Cyclical Patterns of Systemic Risk Metrics: Cross-Country Analysis1
Author: Mr. Plamen K Iossifov

Contributor Notes

Author’s E-Mail Address: piossifov@imf.org.
Publication Date:
05 Feb 2021
eISBN:
9781513568652
Language:
English
Keywords:
credit cycle.; WP; risk metrics; risk index; risk indices; solvency risk; interest rate; mispricing risk; solvency indicator; risk indicator; risk signal; interconnectedness risk; pricing risk metrics; risk concept
Download PDF (1.9 MB)

We analyze a range of macrofinancial indicators to extract signals about cyclical systemic risk across 107 economies over 1995–2020. We construct composite indices of underlying liquidity, solvency and mispricing risks and analyze their patterns over the financial cycle. We find that liquidity and solvency risk indicators tend to be counter-cyclical, whereas mispricing risk ones are procyclical, and they all lead the credit cycle. Our results lend support to high-level accounts that risks were underestimated by stress indicators in the run-up to the 2008 global financial crisis. The policy implications of conflicting risk signals would depend on the phase of the credit cycle.

Abstract

We analyze a range of macrofinancial indicators to extract signals about cyclical systemic risk across 107 economies over 1995–2020. We construct composite indices of underlying liquidity, solvency and mispricing risks and analyze their patterns over the financial cycle. We find that liquidity and solvency risk indicators tend to be counter-cyclical, whereas mispricing risk ones are procyclical, and they all lead the credit cycle. Our results lend support to high-level accounts that risks were underestimated by stress indicators in the run-up to the 2008 global financial crisis. The policy implications of conflicting risk signals would depend on the phase of the credit cycle.

I. Overview

Modern approaches to financial system oversight aim to contain systemic risk, but their practice is complicated by lack of consensus on how to quantify such risks. At conceptual level, systemic risk is the risk of disruptions in the provision of financial services, caused by financial system impairment, that creates serious negative effects on the real sector (IMF, FSB, and BIS, 2009). Defined in this way, it is clear that systemic risk is multidimensional—reflecting the complexity of the financial system—and its intensity is directly observable, only when risks materialize, in the size of the resultant financial and real sector losses. There are many empirical approaches for measuring systemic risk, but an industry standard or set of best practices are yet to emerge (see Bisias and others (2012) and Blancher and others (2013) for surveys of the field). The focus of this paper is on macro-level approaches for measuring systemic risk that provide a lay-of-the-land snapshot of risks across time and countries. Macro-level approaches rely on analyzing the dynamics of sectoral and market aggregates, and as such are not well suited to capture the early stages of a buildup of risks in individual financial institutions and market participants. It is also difficult to capture with aggregate indices risks arising from interconnectedness of financial institutions and market participants, weaknesses in financial supervision, financial integrity, financial market infrastructure, and so on in a timely and consistent manner. A comprehensive analysis of these pertinent issues requires stress testing and other in-depth tools, such as those used in the Financial Sector Assessment Program (IMF and WB, 20052).

Existing macro-level approaches for measuring systemic risk can be classified in two broad categories: “bottom-up” or indicator-based and “top-down” or model-based. Bottom-up approaches proxy systemic risk by vulnerabilities of the financial sector and asset markets that are more directly measurable. The degree of risk is then judged by the distribution of its proxy metrics across time and potentially across peer countries. In contrast, top-down methods estimate empirical models, in which the dependent variable is the incidence or magnitude of a financial system disruption that affects negatively the real sector. The fitted values of the model over a given time horizon or their change over time are then used as measures of systemic risk. Bottom-up approaches are typically feasible even with sparse data and are capable of revealing new risk patterns. On the downside, they do not control for mitigating factors—as extreme values always raise red flags—and involve a higher degree of subjectivity in interpreting risk signals to arrive at a bottom-line assessment. Top-down approaches seek to redress these issues, but are prone to overfitting—as models are selected for their ability to match historical risk patterns—and could be very data intensive. Examples of top-down and bottom-up approaches are, respectively, the Growth-at-Risk model (Adrian, Grinberg, Liang, and Malik, 2018) and the Financial Stability Monitoring Framework (Adrian, He, Liang, and Natalucci, 2019) used in the IMF’s Global Financial Stability Report (GFSR).

In this paper, we present a streamlined, bottom-up approach for measuring cyclical systemic risks with macroeconomic data, applicable across a diverse set of economies and geared to a broad audience. Our approach builds upon similar frameworks used in the U.S. Office of Financial Research (OFR) Financial System Vulnerabilities Monitor (OFR, 2020), the European Financial Stability Board Risk Dashboard (ESRB, 2020), and the GFSR’s Financial Stability Monitoring Framework. Whereas existing approaches tend to apply different risk concepts to different sectors of the economy,3 we propose a streamlined risk nomenclature—liquidity, solvency, and mispricing risks—that can be used uniformly, as applicable, across all sectors of the economy. The concept of mispricing risk—which aims at capturing possible asset-price misalignments or weakening credit standards—is related to Adrian, Covitz, and Liang (2015) notion of pricing of risk, and extends it further to economic agents’ choice of balance-sheet exposures. Adopting a streamlined bottom-up approach allows us to apply a harmonized analytical framework to a bigger and more diverse set of countries, as we can select risk proxies on a case-by-case basis, depending on data availability.

In the empirical part of the paper, we construct indices of liquidity, solvency and mispricing risks for 107 countries and analyze their patterns over the financial cycle. Our objective is to identify combinations of liquidity, solvency, and mispricing risk metrics typical for various phases of the financial cycle that can be used to inform policy responses.

We contribute to the existing literature by using macro-level risk metrics for a bigger and more diverse set of countries and analyzing their evolution over the credit cycle at different leads/lags. The interest in this topic has been primarily driven by bank regulators’ and international financial institutions’ efforts to document financial system developments that led to the 2008 Global Financial Crisis (GFC). A key takeaway from this literature is that risks were underestimated in the run-up to the crisis, with risk-based solvency indicators remaining broadly stable or modestly improving (BIS, 2009; Shin, 2014) and measures of market risk falling through mid-2006 (Shin, 2014). The sparse empirical literature on the topic supports the countercyclicality and leading behavior of market risk metrics over the financial cycle (e.g., Aikman, Lehnert, Liang, and Modungno, 2020), but finds that bank-level, solvency indicators also tend to be countercyclical (i.e., they tend to decrease in the upswing phase of the business/credit cycle and increase in downturns) (Brei and Gambacorta, 2016; Montagnoli, Mouratidis, and Whyte, 2020). However, existing studies focus only on the contemporaneous link between solvency and business/credit cycle indicators, leaving open the possibility that the interrelation between them can be of different sign and strength at different lags/leads. We contribute to the literature by using macro-level risk metrics (instead of bank-level data for individual indicators) for a bigger and more diverse set countries and applying statistical techniques capable of analyzing cyclicality at different leads/lags.

