke et al. (2014) show that unlike banking crises, corporate default crises do not appear to have a significant or systematic effect on aggregate activity (GDP or industrial production). Interestingly, Azariadis, Kaas, and Wen (2016) find that, for the US economy over the period 1981–2012, unsecured debt is quite strongly procyclical and with some tendency to lead GDP. On the other hand, secured debt seems more or less acyclical. They, therefore, argue that the Kiyotaki and Moore (1997) model is not supported by the data. However, it is worth noting that the Giesecke et al. (2014) analysis is specifically focused on periods of crisis. In any case, this is one of a number of complex signals emerging from recent empirical studies. More such instances are examined later in this article.
One implication, then, from Giesecke et al. (2014) is that in analyzing crisis periods, it is important to distinguish secured from unsecured borrowing, both for corporate and noncorporate borrowers, because the credit and collateral mechanisms are largely absent in bond default crises. Another lesson is the importance of credit growth for wider economic stability.
Schularick and Taylor (2012) is an influential study of credit growth and banking crises. Banking crises are events during which a country’s banking sector experiences bank runs, sharp increases in default rates leading to capital losses, public intervention, bankruptcy, and possibly forced merger of financial institutions. Their crisis dates are largely derived from studies by Bordo, Eichengreen, Klingebiel, and Martinez-Peria (2001) and Reinhart and Rogoff (2009) for the pre–World War I era, and from Laeven and Valencia (2008) and Cecchetti, Kohler, and Upper (2009) for the post–World War II period. Their data include money and credit aggregates, along with various other macroeconomic variables, covering 14 developed countries from 1870–2008.15 In total, there are 79 major banking crises in their sample.
They find that the ratio of credit to money remained broadly stable between 1870 and 1930, while during the Great Depression, there was a marked deleveraging of the banking system. In the postwar period, banks first grew their loan books relative to available deposits, before sustaining high credit growth through increasing reliance on nonmonetary liabilities (debt securities and interbank borrowing). As Jordà, Schularick, and Taylor (2016c) show, much of that lending is related to property: The share of mortgage loans in banks’ total lending portfolios averaged across 17 advanced economies has roughly doubled over the past century—from about 30% in 1900 to about 60% in 2014.
Schularick and Taylor (2012) suggest that the increasing reliance on debt securities and markets to fund balance-sheet-lending growth may be a serious problem if it constitutes an unstable source of funding. That is because borrowing conditions, liquidity, and market confidence move center stage, and these may be highly unstable variables in difficult times. Their central empirical finding is that “all forms of [their empirical] model show that a credit boom over the previous five years is indicative of a heightened risk of a financial crisis.” That is, they estimate that sustained credit growth one standard deviation above the mean increases significantly the likelihood of a crisis.16 There is also in their data set some indication that changes in the rate of credit growth—that is, accelerating credit growth—may indicate imminent problems. In related work, Jordà et al. (2016c) calculate that three-quarters of all episodes during which credit to GDP rose by at least 30 percentage points over a five-year period ended in a systemic crisis.17 Figure 2 shows the relationship between credit growth and crisis probability in the expanded data from Jordà et al. (2016a). It captures their key finding rather strikingly, as strong growth in credit is reflected in sharp increases in crisis probability.
Leverage and Asset Prices
Jordà, Schularick, and Taylor (2015) build on Schularick and Taylor (2012), analyzing an annual panel of data for 17 countries since 1870, including various macro, equity, and house price data. Their interest is in the economic impact of debt-financed bubbles. If prices are above trend by more than one standard deviation, they label this a “ price elevation episode.” They also require, for a bubble to be identified, that at some point during such an episode, a large price correction occurs (“ the bubble bursts”), and real asset prices fall by more than 15% over a three-year window looking forward from any year in the episode. In the pre–World War II era, they find that financial crises were, as likely as not, to take place in association with a bubble episode in equities and/or housing (and mostly nonhousing related). In the post–World War II era, 21 of 23 financial crises are associated with a bubble episode in equities and/or housing, with 11 of 21 bubble-related financial crisis recessions linked to bubbles in both asset classes.
The authors find that post–World War II numerous equity price bubbles did not turn into financial crisis episodes while housing price bubbles, although less frequent, appeared to be more disruptive and more likely to be associated with a financial crisis episode. In addition, credit fueled asset price bubbles, especially those in housing markets after World War II, are associated with a higher likelihood of a financial crisis recession. Thus, what seems to be important is whether the price bubble relates to equities or houses, as well as whether the bubble is accompanied by rapid growth in private credit; the worst outcomes occur when the bubble is in house prices and there is a credit boom. In that case, even after five years, the economy typically has not yet quite recovered from the recession and is still struggling to regain its precrisis peak level of real GDP per capita.
More generally, Jordà, Richter, Schularick, and Taylor (2017) find that the ratio of credit to income (what they call the level of financialization) may be systematically related to business-cycle features of advanced economies: More financialized economies, they find, tend to exhibit lower real volatility, but also lower growth and more tail risk. That is, leveraged economies appear to be more at risk of steeper downturns and slower recoveries (the latter often related to financial crisis recessions). That is what Figure 3 captures; higher debt going into a recession is associated with deeper subsequent recessions and slower recoveries. The authors suggest that macroeconomists ought to reflect such features in their core models. These models, they argue, ought to reflect that real consumption and real investment exhibit a strong degree of comovement with credit in more leveraged environments.18
Good Booms, Bad Booms?
One of the issues identified by Jordà et al. (2015) and other researchers is that not all periods of boom end in bust. For example, Bordo and Meissner (2016) argue that not all banking crises are driven by credit booms, and that not all housing or equity booms, or periods with high capital inflows, end in crises. Gorton and Ordoñez (2016) argue that credit booms are not unusual historically and set out to understand why some booms end in crises (bad booms), while others do not (good booms). The key to understanding what determines whether a boom will be “good” or “bad,” they argue, is how sustained is the increase in productivity that appears to give rise to all the booms, good or bad, in their data set. Gorton and Ordoñez (2016) study 34 countries (17 advanced countries and 17 emerging market countries) over the period 1960–2010. The measure of domestic credit to the private sector that they adopt is wider than just bank credit.19 They adopt two measures of productivity: the Solow residual and labor productivity. The definition of financial crisis follows Laeven and Valencia (2012), which defines a systemic banking crisis as occurring if (a) there are “significant signs of financial distress in the banking system (as indicated by significant bank runs, losses in the banking system, and/or bank liquidations)”; and (b) if there are “significant banking policy intervention measures in response to significant losses in the banking system.”20
Gorton and Ordoñez (2016) adopt the following norm for their core analysis: A credit boom is identified whenever a country experiences three consecutive years of positive credit growth (as a fraction of GDP) averaging more than 5%. The boom ends when the country experiences at least two years of credit growth (also as a fraction of GDP) not higher than 0%. Given their country/time sample and these norms, they identify 87 booms. In their sample, 47 crises also are identified by Laeven and Valencia (2012). It turns out that 34 of those crises happened at the end of one of the 87 booms. As a result, they conclude that there are 34 bad booms and 53 good booms in the sample. On the other hand, there were 8 crises that occurred not at the end of a boom, but during a boom, and there were 5 crises that were not associated with any boom. So, while there are both good and bad booms in their data set, there are also crises that do not occur at the end of booms, and some in fact that are unrelated to booms.
Boom periods have certain statistical regularities in the Gorton and Ordoñez data. First, investment growth is significantly higher during booms than nonbooms (suggesting that investment booms typically coincide with credit booms). Similarly, credit extended both to the corporate and household sectors is higher during booms, as is real GDP growth (the latter higher in good than in bad booms). Importantly, average growth in both total factor and labor productivity is significantly higher in good booms than in bad booms—a feature consistent across both developed and developing countries. Thus, while credit growth is a predictor of financial crises in their data set, as for Schularick and Taylor (2012), the likelihood of a crisis is reduced by the occurrence of productivity growth.
The authors show that a credit boom starts with a positive innovation to productivity, but the subsequent growth trajectories differ across good and bad booms. In bad booms, productivity, real GDP, and investment growth rates tail off. In good booms, labor productivity growth is high and sustained, while in bad booms, it collapses sharply by the fourth year.21
The important lesson that seems to emerge from this literature is that banking crises appear to be linked quite strongly to prior (often secured) credit growth. That lesson has been reflected in a number of macroprudential policy innovations in some countries. However, some argue that the lessons from the historical record are not quite so clear. We turn now to sketch out some of those arguments.
Facts and Features of Financial Crises: Some Issues
Dating and Definitions
Bordo and Meissner (2016) provide an interesting comparison of leading data sets concerning the definition of crises. They argue that there are substantial discrepancies across researchers over what constitutes a crisis, and ultimately, that leads to different conclusions about the number, impact, and possibly the causes of crises. From the perspective of policymakers and model builders, those disagreements may lead to uncertainty about which facts and leading indicators to focus on. Of particular interest to Bordo and Meissner (2016) are longer-run data sets (covering more than just banking crises).22 They emphasize three in particular: Bordo et al. (2001) document banking, currency, and twin (banking plus currency) crises for all years between 1880 and 1997. For the years 1880–1945, their sample includes 21 now mostly advanced countries, and from 1945, data from 56 countries is available. Next is the celebrated work of Reinhart and Rogoff (2009), providing data on banking, currency, and sovereign debt crises for more than 70 countries, with some data going as far back as the medieval period. Finally, Taylor (2015) provides the dates for systemic financial crises (mainly banking crises) for 17 countries from 1870–2010.23 Bordo and Meissner (2016) also consider Laeven and Valencia (2008, 2012), who compile a comprehensive data set covering banking, currency, and debt crises for the period 1970–2011. Their data includes 162 advanced, emerging, and less-developed economies.
Table 1 in Bordo and Meissner (2016) gives the definitions for dating the various types of crises in each of the leading data sets: Bordo et al. (2001), Laeven and Valencia (2008, 2012), Reinhart and Rogoff (2009), and Jordà et al. (2016b). They note that
for banking crises, authors disagree about how many banks must be closed or what percentage of the financial system’s capital must be impaired for a crisis to be classified as systemic. Laeven and Valencia require that major policy interventions take place. Reinhart and Rogoff classify more crises than other authors, likely because they only require bank runs to lead to the “closing of one or more financial institutions.” (p. 374)
Data sets may differ in what constitutes a crisis, as previously noted. Related to that, they also may differ as to whether they document a crisis as a twin crisis (data sets perhaps agreeing on one crisis, but not the second one). In addition, data sets can disagree about the precise timing of a crisis—what Bordo and Meissner (2016) call “near misses.” Bordo and Meissner (2016) argue that the correlation between dating methodologies “is not extremely high even within constant country samples.” It is difficult to tell how serious the differences are for the dating specifically of banking crises (our main interest here). It appears that Taylor’s (2015) dating of banking crises is closer to that of both Laeven and Valencia (2008, 2012) and Reinhart and Rogoff (2009), rather than of Bordo et al. (2001).24
Romer and Romer (2016) also emphasized definitional issues in the dating of crises, as well as the binary nature of these and other researchers’ classification schemes. They propose an index of financial crisis that is somewhat more continuous in nature, thus moving away from what they regard as the overly simplistic “crisis/no crisis” distinction. We review their contribution in more detail next.
Output and Fiscal Costs of Financial Crises
If researchers come to different opinions about the timing and nature of financial crises, it seems likely that they also may come to different opinions about the costs of those crises. In addition to that issue, researchers use different methodologies for measuring the lost output as a consequence of a financial crisis. Typically, most authors try to define output losses as deviations from a precrisis peak in output or a precrisis output trend; some emphasize how long it takes to return to the precrisis norm; and others measure cumulative output losses over the period of the acute phase of the crisis.
Bordo et al. (2001) find that financial crises typically are associated with higher output losses than are recessions without financial crises, as do Gourinchas and Obstfeld (2012) and Jordà et al. (2013) among others. Reinhart and Rogoff (2009, 2014) find that recessions with financial crises typically are followed by slower-than-usual recoveries. A number of researchers have confirmed that finding with different data samples. More significant, perhaps, is that Jordà et al. (2011) find that output losses in financial recessions are positively associated with the size of the precrisis increase in the credit-to-GDP ratio. That suggests that the financial crisis is indeed a significant factor in causing output losses. However, it may be that selection bias in constructing databases of financial crises pairs large recessions with episodes of financial frictions in a way that exaggerates the apparent effect of crises on output losses. Alternatively, it may be that in anticipation of a large recession, agents try to reduce leverage in anticipation of future credit supply difficulties, making causality difficult to disentangle. More work in this area would be useful.
Despite numerous differences in calculating trends, crisis-dating methodologies, and samples (time periods and countries), most researchers appear to agree that financial crises typically are associated with economically significant downturns in output and output growth. Indeed, there is evidence that output losses may be even larger in the post–Bretton Woods era than the pre–World War I period. Whether that reflects that the earlier period was able to recover from crises because economies were more flexible, as Bordo and Meissner (2016) hypothesize, because the financial sector and leverage was smaller, or both remains an open question.25
On the other hand, Bordo and Haubrich (2017) compare recovery from recessions with and without crises across 22 business cycles in the United States over the period 1880–2010. They find that recessions with financial crises indeed were deeper than nonfinancial recessions, but that recoveries were stronger than those from nonfinancial recessions. Looking across the OECD economies post-1967, Romer and Romer (2016) reach a similar conclusion as to the speed of recovery.
The work of Reinhart and Rogoff (2009) has emphasized the possible impact of banking crises on the probability of a debt crisis; if sovereign debt is called into question in advanced countries, it seems likely that, wars aside, that will be due to banking crises. However, the fiscal costs of crises may be due both to the direct cost of the bailout and the indirect costs as the economy slows, tax revenues grow more slowly, and government expenditure on welfare rises. Schularick (2012) shows that the systemic crises of the late 20th century are associated with large rises in the debt-to-GDP ratio, but that, in the same sample of 14 advanced countries as Schularick and Taylor (2012), crises prior to the 1970s were not associated with significant rises in this ratio.
Laeven and Valencia (2012) analyze the rise in debt-to-GDP ratios for all the systemic banking crises in their data set. They find that the median rise in the debt-to-GDP ratio across all such crises is 12% of GDP, while in advanced economies, the figure is somewhat higher, at 21.4% of GDP. Fiscal costs, measured as the rise in outlays due to restructuring the financial sector, had a median of 6.8% of GDP. Deducting the rise in fiscal outlays from restructuring from the rise in total debt provides a simple measure of the degree of discretionary fiscal policy. The median for this variable is 7% of GDP. Furthermore, Laeven and Valencia (2012) suggest that countries with large financial sectors, large credit booms, or both also face the largest fiscal costs.
Crisis? What Crisis?
Romer and Romer (2016)26 create a new semiannual series of financial distress in 24 advanced countries for the period 1967–2012. The measure is derived from contemporaneous narrative accounts of country conditions given in the Organization for Economic Co-operation and Development (OECD) publication, OECD Economic Outlook. They classify financial distress on a scale of 0–15 rather than treating it as an 0–1 variable, with a 7 correlating, they claim, with a moderate or systemic crisis.27 Motivated by Bernanke (1983), the definition of financial distress centers not on banking crises or measures of excess credit growth, but on increases in the cost of credit intermediation identified in the OECD Economic Outlook. Overall, the Romer and Romer (2016) index appears to pick up many of the same episodes as other crisis indicators (their main comparators are Reinhart and Rogoff’s crises and Laeven and Valencia’s). That said, some episodes of crises included in alternative chronologies do not show up in their measure, and the timing of financial distress is often quite different.28 For example, Reinhart and Rogoff (2009) sometimes appear to date crises somewhat earlier than do Romer and Romer, while these three chronologies quite often come to different views as to how long-lasting a crisis was, and sometimes how severe.