The paper is organized as follows. Section II presents the analytical underpinnings of the proposed disaggregation of cyclical systemic risk into underlying liquidity, solvency, and mispricing risks. Section III presents the dataset of macrofinancial indicators used to extract signals about underlying risks. These indicators are then used to construct economy-wide risk indices for 107 countries since 1995 that are then optimized. In Section IV, we analyze the cyclicality patterns of our preferred indices of liquidity, solvency, and mispricing risk, using Stock and Watson (1999) cyclicality analysis and an event study of their behavior around systemic bank crises. In Section V, we interpret the identified patterns of the optimized economy-wide risk indices over the financial cycle from a policymaker’s point of view. Section VI concludes with a summary of our main findings.

II. Analytical Framework

Systemic risk is the risk of disruptions in the provision of financial services, caused by impairment of all or parts of the financial system with serious negative effects for the real sector (IMF, FSB, and BIS, 2009). The definition of the financial system potentially encompasses all financial institutions, financial markets, and the financial infrastructure (Houben, Kakes, and Schinasi, 2004). The disruption in the provision of financial services can be triggered by negative shocks originating within the financial system, in the rest of the economy, or from abroad, and manifests itself in: (1) falling asset prices and increased volatility (Eichengreen and Portes, 1987; Bordo and Schwartz, 2000; Illing and Liu, 2003); (2) exchange rate depreciation or losses of official foreign reserves (Sachs, Tornell, and Velasco, 1995; Eichengreen, Rose, and Wyplosz, 1996); (3) widespread insolvencies and defaults of borrowers, lenders and market participants (Bordo, Dueker, and Wheelock, 2000; Breuer, 2004; Claessens and Kose, 2014); and (4) rising interest rates or disruption in the provision of credit (IMF, 1998; Kaminsky and Reinhart, 2001).

Financial stability risks capture different aspects of systemic risk. Financial stability refers to the state of the financial system that minimizes the probability of systemic risk materialization.4 Financial instability is triggered by negative shocks to the financial system that propagate through existing financial vulnerabilities (Adrian, He, Liang, and Natalucci, 2019). The pricing of risk—measured by the slack or tightness of financial conditions5— affects economic agents’ optimal exposure to financial vulnerabilities and is, itself, impacted by shocks, thus acting as an amplifier of negative shocks in the financial system (Adrian, Covitz, and Liang, 2015). Given the unpredictability of the timing and nature of shocks (IMF, 2019), financial stability risks are typically proxied by metrics of financial vulnerabilities and pricing of risk.

In this paper, we decompose the “time dimension” of systemic risk (IMF, FSB, and BIS, 2016) into three underlying risks—solvency, liquidity and mispricing risks. They encompass, respectively, the three categories of risks identified as pertinent to systemic risk analysis in the IMF’s Guidance Note for Surveillance under Article IV Consultation (IMF, 2015):

  • Solvency risk—refers to the potential inability of economic agents to pay off all of their liabilities to other agents even after liquidating all assets. Solvency risk often arises from excessive leverage or exposure to risky assets.

  • Liquidity risk—reflects the potential inability of borrowers to meet their obligations as they fall due, without incurring losses large enough to deplete available liquidity buffers.

  • Mispricing risk—captures the potential for mispricing of risk by economic agents or asset markets. It encompasses both the possibility of under/over-estimation of risk in asset markets—reflected in the slack/tightness of financial conditions6—and by economic agents through excessive (de)leveraging and under/over-exposure to specific financial instruments and asset classes:

    • In asset markets, underestimation of risk often manifests in protracted risk-on sentiment (in the extreme in irrational exuberance), characterized by loose financial conditions. Trading on market momentum can drive a wedge between asset prices and their fundamental determinants. This, in turn, can give rise to self-reinforcing dynamics/ amplification resulting in boom-bust asset price cycles with associated underestimation/overestimation of risks.

    • At a sectoral level, underestimation of risk manifests in loosening of credit standards that often becomes evident only ex-post. This can give rise to overextension (leveraging) of balance sheets of creditors and debtors and concentration of balance-sheet exposures to specific sectors (such as households) or specific financial instruments (such as unhedged foreign currency loans.) and, through them, to asset prices (such as house prices and exchange rates.). On bank balance sheets, the speed of accumulation7 and resultant concentration of exposures result in the build-up of credit/market/interest rate risks.

Mapping of Risks to 2015 Art IV Guidance Note Risk Taxonomy

Citation: IMF Working Papers 2021, 028; 10.5089/9781513568652.001.A001

Source: IMF (2015).Note: For simplicity, the mapping reflects only the first-round effects of shocks on sectoral balance sheets.

The other dimension of systemic risk—“cross-sectional” or “structural” (IMF, FSB, and BIS, 2016)— arises from interlinkages between economic agents both domestically and cross-border. Interconnectedness risk reflects the potential of spread of solvency, liquidity, and mispricing risks to economic agents other than those, on whose balance sheets the risks originated. Lack of harmonized data across a broad set of countries prevents us from including interconnectedness risk in our analysis.

The above risk taxonomy is a stylized representation of the first-round, differential impact of shocks on the balance sheets of economic sectors. In reality, the three types of risks are interlinked and prone to negative feedback loops.8 For example, negative real sector shocks carry the potential of eroding the debt servicing capacity of different sectors of the economy, increasing their exposure to debt-related liquidity risks. Investors’ flight to safety can increase the precautionary demand for liquidity, putting pressure on the price of risky and illiquid assets, and triggering deflation of asset price bubbles and widening of credit risk spreads. The correction in asset markets can overshoot and morph into upside mispricing risks (i.e., increased potential for overestimation of risks). Falling asset prices and higher debt service costs erode sectoral net worths, ratcheting economic agents’ exposures to liquidity and solvency risks.

Systemic risk builds up in the expansion and peak phases of the financial cycle and can materialize and subside in its correction/trough phase. The financial cycle captures “the level and evolution of slack (or excess) in the financial sector” (IMF, 2015). Systemic risk is procyclical (Caruana, 2010), as by definition when properly identified it should be highest prior to its materialization that manifests in financial retrenchment in the trough of the cycle. At the same time, a central finding of the literature on financial frictions (Brunnermeier, Eisenbach, and Sannikov, 2012) is that they can give rise to feedback loops between asset prices and economic agents’ net worths and liquidity spirals. As a result, metrics that serve as proxies for liquidity and solvency risks can exhibit countercyclical behavior, resulting in underestimation of risks and accentuation of the cycle.9 High-level reviews of financial system developments that led to the 2008 Global Financial Crisis provide empirical support of this conjecture, as risk-based solvency indicators were found to have remained broadly stable or modestly improved (BIS, 2009; Shin, 2014) and measures of market risk to have fallen through mid-2006 (Shin, 2014).

III. Data Description and Index Construction

A. Data Compilation

The starting point of the analysis is the compilation of data on relevant macrofinancial indicators that can serve as proxies for the three types of risks. The raw dataset consists of quarterly data for 180 plus countries and 48 variables over 1995:Q1–2020:Q3 (Table 1).10 Data are drawn from publicly available data sources, including various IMF databases,11 Bloomberg (for financial market data), OECD and Global Property Guide (for housing market data), the Credit Research Initiative (CRI) database of the National University of Singapore’s Risk Management Institute (for probabilities-of-default data), and ECB and OECD sectoral accounts data (see Appendices I and II for details). The selection of indicators aims at ensuring maximum country coverage beyond G-20 and OECD countries. Whereas data coverage is uneven both in the time and cross-country dimensions, the sample includes information on more than 15 variables for over 100 countries since 2002.

Table 1.

Initial Set of Indicators for Construction of Aggregate Risk Metrics

article image
Note: Internally at the IMF, the compiled data are housed in the Systemic Risk Tracker, developed by staff of the Strategy, Policy, and Review Department as a central repository of macrofinancial indicators that can be used as proxies of different aspects of systemic risk.

The selection of indicators is guided by the nature of the risks being proxied, a review of the empirical financial crisis literature, and the goal of maximizing country coverage:

  • Solvency risk is proxied by measures of the adequacy and resilience of sectoral capital buffers;

  • Liquidity risk is proxied by measures of the adequacy and resilience of the debt servicing capacity of borrowers and lenders, as well as signs of funding difficulties;

  • Mispricing risk is proxied by measures of slack/tightness of financial conditions, as reflected in market returns, interest rates and market volatility, and signs of loosening of credit standards, as reflected in the speed of accumulation and resultant concentration of balance-sheet exposures.

Given the link between systemic risk and the financial cycle, the empirical, financial crises literature offers insights on the types of variables that can signal financial stress either ahead of time or in the credit cycle downturn. A review of the literature validates the focus on sectoral financial vulnerabilities and signs of excess in financial markets, as their proxies frequently appear as explanatory variables in regression models of financial crises (Appendix III).

Table 1 presents the initial set of indicators used in the analysis. Risk indicators that meet the above criteria are grouped by sector of the economy. In our analysis, we use data only for the private sector and limit the analysis of the financial sector to banks, due to the sparse data availability for the non-bank financial sector. Where necessary, variables are transformed, so that higher values correspond to heightened underlying risks. This is achieved by taking the absolute value of indicators, for which extreme values in both directions could constitute a risk, as well as taking the inverse of ratios and multiplying growth rates by minus one for indicators, for which low raw values signal the buildup of risk. The performed transformations are noted in brackets in the last column of Table 1.

The distribution of indicators across time provides a measure of the relative intensity of underlying risks in individual countries. The data for each variable are transformed into percentile ranks,12 based on the distribution of its values over the entire time span for each individual country. The percentile ranks can be, alternatively, calculated across countries at similar level of economic development over time. Basing the analysis on country-specific percentile ranks allows us to control more granularly for idiosyncratic characteristics of the sample countries not captured by their level of development. Expressed in percentile ranks, the macrofinancial indicators can be interpreted as risk metrics, the extreme values of which raise red flags. Higher values of liquidity and solvency risk metrics signal build-up of risks (either intrinsic in balance-sheet weaknesses or sentiment-driven). Higher values of the mispricing risk metrics reflect heightened risk of financial excess (e.g., high market returns and low volatility) or loosening of credit standards (e.g., overextension of balance sheets of creditors and debtors and concentration of balance-sheet exposures to specific sectors or specific financial instruments.

As noted in the overview section, indicator-based risk monitoring frameworks, such as ours, extract signals from data without attempting to single out deviations from fundamentals or control for policies. As all extreme values of risk proxy indicators raise red flags, the extracted signals may be noisy at times, especially when it comes to asset price-based indicators that are highly sensitive to shocks and changes in the real economy. In analyzing risk signals from mispricing risk indicators, in particular, care should be taken not to interpret them literally as an accurate measure of asset-price misalignments, but rather as a signal that misalignment could be developing or worsening. Indicator-based risk metrics can be used in second-stage analysis by country experts to control for fundamentals, policies and special circumstances in individual countries.

The composite risk metrics, constructed from percentile ranks of the indicators shown in Table 1, would generally move in the opposite direction of untransformed indicators used in the received literature. This is due to the transformations we carry out, so that higher percentile rank corresponds to heightened underlying risk (in the case of mispricing risk: heightened risk of financial excess). Existing studies generally analyze untransformed individual indicators, high values of which indicate low degree of risk (in the case of mispricing risk: low risk of financial excess). For example, Brei and Gambacorta (2016) and Montagnoli, Mouratidis, and Whyte (2020) examine solvency indicators, in which bank capital is in the nominator. High values of such indicators are then associated with low solvency risk. Similarly, Shin (2014) discusses CDS spreads, which when high would indicate low risk of financial excess.

We use the credit cycle as a stand-in for the broader financial cycle, leveraging on Schularick and Taylor (2012) influential finding that “financial crises throughout modern history can be viewed as credit booms gone wrong”. Following Claessens, Kose, and Terrones (2012), we use the dynamics of the real growth of private debt to capture the credit cycle. We choose a measure of credit dynamics that is not conditioned on the notion of long-term trend relative to real sector developments, in order to avoid the well-known beginning and end-point measurement problems in calculating other widely used measures, such as the credit-to-GDP gap.13

B. Construction of Composite Risk Metrics

Raw indices of underlying risks can be constructed by averaging of the percentile ranks of all available indicators. However, despite the conceptual appeal of the initial set of indicators (Table 1), the quality of their empirical signals may differ due to issues such as measurement errors and reporting delays.

We enhance the signal-to-noise ratio of our composite risk metrics, using the Cronbach’s Alpha methodology and Stock and Watson cyclicality analysis on individual indicators:

  • We first narrow down the list of indicators, based on the Cronbach’s Alpha estimate of the reliability, with which they proxy the same economy-wide risk concept when taken as a group (OECD, 2008). The square root of the Cronbach’s Alpha can be interpreted as providing an estimate of the correlation of a composite metric—called “test scale” and constructed by summing up the standardized values of individual risk metrics— with an underlying factor—in this case economy-wide solvency, liquidity or mispricing risk (StataCorp, 2019b).