Romer and Romer (2016) also study the aftermath of financial crises. They find, as do other researchers, that real GDP falls significantly and persistently. However, for the advanced countries in their 1967–2012 sample, those that fall in output following “a typical crisis” is, they argue, moderate overall. The peak decline in real GDP is approximately 6%, while the falls in industrial production and the rise in the unemployment are more modest. Moreover, their characterization of the typical aftermath of financial crises is not substantially different from that derived using existing crisis chronologies (conditional on using the same time period and sample of countries).
Romer and Romer (2016) argue that particular episodes are important outliers, showing that including the 2008 crisis and the fall in Greek national output strongly influences the results; excluding Greece from the sample lowers the estimated average output decline following a crisis by more than a percentage point. They identify 19 episodes in their sample when distress reached at least a 7 (a “moderate/systemic” crisis). They examined the path of output in the wake of these crises, concluding that even here, the evidence does not support the view that the impact of crises are exceptionally damaging. They also argue that there is little evidence of nonlinearities in the aftermaths of crises; more severe crises do not appear to have disproportionately negative aftermaths. They do find, however, that the size and persistence of their measure of financial distress help explain the variation in aftermaths; it may be that financial distress that persists at an elevated level is more likely to lead to more damaging contractions in output. And in 6 cases of extremely adverse aftermaths, including the actual behavior of financial distress explains a substantial portion of the shortfall of output from a forecast of what output would have been (based solely on output).
Conclusions From the Empirical Literature
Systemic banking crises and financial recessions appear to be relatively infrequent in advanced economies. Hence, the historical data required to build up a picture of these events covers long periods of time and diverse countries with somewhat differing financial sectors. It is not surprising, therefore, that uncertainties exist on some issues, quite apart from the dating and definitional discrepancies. For Bordo and Meissner (2016), an overemphasis on credit growth may lead policymakers to ignore other relevant indicators of impending crisis. For Romer and Romer (2016), recent crises in advanced economies just don’t seem that bad—at least most of the time.
However, the historical record also suggests that banking crises can be very costly, possibly the more so the larger the financial sector is. Broad trends in the post–World War II era suggest that the financial sector has become highly interconnected and leveraged. Moreover, it is also clear that financial regulation has not ended systemic bank crises. The impact of widespread bailouts on incentives would appear to be significant, although the historical research for the most part has not studied the impact of regulation.29 Therefore, some quantitative theorists have focused on modeling explicitly the credit intermediation process in DSGE environments and the size of the financial sector, the potential costliness of financial recessions, and the relative infrequency of crises. Less effort has been directed at understanding the quantitative significance in these models of financial regulation.30 We now turn to recent research extending the core models.
Financial Frictions in DSGE Models: New Directions
As noted previously, Bernanke and Gertler (1989), Kiyotaki and Moore (1997), and others focus on credit constraints faced by nonfinancial borrowers.31 For instance, in the Bernanke et al. (1999) model, the size of the external finance premium depends on the health of the borrower’s balance sheet. In essence, as the borrower’s interests become more aligned with the lender’s—that is, as the borrower’s own funds invested in the outcome of an investment project increase—the incentives to deviate from the interests of the lender decline. The external finance premium goes down in consequence. The upshot of this process is that a financial accelerator mechanism emerges. As wider economic conditions improve, balance sheets strengthen and the external finance premium declines. That results in a boost to the borrower’s spending, further bolstering economic activity. And the process works in reverse too; worsening economic conditions are exacerbated as the external finance premium increases and asset prices decline.
During the recent financial crisis, many commentators pointed to widespread evidence that there was disruption in the process of financial intermediation. Consistent with some of the findings of Giesecke et al. (2014), Adrian, Colla, and Shin (2012) conclude that disruptions in the supply of credit intermediated by banks and other financial intermediaries were central drivers in the recent financial crisis; there is evidence that firms that could do so resorted to direct (i.e., bond) financing.32 That disruption in credit intermediation appeared to impinge on the supply of credit to nonfinancial firms—reminiscent of the properties of the two workhorse models described earlier in this article—but also on bank leverage and interbank lending. Adrian et al. (2012, p. 207) argue:
The leverage of the banking sector emerges as being a key determinant (and reflection) of financial conditions. As such, understanding how the leverage of financial intermediaries fluctuates over the cycle emerges as perhaps the most pressing question in the study of macroeconomic fluctuations.
The Interbank Market in DSGE Models: Part I
Gertler and Kiyotaki (2010) have developed a model to begin to address these issues. In effect, these authors introduce a financial accelerator mechanism into the process of financial intermediation itself: the emphasis has shifted from how much final producers were able to borrow to how much banks were able and willing to lend.
In the model of Gertler and Kiyotaki (2010), financial intermediaries exist because they are assumed to possess some specialized skills, such as evaluating and monitoring borrowers and enforcing contracts. That assumption explains why credit flows from lenders (households) to nonfinancial borrowers (firms) via financial intermediaries. The bank takes an equity stake in the firm to which it lends (i.e., it makes a loan and absorbs the resulting risk). However, the use of the bank is not without problems. In the background, there is also assumed to be an agency problem. Specifically, bank managers face the temptation to abscond with a proportion of the bank’s assets; call that proportion . That means that households’ deposits (and deposits from other banks) may not be repaid in full. If the managers do abscond with any assets, this is observed by all, and the bank is deemed to have defaulted. The bank is then wound up and the remaining proportion of the bank’s asset, , are allotted to the depositors.
The risk of losing their funds means that depositors require some assurance that the intermediaries will be able to honor their commitments. The upshot is that lending banks need to hold sufficient own funds to counter the agency risk. That assurance is reflected in the following incentive compatibility condition:
Here, is the value of the banking firm; is the value of the bank’s loan book, being the price of the loan; and the volume is denoted by . As previously noted, is the proportion of assets that the banker is able to steal, is the volume of interbank deposits, and . The higher is, the harder it is for the bank to divert assets funded by interbank deposits. Therefore, implies frictionless interbank markets. Moreover, when the constraint is binding or expected to bind, the intermediary’s balance sheet limits its ability to obtain deposits and lend. When adverse shocks are experienced, the spread widens, which raises the cost of credit to nonfinancial borrowers. Thus, a decline in intermediary net worth induces a fall in the value of assets that the intermediary can hold, given the constraint on its leverage ratio—the latter occurring due to the principal-agent problem.
The modeling of the interbank market in Gertler and Kiyotaki (2010) is relatively simple. It is assumed that financial institutions experience idiosyncratic liquidity shocks; some institutions experience a surplus of funds, others a deficit. However, it is costly to reallocate funds from surplus to deficit institutions. That is because first, the agency problem constraining an intermediary’s ability to obtain funds from depositors also may constrain its ability to obtain funds from other financial institutions. Second, nonfinancial firms may be able to raise deposits from only a subset of financial intermediaries. The upshot is that such frictions in the interbank markets distort real activity, relative to the no-frictions benchmark case.
Thus, the friction between depositors and banks is exacerbated when the interbank market is compromised. Moreover, there is no friction distorting the interaction between banks and nonfinancial debtors, as in the two core (precrisis) workhorse models.
A “ crisis” in this setup is sparked by an exogenous decline in the “quality of capital”; that is, a decline in the sequence of dividend payouts expected to be remitted to the bank. Intuitively, the nonfinancial companies in which banks own equity stakes experience a persistent decline in their productive potential, and hence their dividend payouts. This causes the bank, due to its leveraged position, to have to reduce its borrowing by proportionally more than the initial drop in net worth. In doing so, this so-called fire sale further reduces the value of nonfinancial equity, tightening still further the bank’s borrowing constraint. This decline in asset values depresses real investment. How damaging the ensuing downturn is depends on the size of the negative shock, the efficiency of the interbank markets, and government intervention.