  • We then weed out indicators that behave differently over the financial cycle than the majority of proxies for the same type of risk. As noted above, we use the credit cycle—as captured by the dynamics of real growth of private debt—as a stand-in for the broader financial cycle. Using Stock and Watson (1999) cyclicality analysis, we drop indicators that exhibit markedly different intertemporal correlation patterns with real growth of private debt—used to capture the credit cycle— than those of the majority of proxies for the same type of risk for a given sector of the economy. This is necessary because we later aggregate the signals from individual indicators using arithmetic averages. Doing so with series that behave differently over the financial cycle would tend to reduce the signal-to-noise ratio of our composite risk metrics.

The two techniques are chosen for their flexibility in handling missing values, which are a prominent feature of our dataset, and empirical tractability and replicability.

Cronbach’s Alpha

The Cronbach’s Alpha (“C-alpha”) is given by the following expression (OECD, 2008):

αc=kk−1(1−Σj=1kσxj2σx02)k−number of items x ;σxj2− variance of item xj;x0=Σj=1kxj−test scale.

“C-alpha measures the portion of total variability of the sample of individual indicators due to the correlation of indicators. It increases with the number of individual indicators and with the covariance of each pair. “ (OECD, 2008, p. 72). We apply the C-alpha analysis on the individual risk metrics, the values of which are transformed, where needed, in the way noted in the last column of Table 1 and then expressed in percentile ranks. The percentile transformation ensures that all variables are of the same scale, which makes further standardization—prior to combining them into the test scale—unnecessary. The C-alpha is especially well suited for the purposes of our analysis, as the sequentially optimized test scale is constructed in the same way as both the raw composite sector risk metrics and the preferred, economy-wide, risk indices, derived as an outcome of the analysis in this section.14 In practice, the values of the calculated C-alpha should be seen as indicative, as its theoretical properties depend on the absence of missing values (SAS Institute Inc, 2016), whereas the latter are allowed in its empirical implementation (StataCorp, 2019b).

Table 2 presents the narrowed down list of indicators obtained from the application of the Cronbach’s Alpha methodology. Cronbach’s Alpha analysis of the initial set of indicators suggests that the reliability of the three economy-wide scales, constructed from available indicators of solvency, liquidity and mispricing risks, can be improved by dropping indicators that are weakly correlated with other proxies of the same type of risk (Appendix Table 1). We proceed to drop items with negative or positive but low correlation with a scale constructed from all items except the one under consideration (see item-rest correlations in Column 5 of Appendix Table 1). The resulting sub-set of risk proxies proxy more reliably underlying risks with estimated Cronbach’s Alpha coefficients in the second-stage of the analysis (Table 2) within the 0.6–0.8 range commonly considered acceptable (OECD, 2005; Goforth, 2015). Several of the remaining indicators are less correlated with other proxies of the same type of risk (Column 5 of Table 2), but dropping them would not improve significantly the Cronbach’s Alpha coefficients (see last column of Table 2), which is why we keep them at this stage.

Table 2.

Cronbach’s Alpha Analysis of Narrowed-Down Set of Indicators of Economy-Wide Risks

article image
Source: Authors’ estimates using StataCorp (2019a).
Stock and Watson (1999) cyclicality analysis

We use the Stock and Watson (1999) methodology to analyze the behavior of the indicators shown in Table 2 over the credit cycle, with the view of optimizing further the signal-to-noise ratio of aggregate risk metrics derived from them. Appendix Figure 2 presents the distributions across sample countries of the cross-correlations of individual indicators, grouped by sector and underlying risk, with leads and lags of real growth of private debt (expressed in percentile ranks over the entire time span for each individual country). The cross-country distributions of these cross-correlations provide an indication of the statistical significance of the observed common patterns in the data. Following Stock and Watson (1999), if the highest correlation is with one of the lags of real growth of private debt, we conclude that the risk metric follows the credit cycle with a delay. Alternatively, if the highest correlation is with a given lead of the variable, we determine that the risk metric leads the credit cycle. If the maximum correlation is positive, the risk metric is procyclical, whereas if it is negative, it is countercyclical vis-à-vis private sector debt.

We proceed to drop indicators, which intertemporal correlation patterns with real growth of private debt deviate significantly from the common patterns by sector/risk or are available only for a small number of countries. We first drop from the list the real growth rates of sub-components of private debt, used to capture different aspects of mispricing risk, in order to avoid mechanical correlation patterns.15 As seen in Appendix Figure 2, the real growth rates of bank loans to households and bank loans and external debt of corporates are contemporaneously correlated with real growth of private debt, of which they are sub-components. Next, we exclude from the list the real domestic government bond yield and house price-to-income and house price-to-rent ratios, as their median sample correlations are clustered around zero for all leads and lags of real growth of private debt. We also drop the LIBOR-OIS spread and interest payments-to-income ratios for households and corporates from the preferred list, as they are available only for a small number of countries (see Table 2). Finally, we remove from the list the share of non-performing loans in total bank loans and interest payments-to-income ratios for households and corporates, as in contrast with the other risk metrics, they lag rather than lead the credit cycle.16

Preferred composite risk metrics

Table 3 presents the preferred set of 20 indicators, obtained from applying the Cronbach’s Alpha and Stock and Watson (1999) cyclicality analyses on the initial set of indicators. The remaining solvency and liquidity risk metrics cover the banking, corporate, and household sectors, whereas the selected mispricing risk metrics are only asset price-based (Table 3). A number of these or derivative indicators have also been found to be useful predictors of financial crises (Appendix III). Examples include capital-to-assets ratio and return-on-equity (Jordà and others, 2017), loan-to-deposit ratio (Navajas, 2013), real stock market returns (Schularick and Taylor, 2012), housing price growth (Babecky and others, 2014), private sector debt service ratio (Drehmann and Juselius, 2014) and bank loans to households-to-GDP ratio (Alessi and Detken, 2018).

Table 3.

Preferred Set of Indicators for Construction of Aggregate Risk Metrics by Sector/Market

article image

We use the preferred set of 20 indicators to construct economy-wide indices of underlying risks for 107 countries over the period 1995:Q1–2020:Q3. For each type of risk, the economy-wide index is calculated by averaging the percentile ranks of the indicators first by sector/market and then across sectors, effectively giving each sector/market equal weight in the resultant economy-wide risk index. As noted above, expressed in percentile ranks, the preferred indicators can be interpreted as risk metrics, the extreme values of which raise red flags. We opt for simple averages to enhance the transparency of the constructed risk indices. In the case of mispricing risk, we calculate the index only if it is based on information on more than one financial market.

IV. Empirical Analysis

We analyze the patterns of the constructed economy-wide indices of liquidity, solvency, and mispricing risk over the financial cycle, using Stock and Watson (1999) cyclicality analysis and an event study of their behavior around systemic bank crises.17 As in the index-construction stage of the analysis, we use the credit cycle—as captured by the dynamics of real growth of private debt—as a stand-in for the broader financial cycle.

A. Stock and Watson (1999) Cyclicality Analysis

Analysis of cyclicality patterns shows that liquidity and solvency risk indices are counter-cyclical, whereas the mispricing one is procyclical, and they all lead the credit cycle. Figure 1 presents the distributions across sample countries of the cross-correlations of economy-wide risk indices with leads and lags of real growth of private debt (expressed in percentile ranks over the entire time span for each individual country). Results suggest that liquidity and solvency risk indices are counter-cyclical in levels, whereas the mispricing one is procyclical, and they all lead the credit cycle by at least four quarters. Across sample countries, the intertemporal correlation between the real growth of private debt and the economy-wide liquidity and solvency risk indices peaks in negative territory around the fourth lead of real growth of private debt. The opposite holds for the economy-wide, asset price-based mispricing risk index. Its intertemporal correlation with real growth of private debt peaks in positive territory around the fourth lead of real growth of private debt. Box 1 presents a case study that illustrates the main findings from our cross-country analysis with the experience of Denmark in the run-up and aftermath of the 2008 Global Financial Crisis.

The economy-wide indices of underlying risks cannot be combined directly in an overall index of systemic risk. This is because the metrics of mispricing risk do not co-move with liquidity and solvency risk metrics over the credit cycle (Figure 1).18

Figure 1.
Figure 1.

Sample Distribution of Cross-Correlograms of Preferred Economy-Wide Risk Indices and Real Growth of Private Debt

(Correlation Coefficients of Variables Expressed in Percentile Ranks within Countries’ Own History)

Citation: IMF Working Papers 2021, 028; 10.5089/9781513568652.001.A001

Notes: The box plots show, on the vertical axis, the distribution of individual country correlation coefficients of preferred economy-wide risk indices with different leads and lags of real growth of private debt. All variables are expressed in percentile ranks over the entire time span for each individual country. The leads and lags are shown on the horizontal axis; In box plots, the lower and upper hinges of each box show the 25th and 75th percentiles of the distribution, the line in the box indicates the median, and the end-points of whiskers mark next adjacent values. Following Stock and Watson (1999), a large positive correlation at k=0 indicates that the two series co-move in the same direction; a large negative correlation at k=0 shows that the two series move in opposite directions; a maximum correlation at negative k (e.g., k=-1) indicates that the risk metric follows developments in real credit growth with a lag of k quarters; a maximum correlation at positive k (e.g., k=1) indicates that the risk metric leads developments in real credit growth with a lead of k quarters.

Denmark: Systemic Risk Patterns over a Full Credit Cycle

Over the period 2003–11, the Danish economy witnessed all phases of the credit cycle, in parallel with a house prices boom/bust cycle accompanied by a secular increase in household leverage. The figure below shows the growth of real private debt alongside the economy-wide risk indices, which are centered around their medians and expressed as contributions to an aggregate index, in which they enter with equal weights of one-third.

The dynamics of the different indices of underlying risks are consistent with the findings outlined above:

  • Liquidity and solvency risk metrics are countercyclical and lead the cycle, progressively falling below their medians in the upswing of the credit cycle before reversing course close to its peak and swinging into positive territory in the downturn and trough phases of the credit cycle.

  • The mispricing risk index is, on the other hand, pro-cyclical, and also leads the credit cycle. It signals rising potential for underestimation of risks in the upswing and peak phases of the credit cycle, before receding and starting to flag a growing potential for overestimation of risks in the downturn and trough phases.

Denmark: Evolution of Underlying Risks Over the Credit Cycle, 2003–11

Citation: IMF Working Papers 2021, 028; 10.5089/9781513568652.001.A001

Source: Authors’ estimates.Note: The economy-wide risk indices are in percentile ranks centered around the median (50th percentile rank = 0) and are expressed as contributions to an aggregate index, in which they enter with equal weights of one-third.

B. Event Analysis around Systemic Banking Crises

Systemic bank crises are extreme cases of systemic risk materialization in the correction phase of the credit cycle. Real growth of private debt typically peaks in the year prior to a systemic bank crisis and then steeply declines as the crisis unfolds, bottoming out two years after its start (Appendix Table 2). Taking a closer look at risk patterns at the extreme can bring out salient features of the data that would otherwise remain hidden.

An event study of the behavior of economy-wide risk indices around systemic bank crises reaffirms the findings from the cyclicality analysis and provides further insights of their potential to act as early warning indicators. Figure 2 presents the average values of the economy-wide indices of liquidity, solvency and mispricing risks in levels around 65 systemic banking crises since 1995. The timing of systemic bank crises is taken from Laeven and Valencia (2018). The chart shows the average within-country percentile ranks of these indices, centered around their medians, 12 quarters before and after a systemic banking crisis that occurs in period t=0. Results show that the mispricing risk index is procyclical vis-à-vis the credit cycle and is near its peak two to three years ahead of a systemic banking crisis. It builds and remains above its median in the upswing phase of credit cycles that end with a banking crisis, plateaus and then recedes in the year leading up to the downturn phase, and bottoms out below its median in the trough phase of the cycle. Liquidity and solvency risk indices are, on the other hand, countercyclical in levels—they tend to be lower in the upswing phase than in the downturn phase of credit cycles that end with a banking crisis. At the same time, they do build up in the year leading up to the downturn (moving in the opposite direction of mispricing risk metrics), peaking above their medians in the trough phase of the cycle, before gradually tapering off.

Figure 2.
Figure 2.

Evolution of Preferred Economy-Wide Risk Indices around Systemic Banking Crises

Citation: IMF Working Papers 2021, 028; 10.5089/9781513568652.001.A001

Source: Authors’ estimates.Notes: Shown are averages of indices in levels (centered around their medians) across 65 systemic banking crises since 1995.