If interbank markets are frictionless, bankers cannot divert assets funded by interbank borrowing and the impact of a quality of capital shock is more limited; the incentive compatibility constraint tightens less than it otherwise would. The contraction in the capital stock drives up the return to capital, encouraging a return to investment activities. In their baseline model, Gertler and Kiyotaki (2010) show that compared to a real business cycle model benchmark, the model with no interbank financial frictions experiences a deeper contraction in aggregate economic activity and endures a slower recovery following the initial negative shock. On the other hand, with frictions in the interbank market, the contraction in the economy is even more severe. The efficiency of the interbank market in this setup works to ameliorate the severity of the contraction in economic activity; the interbank market, if efficient, is in effect a partial solution to the underlying agency problem faced by households when placing their deposits with banks.
However, if banks are no better than households in monitoring bank activity (i.e., the agency problem faced by households is as severe for banks in their lending), then the downturn in economic activity consequent on a quality of capital shock to nonfinancial borrowers is more pronounced. The leverage and fire-sale effects just noted are somewhat more severe under inefficient interbank markets, as the spread between the expected return on equity and the risk-free rate rises further (as a result of the decline in bank net worth). Regardless of how efficient the interbank market is, it takes time for banks in the model to rebuild their capital base (net worth)—the model’s version of deleveraging—during which time the spread between the expected return on equity and the risk-free rate (the credit spread) remains elevated. And that spread is wider the less efficient the interbank market is.
Gertler and Kiyotaki (2010) also examine whether direct lending by the central bank may alleviate financial recessions in their model. They assume a simple rule for the central bank, whereby private credit is intermediated by the central bank in proportion to the rise in the credit spread; the higher the credit spread relative to its long-run (i.e., steady-state) value, the more the central bank lends to private borrowers. This policy significantly reduces the rise in the spread, which in turn ameliorates the drop in investment. The overall decline in output is reduced significantly.
Again, there appears to be a significant difference between how important this intervention is with frictionless interbank markets compared to the case with frictions. When interbank markets are compromised, and the central bank is able to identify good lending opportunities, central bank intervention is even more effective.
Gertler and Kiyotaki (2010) is an important contribution, as it maps out a way to model financial intermediaries in a quantitative, mainstream DSGE environment that is usable in principle for policy purposes. Moreover, in introducing an interbank market, it moves the DSGE model environment in a realistic direction, given the increasing importance, and possible fragility, of interbank funding.33 However, there are some shortcomings. For example, there is no role for outside equity for banks. The only funds that banks in the model obtain from investors are one-period deposits, while the own funds are underpinned by the incentive compatibility constraint. In practice, retail deposits are typically covered by deposit protection schemes and capital is regulated by prudential authorities. In addition, some firms, especially larger firms, are able to substitute direct financing for bank loans, as noted earlier. Finally, the Gertler and Kiyotaki (2010) model is analyzed in approximate linear form, with the incentive compatibility constraint always binding, which misses key aspects of crisis dynamics. We turn now to that latter issue.
Financial Frictions and Financial Crises
For some researchers, the sharp contraction and slow recovery in some countries following the recent crisis are important features to be explained. Brunnermeier and Sannikov (2014) develop a stochastic, continuous-time model to study full equilibrium dynamics (as opposed to local analysis based on linear approximations).
In their model, they distinguish between so-called experts, who are more efficient than households at turning capital into output. Experts, however, are limited in how much equity they may issue—a reflection of an underlying agency problem—but in principle, they are not constrained in the issuing of debt. Their model features occasionally binding borrowing constraints, which are the central element in producing highly nonlinear dynamics of a sort that is intended to reflect crisis periods.34 Indeed, the effects are asymmetric, arising only in downturns. Broadly, in times of relative stability, the borrowing constraint is not binding and the economy is stable, not deviating much from its long-run (i.e., stochastic) steady state. However, following a large enough negative disturbance (or a sequence of smaller, negative disturbances), the economy may shift into a region where the constraint is binding, amplifying the effect of the shock on the downturn.
In the aggregate, the so-called experts act in a risk-averse manner: they anticipate possible adverse shocks and optimally determine a level of net worth that can handle a range of shocks while still meeting their debt obligations. Thus, in supposedly normal times, near the stochastic steady state, amplification effects are subdued. However, when faced with rare, large, negative shocks, they delever. As in Gertler, Kiyotaki, and Prestipino (2016) (discussed next), during a period of deleveraging, capital is used less efficiently (i.e., aggregate production falls), asset prices are negatively affected and volatile, and financial amplification is substantial and long-lasting. Brunnermeier and Sannikov (2014) emphasize that this process leads to high levels of asset price volatility and wider economic volatility due to endogenous risk formation. In addition to being able to study the global solution of the model, the stochastic continuous time framework connects with the asset-pricing literature.35
Unlike in Gertler and Kiyotaki (2010), there is no role for an interbank market in the Brunnermeier and Sannikov (2014) model. In the recent crisis, many analysts point to that market as being central to the worsening of the crisis. We now examine new models that can combine nonlinear dynamics, as in Brunnermeier and Sannikov (2014), along with simple models of the interbank market.
The Interbank Market in DSGE Models: Part II
This section focuses on two recent contributions introducing a nontrivial interbank market into a DSGE environment. The first is that of Boissay, Collard, and Smets (2016). The role of the interbank market, which is subject to a “lemons” problem, is central to their model. The explicit motivations behind their model are the infrequent nature of financial crises, the interbank market, and the “credit booms gone bust” characterization of banking/credit crises of Schularick and Taylor (2012) reviewed previously.
In the model of Boissay et al. (2016), banks are heterogeneous in how efficient they are at intermediation. One may think that there are costs to finding good loan opportunities, with some banks more proficient than others and the better banks incurring lower search costs. Banks, as in Gertler and Kiyotaki (2010), have an incentive to divert borrowed funds, and that incentive limits how much they can borrow. Importantly, a bank’s type is also private information. If information on type were publicly known, with no diversion risk, it would be optimal to channel all lending via the most efficient bank. When the bank type is private information, that is not possible; lenders cannot factor bank type into their assessment as to how likely a bank is to divert resources. Therefore, the interbank loan contract is the same for all banks. Let be the gross interbank loan rate and be the gross return on diverted funds. A bank may divert its own funds and any borrowed funds, , although such diversion is costly. Let reflect that cost such that, per unit of deposit, the requirement for banks not to divert funds is that the gross interbank rate is high enough:
This is an incentive compatibility constraint, and it potentially limits the amount of interbank lending that can take place in equilibrium. Notice that as the gross interest () rate goes down, so too might the amount of interbank borrowing ().
This model features a more-or-less familiar financial accelerator effect. However, on top of that, and similar to the model of Brunnermeier and Sannikov (2014), it features highly nonlinear dynamics as periods of strong economic growth give birth to strong credit growth and subsequent slumps as the economy deleverages. That is, as the economy grows and productivity rises, the banking system expands credit to final borrowers. The more efficient banks naturally expand their firm lending activities by borrowing from less efficient banks. However, as the economy slows (as the underlying elevated level of productivity comes to an end), agents in both the household sector and the corporate sector naturally respond. Households increase their savings to smooth consumption and firms borrow less anticipating lower future demand. This rise in net savings depresses economywide interest rates, including in the interbank market. As the interbank rate falls, less efficient banks are more tempted to borrow and divert funds. And since bank type is private information, the interbank market becomes very risky and interbank lending declines. This reflects the so-called lemons problem.36 Boissay et al. (2016) show that there is a threshold level of interest rates, below which the interbank market freezes entirely. The result is a credit crunch and a very deep recession. In this way, the interbank market plays a critical role in both the credit boom and the credit bust.
In this model, all crises would appear to follow “good” booms, in the terminology of Gorton and Ordoñez (2016). Indeed, all good booms run the risk of precipitating a crisis. The interbank market, as well as its lack of transparency, is the ultimate source of any serious instability. There is no role in the model for bank equity or secured lending. And since bank type is not verifiable, the model’s implications for financial regulation are unclear. Nevertheless, the model is a significant step forward in modeling the role of credit intermediation.