V. Interpretation of Findings

Our main finding—that economy-wide liquidity and solvency risk metrics are countercyclical, whereas the mispricing ones are procyclical, and they all lead the credit cycle—provide a blueprint of expected patterns of risks and their proxy metrics over the phases of the financial cycle:19

  • Build-up phase that in the extreme can give rise to financial manias, credit booms, or asset price bubbles. Solvency, liquidity, and mispricing risks are initially generally benign in both the financial and real sectors of the economy. Over time, the virtuous cycle of financial deepening and real growth acceleration can morph in a self-perpetuating cycle of credit expansion, unsustainable pace of income growth, and asset price inflation. The resulting build-up of downside mispricing risk (i.e., increased potential for underestimation of risks) is characterized by high rates of return, low volatility, compressed risk premia, and rapid credit growth. Rising asset prices increase collateral valuations, making existing loans appear increasingly better provisioned, and encouraging banks and their clients to take more risk (solvency risk is initially underestimated by its proxy metrics). Higher net worths and faster income growth compress risk premia, which alongside greater credit availability, initially mask liquidity risk as captured by its proxy metrics. Risk management can become laxer, further amplifying downside mispricing risks, as financial market participants increasingly trade on market momentum and financial institutions’ management and operation processes become progressively strained by the increased volumes of business. As leverage of debtors and creditors rises, their sensitivity to abrupt changes in incomes, interest rates, and—where credit is extended in foreign currency—exchange rate shocks, making it more likely that rising solvency and liquidity risks are registered by their proxy metrics.

  • Correction phase that in the extreme can turn into financial market crashes, financial sector panics, or credit crunches. Excessive risk-taking during the build-up phase make the economy vulnerable to negative external and internal shocks, increasing the likelihood of “hard landing”, as these shocks erode the debt servicing capacity of different sectors of the economy. Investors’ flight-to-safety increases the precautionary demand for liquidity, putting pressure on the price of risky and illiquid assets, triggering deflation of asset price bubbles and widening of credit risk spreads. The correction of mispricing risks can overshoot and morph into upside mispricing risks (i.e., increased potential for overestimation of risks), characterized by negative rates of return, high volatility, large risk premia, increase of non-performing loans in banks’ portfolios, and credit crunch. Falling asset prices and higher debt service costs erode sectoral net worths, ratcheting economic agents’ exposure to liquidity and solvency risks and raising the potential for overestimation by their proxy metrics. These can be further magnified by changes in investor sentiment and herding behavior. The decline of collateral valuations and the heightened liquidity and solvency risks could trigger a tightening of credit underwriting standards, further ratcheting upside mispricing risks and draining liquidity from the financial system. Solvency and liquidity risks, as captured by their proxy metrics, may initially continue to rise before subsiding, as incomes and net worths can fall faster than the pace of deleveraging, giving rise to Fisherian debt deflation spiral (Fisher, 1933).

  • Peak and trough phases—characterized by inflection points in the level or change of intensity of underlying risks. In the peak phase, the self-reinforcing dynamics of downside mispricing risk stall, making it even more likely for solvency and liquidity risk metrics to register rising risks and give impetus for a negative feedback loop between rising solvency and liquidity risk and falling mispricing risk metrics. In the trough phase, the process of repair of overleveraged balance sheets eventually improves the creditworthiness of borrowers, restores the capacity of creditors to underwrite risks, and rebuilds investor confidence, setting the stage for a new systemic risk cycle.

Our results suggest that low values of liquidity, solvency, and mispricing risk indices need to be evaluated in reference to the phase of the financial cycle. Low levels of liquidity and solvency aggregate risk metrics would tend to underestimate risks in the buildup phase of the credit cycle, due to the interplay between liquidity and net worth of creditors/debtors and credit-driven booms of real activity and asset prices. Low values of mispricing aggregate risk metrics in the correction phase of the cycle may not be benign, as the correction of financial market excess and concentration of balance-sheet exposures may overshoot, potentially spreading solvency, liquidity, and mispricing risks to otherwise sound parts of the financial system and the real economy.

Economy-wide risk indices have the potential to serve as early warning indicators of banking crises. The event study suggests that mispricing risk metrics in levels provide useful early warning signals two to three years ahead of banking crisis. Increases in liquidity and solvency risk indices, especially when they occur alongside downward correction in mispricing risk metrics can provide additional early warning signals of banking crises, albeit with shorter lead time of about a year.

VI. Conclusion

In this paper, we advance the literature on indicator-based metrics of systemic risk by proposing a harmonized risk taxonomy across sectors and applying it to a bigger and more diverse set of countries. We group systemic risk metrics into proxies for liquidity, solvency, and mispricing risks. The first two risks are standard in the literature. Mispricing risk aims at capturing possible asset-price misalignments or weakening credit standards. We then use macrofinancial indicators to construct optimized, economy-wide risk indices for 107 countries over 1995–2020 . We use them to show that liquidity and solvency risk metrics are countercyclical, whereas the mispricing one is procyclical. All risk metrics lead the credit cycle by at least a year, pointing to their potential to act as early-warning indicators of banking crises.

In contrast to evidence from bank-level studies of solvency dynamics, our results lend support to high-level accounts that risks were underestimated by stress indicators in the run-up to the 2008 GFC. When comparing our findings to those in the received literature, one needs to take into account that high values of our composite risk metrics indicate heightened underlying risks (in the case of mispricing risk: higher risk of financial excess). As a result, the solvency risk index would move in the opposite direction of ratios, such as the leverage ratio and the capital-to-risk weighted assets ratio, in which capital buffers are in the nominator. In the same vein, the composite mispricing risk index would move in the opposite direction of financial price indicators that capture financial stress (for example, CDS spreads). With this in mind, our results strongly support high-level accounts that risks were underestimated by stress indicators in the run-up to the 2008 GFC. In our view, conflicting evidence from existing bank-level studies of bank solvency dynamics can be explained by the studies’ focus on their contemporaneous link with the business/credit cycle.

The relative importance assigned to signals from the various risk metrics would depend on the relevant time horizon for policy action. In the correction and trough phases of the cycle, the focus of the policy response is on contemporaneous prevention of risk spillovers. The need for and design of policies to engineer a soft-landing of the economy and crisis response policies (for example, relaxation of the macroprudential regime, liquidity and equity support for businesses, bank recapitalization) can then be informed by high levels of solvency and liquidity risk indices and low values of the mispricing risk metrics. In the buildup and peak phases of the cycle, policy response should, instead, be forward-looking, aiming to stem risks before they appear on sectoral balance sheets and in asset valuations. Tightening of macroprudential policies aimed at containment of systemic risk can then be informed by high levels of mispricing risk indices and rising solvency and liquidity risk metrics.

Areas for future research include testing the early-warning indicator properties of risk indices and reconciliation of the findings from macro and micro-level empirical studies. With respect to the latter, our finding that all three, economy-wide risk indices lead the credit cycle points to the need for further research of the interrelation between bank-level solvency indicators and the business/credit cycle at different lags/leads.