The second contribution that we focus on here is that of Gertler et al. (2016). Building on the earlier contribution of Gertler and Kiyotaki (2010), the authors suggest that it is not the interbank market as such that is the problem. They argue that a combination of financial innovation and regulatory constraints on traditional banks has created a network of financial intermediaries that has increased the equilibrium level of leverage in the economy. Specifically, financial innovation has created specialist institutions with comparative advantages over traditional banks in certain forms of loan origination, securitization, and funding. Given these skills, and the distortive impact of capital and other regulation on traditional banks (although none of these factors are modeled explicitly), an increasing proportion of lending has migrated to the shadow banking/wholesale sector. And it was this sector that played a key role in the recent financial crisis. Gertler et al. (2016) argue that the freezing of the interbank market was a defining characteristic of the crisis since, as in Boissay et al. (2016) and Brunnermeier and Sannikov (2014), it results in the economy operating in a highly inefficient way, unable to direct capital to its most efficient uses.
The model builds on the frameworks of Gertler and Kiyotaki (2010) and Gertler and Kiyotaki (2015).37 Private households may invest directly in nonfinancial firms or in retail or wholesale banks. It is costly to make such investments for all agents, but regulations applied to retail banks and not to wholesale banks provide the latter with a competitive advantage at the margin. Similarly, retail banks have expertise not available to households, giving the latter an advantage over the former at the margin. The wholesale banks are set out such that they fund themselves optimally largely via the interbank market (i.e., they hold few retail deposits). Consequently, the size of the wholesale banking market arises endogenously depending on two factors. The first is the advantage that wholesale banks have over retail banks in managing assets; and the second is the advantage of retail banks over households in overcoming an agency friction that impedes lending to wholesale banks. We first describe this model setting to one side the issue of bank runs, to which we return later.
The model is an endowment economy with a fixed amount of nondepreciating capital and a single nondurable good. There are three classes of agents: households, retail banks, and wholesale banks. In principle, all agents may hold capital directly and invest their wealth in each of the other agents. The capital held is used to produce the nondurable good and to be carried over into the next period to be used in goods production. The quantity of goods produced is stochastic, as it depends on an aggregate technology productivity shock. The productivity of capital also will depend on which agent operates that capital.
Households consume and save. Their savings are in the form of bank deposits and direct capital holdings. If they hold capital directly, they need to pay an operating or absorption cost; it is costly for agents to hold capital directly, and increasingly so at the margin. In the absence of bank runs, bank deposits are one-period debt instruments, paying a noncontingent return. Each period, households receive income from their capital holdings, their bank deposit portfolio, and they receive an endowment (attended by the aforementioned productivity shock).
As noted previously, there are two types of banks: retail and wholesale. They are identical but for their ability to absorb capital at the margin; wholesale banks face a zero marginal cost, while retail banks face a positive cost (that is nevertheless lower than households’ cost). Both types of bank may raise funding from households and other banks, and both may hold capital directly. Retail banks combine their own funds (retained earnings plus an initial endowment; there is no outside equity) with deposits from households and other banks.
As in Gertler and Kiyotaki (2010), there is an assumed moral hazard problem that limits banks’ ability to raise funds. The banker may divert a fraction, (), of nonfinancial loans funded via retail deposit or own funds. However, the banker is able to divert only () proportion of loans funded by interbank deposits. As before, bankers are assumed to possess superior monitoring skills compared with households. There are also retail and wholesale funding markets. On the other hand, if the bank lends to other banks, the assumption is that such loans are easier to monitor and, as such, more difficult to divert. Given the incentives to divert and the ability and constraints on the lenders to the banks, the banks will choose to act honestly only if the following incentive constraints are respected:
Here, the value of the banking firm (, for retail banks, for wholesale banks) is denoted by . The nonfinancial loan book is funded in the amount by retail deposits and the bank’s own net worth, and by in interbank-borrowed funds. The first constraint is relevant to banks who borrow on the interbank market and the second to those who lend (in which case, ). denotes the price of capital, and is the marginal cost of managing the loan book. For retail banks, , while for wholesale banks, Finally, , such that reflects an assumption that bankers can divert only a fraction of its loans to other banks. Thus, while low monitoring costs () may encourage expansion of the wholesale banking sector at the expense of the retail sector, retail banks have a superior ability over households to monitor wholesale banks, who also face incentives to behave dishonestly. By the appropriate choice of and , Gertler et al. (2016) are able to generate what they argue is an empirically plausible amount of interbank lending, while still having retail banks making nonfinancial loans.38 In fact, they are able to analyze equilibria wherein wholesale banks raise no funds from households, instead relying on own funds and interbank loans and where retail banks are able to increase their leverage by making interbank loans.39
Figure 4 is a visual characterization of these new models, which attempts to relate them to the earlier core models shown in Figure 1. The idea is to show that interbank lending may complicate the process of deleveraging when that market also suffers from frictions, as in Boissay et al. (2016) and Gertler et al. (2016). In the top section, the banking sector acts to stabilize the economy; it channels funds toward profitable opportunities. However, when there are financial frictions (the lower section), the fall in prices tightens constraints that banks face. That tightening can exacerbate the initial shock, and jointly banks’ creditworthiness suffers, affecting the wider economy.
Equilibrium With Bank Runs
The possibility of nonlinear dynamics in the Gertler et al. (2016) model comes from the possibility of bank runs (either anticipated or unanticipated).40 The condition for a bank run is when the wholesale sector may not be able to repay its interbank creditors—the more likely the higher wholesale bank leverage is (and thus the lower the resale value of capital), the higher the interbank borrowing rate is and the lower aggregate productivity is. If there is a run on the wholesale banking sector, there is in effect a fire sale; wholesale banks liquidate their portfolio, which has to be absorbed by retail banks and private agents, both of whom are less efficient at managing capital, as in Brunnermeier and Sannikov (2014). The wholesale sector then has to rebuild itself slowly as new banks enter the market.
The economy thus finds it difficult to reequilibrate following a negative shock to productivity, as retail banks and households face costs absorbing additional capital. These assumed costs make it increasingly costly at the margin for retail banks and households to absorb capital directly. In this way, agents with wealth but lack of expertise are constrained in their purchases of assets during a resale.
The lower the agency friction, , between retail banks and wholesale banks, the larger equilibrium leverage is. On the one hand, the lower that is [and the higher the financial innovation, as Gertler et al. (2016) interpret it], the more stable is the economy, other things constant. That is because retail banks are in a better position to absorb the asset sales of wholesale banks, stabilizing asset prices and reducing the financial accelerator effects. On the other hand, the lower that is, the more likely are bank runs; financial efficiency entails higher equilibrium leverage and asset liquidation values are lower than they would be otherwise.
The financial accelerator mechanism is present here, as in Gertler and Kiyotaki (2010), although the wholesale sector plays an especially significant part. Leverage amplifies the impact of the drop in aggregate productivity on bankers’ net worth, resulting in a tightening of financial constraints as credit spreads increase. As a result, wholesale banks sell off loans, pushing down asset prices and bank net worth. The higher wholesale bank leverage is, the stronger this feedback typically is, forcing large-scale liquidations of their assets and reducing their demand for interbank loans. As a result of this fire sale, retail bankers increase their asset holdings and absorb, along with households, the capital that the wholesale banks place on the market. However, as retail banks and households are less efficient in intermediation, the fire sale is costly; the cost of bank credit to nonfinancial borrowers rises and the drop in aggregate output is amplified.
As with Boissay et al. (2016), Gertler et al. (2016) present an innovative way to extend the core DSGE framework to incorporate credit intermediation. However, the policy implications of these models are as yet tentative. On the one hand, it remains the case that banks in these models have no equity capital—a major lever that, in practice, regulators may adjust to address the robustness of the banking sector. Also, the moral hazard due to bailouts and other banking support operations are difficult to assess. For example, Gertler et al. (2016) emphasize the excessive costs on the economy of wholesale banks liquidating their assets, which is especially costly in the presence of a bank run. They argue that the anticipation of ex post intervention by the central bank may help stabilize asset prices and reduce the risk of a bank run. Indeed, leverage actually could decrease as anticipated intervention puts a floor under asset prices. However, the moral hazard implications of such actions, actual or anticipated, are unclear, as it would seem that anticipated bailouts run the risk of increasing leverage and the frequency of bailouts. This is not studied in their model. Kareken and Wallace (1978) present a discussion of the potential risks.