References

  • Adrian, Tobias, Dong He, Nellie Liang, and Natalucci, Fabio, 2019, “A Monitoring Framework for Global Financial Stability,” IMF Staff Discussion Notes, 19/06, available on the internet at: www.imf.org/~/media/Files/Publications/SDN/2019/SDNEA2019006.ashx, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Adrian, Tobias, Federico Grinberg, Nellie Liang, and Malik, Sheheryar, 2018, “The Term Structure of Growth-at-Risk,” IMF Working Paper, 18/180, available on the internet at: www.imf.org/~/media/Files/Publications/WP/2018/wp18180.ashx, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Adrian, Tobias, Daniel Covitz, and Liang, Nellie, 2015, “Financial Stability Monitoring,” Annual Review of Financial Economics, 7, December, available on the internet at: doi.org/10.1146/annurev-financial-111914-042008, pp. 357–95.

    • Search Google Scholar
    • Export Citation

  • Aikman, David, Andreas Lehnert, Nellie Liang, and Modungno, Michele, 2020, “Credit, Financial Conditions, and Monetary Policy Transmission,” International Journal of Central Banking, 62, available on the internet at: www.ijcb.org/journal/ijcb20q2a4.htm, Jun, pp. 141–179.

    • Search Google Scholar
    • Export Citation

  • Alessi, Lucia and Detken, Carsten, 2018, “Identifying Excessive Credit Growth and Leverage,” Journal of Financial Stability, 35, Apr, available on the internet at: doi.org/10.1016/j.jfs.2017.06.005 pp. 215–225.

    • Search Google Scholar
    • Export Citation

  • Babecky, Jan, Tomas Havránek, Jakub Matějů, Marek Rusnák, Katefina Šmídková, and Vašíček, Bofek, 2014, “Banking, Debt, and Currency Crises in Developed Countries: Stylized Facts and Early Warning Indicators,” Journal of Financial Stability, 15, Dec, available on the internet at: doi.org/10.1016/j.jfs.2014.07.001, pp. 1–17.

    • Search Google Scholar
    • Export Citation

  • Blancher, Nicolas, Srobona Mitra, Hanan Morsy, Akira Otani, Tiago Severo, and Valderrama, Laura, 2013, “Systemic Risk Monitoring (“SysMo”) Toolkit—A User Guide,” IMF Working Paper, WP/13/168, available on the internet at: www.imf.org/~/media/Websites/IMF/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13168.ashx, (Washington: IMF).

    • Search Google Scholar
    • Export Citation

  • BIS, 2009, The role of valuation and leverage in procyclicality, CGFS Papers 34, Apr., available on the internet at: www.bis.org/publ/cgfs34.pdf, (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • BIS, 2017, Basel III Leverage Ratio Framework – Executive Summary, BIS FSI Executive Summaries, Oct., available on the internet at: www.bis.org/fsi/fsisummaries/b3_lrf.pdf, (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Bisias, Dimitrios, Mark Flood, Andrew W. Lo, and Valavanis, Stavros, 2012, “A Survey of Systemic Risk Analytics,” Annual Review of Financial Economics, 4(1), available on the internet at: doi.org/10.1146/annurev-financial-110311-101754, Oct., pp. 255–296.

    • Search Google Scholar
    • Export Citation

  • Bordo, Michael, Michael Dueker, and David Wheelock, 2000, “Aggregate Price Shocks and Financial Instability: Historical Analysis,” NBER Working Paper, 7652, available on the internet at: www.nber.org/papers/w7652.pdf, (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Bordo, Michael and Schwartz, Anna, 2000, “Measuring Real Economic Effects of Bailouts: Historical Perspectives on How Countries in Financial Distress Have Fared with and without Bailouts,” NBER Working Paper, 7701, available on the internet at: www.nber.org/papers/w7701.pdf, (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Borio, Claudio and Lowe, Phillip, 2002, “Asset Prices, Financial and Monetary Stability: Exploring the Nexus,” BIS Working Paper, No. 114, available via Internet: www.bis.org/publ/work114.pdf, (Basel: Bank for International Settlements) .

    • Search Google Scholar
    • Export Citation

  • Brei, Michael and Gambacorta, Leonardo, 2016, “Are Bank Capital Ratios Pro-Cyclical? New Evidence and Perspectives,” Economic Policy, 31(86), Apr, available on the internet at: https://doi.org/10.1093/epolic/eiw001, pp. 357–403.

    • Search Google Scholar
    • Export Citation

  • Breuer, Janice, 2004, “An Exegesis on Currency and Banking Crises,” Journal of Economic Surveys, 18(3), July, available on the internet at: doi.org/10.1111/j.0950-0804.2004.00223.x, pp. 293–320.

    • Search Google Scholar
    • Export Citation

  • Brunnermeier, Markus, Thomas Eisenbach, and Sannikov, Yuliy, 2012, “Macroeconomics with Financial Frictions: A Survey,” NBER Working Paper, 18102, available on the internet at: www.nber.org/papers/w18102.pdf, (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Caruana, Jaime, 2010, Systemic risk: how to deal with it?, available on the internet at: www.bis.org/publ/othp08.htm, BIS Other Publications, (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Claessens, Stijn, M. Kose, and Terrones, Marco, 2012, “How do Business and Financial Cycles Interact?,” Journal of International Economics, 87, available on the internet at: https://doi.org/10.1016/j.jinteco.2011.11.008, pp. 178–190.

    • Search Google Scholar
    • Export Citation

  • Claessens, Stijn and Kose, M. Ayhan, 2014, “Financial Crises: Explanations, Types, and Implications,” Chapter 1 in Stijn Claessens, M. Kose, L. Laeven, and F. Valencia, eds, Balance Sheet Recession is the Reason for Secular Stagnation, available on the internet at: www.elibrary.imf.org/doc/IMF071/20264–9781475543407/20264–9781475543407/Other_formats/Source_PDF/20264–9781484324295.pdf, pp. 3–59, (Washington: IMF).

    • Search Google Scholar
    • Export Citation

  • Drehmann, Mathias and Juselius, Mikael, 2014, “Evaluating Early Warning Indicators of Banking Crises: Satisfying Policy Requirements,” International Journal of Forecasting, 30(3), Jul, available on the internet at: doi.org/10.1016/j.ijforecast.2013.10.002, pp. 759–780.

    • Search Google Scholar
    • Export Citation

  • Drehmann, Mathias and Tsatsaronis, Kostas, 2014, “ The Credit-to-GDP Gap and Countercyclical Capital Buffers: Questions and Answers,” BIS Quarterly Review, Mar, available on the internet at: www.bis.org/publ/qtrpdf/r_qt1403g.pdf, pp. 55–73.