Financial Frictions: International Dimensions
The notion that financial frictions may bind only periodically has been an important theme in the international macroeconomics literature, particularly in the context of developing countries. Financial frictions are also prevalent in the international macro literature, which focuses on sudden stops to capital inflows. Mendoza (2010) studies a small, open economy with an exogenous interest rate and price of foreign input goods. There exists a collateral constraint limiting the value of outstanding intertemporal debt and intratemporal loans to cover working capital requirements (e.g., labor and intermediate inputs). Mendoza (2010) shows that for his model of an emerging economy, the collateral constraint is only occasionally binding. Similar to Brunnermeier and Sannikov (2014), following shocks, it has little impact on the economy’s dynamics most of the time. In part, the constraints bind infrequently because of precautionary savings behavior (as agents seek to avoid the effects of precipitous decreases in consumption). Leverage buildups are preceded by a period of strong growth. When the leverage constraint binds, there is a cessation of lending to the economy (a sudden stop) and a sharp contraction in the economy as the cost of borrowing rises and fire sales of capital further tighten the collateral constraint. The buildup of risk in the economy is, in a sense, endogenous. That is, a crisis may be sparked by shocks that, at lower leverage levels, would be absorbed by agents rather than by a sharp contraction in activity at a higher level of leverage. The role of financial frictions in the developing economy context is surveyed in more detail in Uribe and Schmitt-Grohé (2017).
Directions for Future Research
The literature on financial frictions remains an active and important area of research. There has been a great deal of progress since the financial crisis in understanding the properties of both that crisis and earlier ones, as well as in how to go about modeling crises in a DSGE environment.
However, there remains much that needs to be better understood around the historical record and the materiality of any inconsistencies across data sets and studies. Furthermore, although much progress has been made in incorporating financial intermediation into workhorse DSGE models, some important features of the landscape are missing, making policy prescriptions based on these models challenging.
For instance, an important omission in all models of banks and financial intermediaries in DSGE models is a theory on how easy it is for classes of final borrowers to switch sources of external finance during difficult economic times. The models reviewed in this paper define a crisis as a period when the economy finds it hard to reallocate capital across agents and operate it efficiently. That behavior, however plausible, is essentially hardwired into these models. Clarifying the microfoundations of these difficulties seems important.
The emerging models of the banking sector in DSGE environments have not yet incorporated the outside equity of banks. The level of bank capital has been at the forefront of supervisory concern for many decades, and now capital regulation has come under the purview of so-called macroprudential policymakers.41 The efficiency and welfare implications of such policies, including the optimal form of these regulations, are as yet far from clear when viewed from the perspective of DSGE models.42 Gertler et al. (2016) argue that the shadow banking sector is, in part, a response to other financial regulations. Understanding the interplay between financial intermediaries’ corporate structures also appears to be an underexplored area.43
Finally, building and econometrically estimating medium-scale, nonlinear DSGE models with financial intermediaries and incorporating the possibility of crisis events (e.g., bank runs and collapses in the interbank market) are desirable goals for applied research. As things stand, the technical challenges appear substantial, and such models may be some way off.
Adrian, T., Colla, P., & Shin, H. S. (2012). Which financial frictions? Parsing the evidence from the financial crisis of 2007 to 2009. NBER Macroeconomics Annual, 27, 159–214.Find this resource:
Akerlof, G. A. (1970). The market for “lemons”: Quality uncertainty and the market mechanism. Quarterly Journal of Economics, 84(3), 488–500.Find this resource:
Allen, F., & Gale, D. (2007). Understanding financial crises: The Clarendon Lectures. Oxford: Oxford University Press.Find this resource:
Azariadis, C., Kaas, L., & Wen, Y. (2016). Self-fulfilling credit cycles. Review of Economic Studies, 83(4), 1364–1405.Find this resource:
Bernanke, B. (1983). Nonmonetary effects of the financial crisis in the propagation of the Great Depression. American Economic Review, 73(1), 257–276.Find this resource:
Bernanke, B., & Gertler, M. (1989). Agency costs, net worth, and business fluctuations. American Economic Review, 79(1), 14–31.Find this resource:
Bernanke, B., Gertler, M., & Gilchrist, S. (1999). The financial accelerator in a quantitative business cycle framework. In J. Taylor & M. Woodford (Eds.), Handbook of macroeconomics (Vol. 1, pp. 1341–1393). Amsterdam: Elsevier.Find this resource:
Bhattacharya, S., Boot, A., & Thakor, A. (Eds.). (2004). Credit, intermediation, and the macroeconomy. Oxford: Oxford University Press.Find this resource:
Boissay, F., Collard, F., & Smets, F. (2016). Booms and banking crises. Journal of Political Economy, 124(2), 489–538.Find this resource:
Bordo, M. D., Eichengreen, B., Klingebiel, D., & Martinez-Peria, M. S. (2001). Is the crisis problem growing more severe? Economic Policy: A European Forum, 16(32), 51–75.Find this resource:
Bordo, M. D., & Haubrich, J. G. (2017). Deep Recessions, Fast Recoveries, and Financial Crises: Evidence from the American Record. Economic Inquiry, 55(1), 527–541.Find this resource:
Bordo, M. D., & Meissner, C. M. (2016). Fiscal and financial crises. In J. B. Taylor & H. Uhlig (Eds.), Handbook of macroeconomics (Vol. 2, pp. 355–412). The Netherlands: Elsevier.Find this resource:
Brunnermeier, M. K., & Sannikov, Y. (2014). A macroeconomic model with a financial sector. American Economic Review, 104(2), 379–421.Find this resource:
Calomiris, C., & Kahn, C. (1991). The role of demandable debt in structuring optimal banking arrangements. American Economic Review, 93(5), 1615–1646.Find this resource:
Carlstrom, C., & Fuerst, T. (1997). Agency costs, net worth, and business fluctuations: A computable general equilibrium analysis. American Economic Review, 87(5), 893–910.Find this resource:
Carlstrom, C. T., Fuerst, T. S., & Paustian, M. (2016). Optimal contracts, aggregate risk, and the financial accelerator. American Economic Journal: Macroeconomics, 8(1), 119–147.Find this resource:
Cecchetti, S. G., Kohler, M., & Upper, C. (2009). Financial crises and economic activity. NBER Working Paper no. 15379.Find this resource:
Christiano, L. J., Motto, R., & Rostagno, M. (2014). Risk shocks. American Economic Review, 104(1), 27–65.Find this resource:
Damjanovic, T., Damjanovic, V., & Nolan, C. (2017). Default, bailouts, and the vertical structure of financial intermediaries. University of Glasgow. Working Paper.Find this resource:
Diamond, D., & Dybvig, P. (1983). Bank runs, deposit insurance, and liquidity. Journal of Political Economy, 91(3), 401–419.Find this resource:
Di Tella, S. (2017). Uncertainty shocks and balance sheet recessions. Journal of Political Economy, 125(6), 2038–2081.Find this resource:
Duncan, A., & Nolan, C. (2015). Objectives and challenges of macroprudential policy. University of Glasgow Economics Discussion Paper 2015–22.Find this resource:
Duncan, A., & Nolan, C. (2017a). Disputes, debt, and equity. University of Glasgow Economics Discussion Paper 2017–08.Find this resource:
Duncan, A., & Nolan, C. (2017b). Financial macroeconomics with complete business cycle risk markets. Working Paper.Find this resource:
Gertler, M., & Kiyotaki, N. (2010). Financial intermediation and credit policy in business cycle analysis. In B. M. Friedman & M. Woodford (Eds.), Handbook of monetary economics (pp. 547–599). North Holland: Elsevier.Find this resource:
Gertler, M., & Kiyotaki, N. (2015). Banking, liquidity, and bank runs in an infinite horizon economy. American Economic Review, 105(7), 2011–2043.Find this resource:
Gertler, M., Kiyotaki, N., & Prestipino, A. (2016). Wholesale banking and bank runs in macroeconomic modeling of financial crises. In J. B. Taylor & H. Uhlig (Eds.), Handbook of macroeconomics (Vol. 2, pp. 1345–1425). The Netherlands: Elsevier.Find this resource:
Gertler, M., Kiyotaki, N., & Prestipino, A. (2017). A macroeconomic model with financial panics. Working Paper.Find this resource:
Giesecke, K., Longstaff, F. A., Schaefer, S., & Strebulaev, I. A. (2014). Macroeconomic effects of corporate default crises: A long-term perspective. Journal of Financial Economics, 111(2), 297–310.Find this resource:
Gilchrist, S., & Leahy, J. V. (2002). Monetary policy and asset prices. Journal of Monetary Economics, 49(2002), 75–97.Find this resource:
Gilchrist, S., Yankov, V., & Zakrajšek, E. (2009). Credit market shocks and economic fluctuations: Evidence from corporate bond and stock markets. Journal of Monetary Economics, 56(4), 471–493.Find this resource:
Gorton, G., & Metrick, A. (2012). Securitized banking and the run on repo. Journal of Financial Economics, 104(3), 425–451.Find this resource:
Gorton, G., & Ordoñez, G. (2014). Collateral crises. American Economic Review, 104(2), 343–378.Find this resource:
Gorton, G., & Ordoñez, G. (2016). Good booms, bad booms. NBER Working Paper no. w22008. Updated 2017.Find this resource:
Gourinchas, P. O., & Obstfeld, M. (2012). Stories of the twentieth century for the twenty-first. American Economic Journal: Macroeconomics, 4(1), 226–265.Find this resource:
Guerrieri, V., & Uhlig, H. (2016). Housing and credit markets: Booms and busts. In J. B. Taylor & H. Uhlig (Eds.), Handbook of macroeconomics (pp. 1427–1496). The Netherlands: Elsevier.Find this resource:
Hall, R. E. (2010). Why does the economy fall to pieces after a financial crisis? Journal of Economic Perspectives, 24(4), 3–20.Find this resource:
Hart, O., & Moore, J. H. (1994). A theory of debt based on the inalienability of human capital. Quarterly Journal of Economics, 109(4), 841–879.Find this resource:
He, Z., & Krishnamurthy, A. (2012). A model of capital and crises. Review of Economic Studies, 79(2), 735–777.Find this resource:
He, Z., & Krishnamurthy, A. (2014). A macroeconomic framework for quantifying systemic risk. Working Paper.Find this resource:
Holmström, B., & Tirole, J. (2011). Inside and outside liquidity. Cambridge, MA: MIT Press.Find this resource:
Iacoviello, M. (2005). House prices, borrowing constraints, and monetary policy in the business cycle. American Economic Review, 95(3), 739–764.Find this resource:
Iacoviello, M., & Neri, S. (2010). Housing market spillovers: Evidence from an estimated DSGE model. American Economic Journal: Macroeconomics, 2(2), 125–164.Find this resource:
Ivashina, V., & Scharfstein, D. (2010). Bank lending during the financial crisis of 2008. Journal of Financial Economics, 97(2010), 319–338.Find this resource:
Jordà, Ò., Richter, B., Schularick, M., & Taylor, A. M. (2016a). Macrofinancial history and the new business cycle facts. NBER Macroeconomics Annual, 31, 213–263.Find this resource:
Jordà, Ò., Richter, B., Schularick, M., & Taylor, A. M. (2017). Bank capital redux: Solvency, liquidity, and crisis. NBER Working Paper No. 23287.Find this resource:
Jordà, Ò., Schularick, M., & Taylor, A. M. (2011). Financial crises, credit booms, and external imbalances: 140 years of lessons. IMF Economic Review, 59(2), 340–378.Find this resource:
Jordà, Ò., Schularick, M., & Taylor, A. M. (2015). Leveraged bubbles. Journal of Monetary Economics, 76(Suppl.), S1–S20.Find this resource:
Jordà, Ò., Schularick, M., & Taylor, A. M. (2013). When credit bites back. Journal of Money, Credit and Banking, 45, 3–28.Find this resource:
Jordà, Ò., Schularick, M., & Taylor, A. M. (2016b). Sovereigns versus banks: Credit, crises, and consequences. Journal of the European Economic Association, 14(1), 45–79.Find this resource:
Jordà, Ò., Schularick, M., & Taylor, A. M. (2016c). The great mortgaging: housing finance, crises and business cycles. Economic Policy, 31(85), 107–152.Find this resource:
Kareken, J. H., & Wallace, N. (1978). Deposit insurance and bank regulation: A partial-equilibrium exposition. Journal of Business, 51(3), 413–438.Find this resource:
Kiyotaki, N., & Moore, J. (1997). Credit cycles. Journal of Political Economy, 105(2), 211–248.Find this resource:
Krasa, S., & Villamil, A. P. (2000). Optimal contracts when enforcement is a decision variable. Econometrica, 68(1), 119134.Find this resource:
Krishnamurthy, A. (2003). Collateral constraints and the amplification mechanism. Journal of Economic Theory, 111(2), 277–292.Find this resource:
Laeven, L., & Valencia, F. (2008). Systemic banking crises: A new database. IMF Working Paper No. 08/224.Find this resource:
Laeven, L., & Valencia, F. (2012). Systemic banking crises database: An update. IMF Working Paper No. WP/12/163.Find this resource:
Meeks, R. (2017, June). Capital regulation and the macroeconomy: Empirical evidence and macroprudential policy. European Economic Review, 95, 125–141.Find this resource:
Mendoza, E. G. (2010). Sudden stops, financial crises, and leverage. American Economic Review, 100, 1941–1966.Find this resource:
Mian, A., & Sufi, A. (2016). Who bears the cost of recessions? The role of house prices and household debt. Handbook of Macroeconomics, 2, 255–296.Find this resource:
Minsky, H. P. (1986). Stabilizing the unstable economy. New York: McGraw-Hill.Find this resource:
Nikolov, K. (2014, August). Collateral amplification under complete markets. Journal of Economic Dynamics and Control, 45, 80–93.Find this resource:
Nolan, C., & Thoenissen, C. (2009). Financial shocks and the US business cycle. Journal of Monetary Economics, 56(4), 596–604.Find this resource:
Quadrini, V. (2011). Financial frictions in macroeconomic fluctuations. Federal Reserve Bank of Richmond Economic Quarterly, 97(3), 209–254.Find this resource:
Reinhart, C. M., & Rogoff, K. S. (2008). This time is different: A panoramic view of eight centuries of financial crises. NBER Working Paper No. 13882.Find this resource:
Reinhart, C. M., & Rogoff, K. S. (2009). This time is different: Eight centuries of financial folly. Princeton, NJ: Princeton University Press.Find this resource:
Reinhart, C. M., & Rogoff, K. S. (2014). Recovery from Financial Crises: Evidence from 100 Episodes. NBER Working Paper No. 19823.Find this resource:
Romer, C. D., & Romer, D. H. (2016). New evidence on the aftermath of financial crises in advanced countries. Berkeley: University of California.Find this resource:
Schularick, M. (2012). Public debt and financial crises in the twentieth century, No 2012/1, Discussion Papers, Free University Berlin, School of Business & Economics.Find this resource:
Schularick, M., & Taylor, A. M. (2012). Credit booms gone bust: Monetary policy, leverage cycles, and financial crises, 1870–2008. American Economic Review, 102(2), 1029–1061.Find this resource:
Taylor, A. M. (2015). Credit, stability, and the macroeconomy. Annual Review of Economics, 7(1), 309–339.Find this resource:
Townsend, R. M. (1979). Optimal contracts and competitive markets with costly state verification. Journal of Economic Theory, 21(2), 265–293.Find this resource:
Uribe, M., & Schmitt-Grohé, S. (2017). Open economy macroeconomics. Princeton, NJ: Princeton University Press.Find this resource:
(1.) The empirical literature on financial crises has focused on banking, currency, and sovereign debt crises, as well as crises combining one or more of these elements. The focus in this paper is mainly on banking crises, specifically those related to periods of robust credit growth.
(2.) Of course, many other models exist that addressed financial crises and frictions. The models reviewed in this discussion became the workhorse models because they could be incorporated into a core neoclassical growth and business-cycle framework. There is also a vast body of literature analyzing financial markets and institutions from partial equilibrium perspectives. See, for example, many of the papers in Bhattacharya et al. (2004) and Allen and Gale (2007); also relevant is Minsky (1986). Many of the key ideas in these other approaches, however, are captured in recent extensions to the core models reviewed. For example, the idea of credit cycles, emphasized by Minsky (1986), is captured in models by Azariadis et al. (2016) and Boissay et al. (2016).
(3.) Duncan and Nolan (2017a) point out that debt contracts are in fact not the optimal contract in standard costly state verification environments. They incorporate audit errors into the model and show that, in that case, debt contracts often emerge as optimal. They pursue the macroeconomic implications of those insights in Duncan and Nolan (2017b).
(8.) For example, see Christiano, Motto, and Rostagno (2014), which explores the role of so-called risk shocks. They extend the Bernanke et al. (1999) model, such that borrowing entrepreneurs face idiosyncratic productivity, the cross-section dispersion of which may widen stochastically. They find that fluctuations in risk are the most important shock driving the business cycle.
(9.) Hall (2010) shows that his variant of the workhorse models does a reasonable job of capturing the initial fall in economic activity following the recent financial crisis, but that its dynamic properties are less impressive. This appears to reflect, as noted in the text, the need for a time-varying financial friction.
(10.) See the analysis in Duncan and Nolan (2017b). For a similar result to Carlstrom et al. (2016) vis-à-vis the Kiyotaki and Moore (1997) model, see Krishnamurthy (2003) and Nikolov (2014). The essence of these results flows from the fact that aggregate shocks are publicly observable. That makes insurance contracts contingent on those risks that are possible to write and enforce. See Di Tella (2017) for an alternative perspective.
(11.) Many of the key empirical and historical contributions to the study of financial crises also have covered currency and sovereign debt crises in addition to banking crises, as well as crisis events characterized by more than one type of crisis. See Bordo et al. (2001), Reinhart and Rogoff (2009), and Laeven and Valencia (2008, 2012).
(12.) That specialness sometimes is referred to as the bank or credit channel, in the context of the transmission mechanism of monetary policy.
(13.) A crisis is taken to be a run of consecutive years when the default rate exceeds 2.5% (five times the median default rate). During the period 1871–1879, more than 50% of all outstanding bonds defaulted; this period is associated with railroad expansion. In total, there are 13 corporate default crises in their data.
(14.) Notably, over the period 1933–1940, the outstanding stock of corporate bonds is larger than bank loans.
(15.) The countries included are Australia, Canada, Denmark, France, Germany, Italy, Japan, Netherlands, Norway, Spain, Sweden, Switzerland, the United Kingdom, and the United States. They (and coauthors) have expanded this data set since then, in terms both of countries included (to 17, adding Belgium, Finland, and Portugal) and many more variables. See Jordà, Schularick, and Taylor (2016b).
(16.) See also the results of Gourinchas and Obstfeld (2012), which argues that over the period 1973–2010, their data suggest that regardless of whether a country is emerging or advanced, domestic credit expansion and real currency appreciation have been the most robust and significant predictors of financial crises.
(17.) Jordà et al. (2016b) analyze total (public plus private) debt levels in the same data set. They confirm that it is private credit that is linked to ensuing crisis episodes. Thus, historically in advanced economies, financial crises do not appear to be caused primarily by public debt problems. However, if a financial crisis strikes and fiscal capacity is low, there is evidence that the ensuing recession is more costly.
(18.) Of less interest for the present purposes, but nevertheless of some importance, they also note that advanced economies have become more synchronized, perhaps lessening the ability to hedge financial risk internationally.
(19.) Credit is defined as financial resources provided to the private sector, including loans, acquisitions of nonequity securities, trade credit, and other receivables. It thus appears to be a mix of secured and unsecured credit.
(20.) Auxiliary criteria employed by Laeven and Valencia (2012) are (a) extensive liquidity support (when central bank claims on the financial sector to deposits exceeds 5% and more than double relative to the precrisis level); (b) bank restructuring gross costs are at least 3% of GDP; (c) significant bank nationalizations; (d) significant guarantees are put in place; (e) there are significant asset purchases (at least 5% of GDP); and (f) there are deposit freezes, bank holidays, or both.
(21.) The authors test and verify that their results are not skewed by the recent financial crisis, and go on to build a model to capture the salient features in their data set. This interesting contribution is not a DSGE macro model and therefore is not included in the present discussion.
(22.) That is, including banking, currency, and sovereign debt crises, as well as twin (banking plus currency) and triple (banking plus currency plus sovereign debt) crises.
(23.) This data, as noted previously, expands on the 14 countries covered in Schularick and Taylor (2012), on which they based their influential argument that banking crises are credit booms gone bust, as reviewed previously. That is, credit extended by the banking sector has been the main predictor of financial crises in the 20th century (in advanced economies). Bordo and Meissner (2016) appear to question some of Schularick and Taylor’s conclusions.
(24.) Bordo and Meissner (2016) report some dissimilarities in the cases of sovereign debt crises and currency crises that identify some significant differences, but these are less relevant for our interests here.
(25.) This issue is revisited in the section “Directions for Future Research.”
(26.) The 2016 version of their paper appears to update the March 2015 NBER Working Paper 2102 of the same name.
(27.) Thus, for the Romer and Romer index, rather than a time period being a period of no crisis or a period of crisis, they define things as follows: 0 corresponds to no financial distress; 1, 2, and 3 to gradations of credit disruptions; 4, 5, and 6 to gradations of minor crises; 7, 8, and 9 to gradations of moderate crises; 10, 11, and 12 to gradations of major crises; and 13, 14, and 15 to gradations of extreme crises.
(28.) As noted, for Romer and Romer’s measure of financial distress, 7 indicates a systemic crisis. Six such episodes occurred before 2007: Finland, Norway, and Sweden in the early 1990s; Japan in the 1990s and early 2000s; Turkey in the early 2000s; and the United States around 1990. The same six episodes show up in the Reinhart and Rogoff (2009) and Laeven and Valencia (2008, 2012) chronologies, although Reinhart and Rogoff do not classify the US episode around 1990 as systemic. Romer and Romer also do not classify Spain in the late 1970s and early 1980s and Turkey in 1980s as systemic crises, although other researchers’ chronologies do.
(33.) See Gorton and Metrick (2012) for a description of the role of the interbank market in the recent crisis. They emphasize the role of securitization in the repo market. Such features have not yet been explicitly worked into DSGE models. See Gorton and Ordoñez (2014).
(34.) As is noted later in this article, international macroeconomists for some time have been building models of small, open economies with similar features (occasionally binding constraints) in order to analyze so-called sudden stops and related issues in developing economies. The methodology and model developed by Brunnermeier and Sannikov (2014) are distinct, however, along a number of important dimensions.
(38.) If there are multiple types of loans, and retail banks have a comparative advantage in booking some of the these alternative types, then it need not be the case that retail banks make loans solely because the wholesale sector has reached loan capacity.
(39.) That is, while retail funding may be cheaper for wholesale banks, it is more heavily rationed than interbank funding. Wholesale banks are able to increase leverage by relying solely on interbank deposits. Nevertheless, interbank funding is also limited by the banks’ incentive to divert funds. Similarly, by making interbank loans, retail banks are able to attract more deposits and further increase leverage per unit of net wealth.
(40.) For example, anticipated runs unfold as follows: The authors assume that the conditional probability of a run is exogenously linked to the ratio of net assets to liabilities; the closer the wholesale bank is to not being able to meet its liabilities, the higher the perceived probability of a run. This assumption in effect attaches risk premia on loan rates to wholesale banks and can lead to a so-called slow run, where credit extended to the wholesale sector is incrementally tightened as the sector’s solvency is increasingly called into question.