    • Search Google Scholar
    • Export Citation

  • ECB, 2011, “The Financial Crisis in the Light of the Euro Area Accounts: A Flow-Of-Funds Perspective,” ECB Monthly Bulletin, October, pp. 99–120, available on the internet at: www.ecb.europa.eu/pub/pdf/mobu/mb201110en.pdf, (Frankfurt: European Central Bank).

    • Search Google Scholar
    • Export Citation

  • ECB, 2013, “Structural Issues Report: Corporate Finance and Economic Activity in the Euro Area,” ECB Occasional Paper, 151, available on the internet at: www.ecb.europa.eu/pub/pdf/scpops/ecbocp151.pdf, (Frankfurt: European Central Bank).

    • Search Google Scholar
    • Export Citation

  • ESRB, 2020, ESRB Risk Dashboard, available on the internet at: www.esrb.europa.eu/pub/rd/html/index.en.html, (Frankfurt: European Systemic Risk Board).

    • Search Google Scholar
    • Export Citation

  • Eichengreen, Barry and Richard Portes, 1987, “The Anatomy of Financial Crises,” NBER Working Paper, 2126, available on the internet at: www.nber.org/papers/w3934.pdf, (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Eichengreen, Barry, Andrew K. Rose, and Charles Wyplosz, 1996, “Contagious Currency Crises,” NBER Working Paper, 5681, available on the internet at: www.nber.org/papers/w5681.pdf, (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Eichengreen, Barry and Arteta, Carlos, 2000, “Banking Crisis in Emerging Markets: Presumptions and Evidence,” UC Berkeley: Center for International and Development Economics Research, available via Internet: https://escholarship.org/uc/item/3pk9t1h2.

    • Search Google Scholar
    • Export Citation

  • EU, 2013, European System of Accounts, available on the internet at: https://ec.europa.eu/eurostat/en/web/products-manuals-and-guidelines/-/KS-02-13-269, (Luxembourg: Publications Office of the European Union).

    • Search Google Scholar
    • Export Citation

  • Fisher, Irving, 1933, “The Debt-Deflation Theory of Great Depressions,” Econometrica, 1, pp. 337–357.

  • Frankel, Jeffery and Saravelos, George, 2010, “Are Leading Indicators of Financial Crisis Useful for Assessing Country Vulnerability? Evidence from the 2008–2009 Global Crisis,” NBER Working Paper, No. 16047.

    • Search Google Scholar
    • Export Citation

  • Goforth, Chelsea, 2015, “Using and Interpreting Cronbach’s Alpha,” University of Virginia Library online article, available on the internet at: data.library.virginia.edu/using-and-interpreting-cronbachs-alpha/.

    • Search Google Scholar
    • Export Citation

  • Hermansen, Mikkel and Rohn, Oliver, 2015, “Economic Resilience: The Usefulness of Early Warning Indicators in OECD Countries,” OECD Economics Department Working Papers, No. 1250, available on the internet at: doi.org/10.1787/eco_studies-2016-5jg2ppjrd6r3, (Paris: OECD).

    • Search Google Scholar
    • Export Citation

  • Holopainen, Markus and Sarlin, Peter, 2017, “Toward Robust Early-Warning Models: A Horse Race, Ensembles and Model Uncertainty,” Journal Quantitative Finance, 17(12), available on the internet at: doi.org/10.1080/14697688.2017.1357972, pp. 1933–1963.

    • Search Google Scholar
    • Export Citation

  • Houben, Aerdt, Jan Kakes, and Garry Schinasi, 2004, “Toward a Framework for Safeguarding Financial Stability,” IMF Working Paper, WP/04/101, available on the internet at: www.imf.org/external/pubs/ft/wp/2004/wp04101.pdf, (Washington: IMF).

    • Search Google Scholar
    • Export Citation

  • Illing, Mark, and Ying Liu, 2003, “An Index of Financial Stress for Canada,” Bank of Canada Working Paper, 14, available on the internet at: ideas.repec.org/p/bca/bocawp/03-14.html, (Ottawa: Bank of Canada).

    • Search Google Scholar
    • Export Citation

  • IMF, 1998, “Financial Crises: Characteristics and Indicators of Vulnerability,” Chapter IV in World Economic Outlook, May, available on the internet at: www.imf.org/~/media/Websites/IMF/imported-flagship-issues/external/pubs/ft/weo/weo0598/pdf/_0598ch4pdf.ashx, (Washington: IMF).

    • Search Google Scholar
    • Export Citation

  • IMF and WB, 2005, Financial Sector Assessment: A Handbook, available on the internet at: www.worldbank.org/en/programs/financial-sector-assessment-program/publication/financial-sector-assessment-handbook, (Washington: International Bank for Reconstruction and Development/World Bank/International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • IMF, 2015, Guidance Note for Surveillance under Article IV Consultation, available on the internet at: www.imf.org/~/media/Websites/IMF/imported-full-text-pdf/external/np/pp/eng/2015/_031915.ashx, (Washington, DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • IMF, 2017, “Financial Conditions and Growth at Risk,” Chapter 3 in Global Financial Stability Report “Is Growth at Risk,” Oct, available on the internet at: www.imf.org/en/Publications/GFSR/Issues/2017/09/27/~/media/Files/Publications/GFSR/2017/October/chapter-3/Documents/c3.ashx?la=en, (Washington, DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • IMF, 2019, “Technical Note to Chapter 1: Vulnerabilities in a Maturing Credit Cycle,” Online Annex 1.1. to Chapter 1 in Global Financial Stability Report “Vulnerabilities in a Maturing Credit Cycle,” Apr, available on the internet at: www.imf.org/~/media/Files/Publications/GFSR/2019/April/English/onlineannex11.as hx?la=en, (Washington, DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • IMF, FSB, and BIS, 2009, Guidance to Assess the Systemic Importance of Financial Institutions, Markets and Instruments: Initial Considerations, Report to the G-20 Finance Ministers and Central Bank Governors, available on the internet at: www.bis.org/publ/othp07.pdf, (Washington, DC: International Monetary Fund, Financial Stability Board, and Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • IMF, FSB, and BIS, 2016, Elements of Effective Macroprudential Policies, available on the internet at: http://www.imf.org/external/np/g20/pdf/2016/083116.pdf, (Washington, DC: International Monetary Fund, Financial Stability Board, and Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • IMF and World Bank, 2005, Financial Sector Assessment: A Handbook, available on the internet at: www.imf.org/external/pubs/ft/fsa/eng/index.htm, (Washington, DC: The International Bank for Reconstruction and Development/The World Bank/The International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Iossifov, Plamen and Khamis, May, 2009, “Credit Growth in Sub-Saharan Africa: Sources, Risks, and Policy Responses,” IMF Working Paper, WP/09/180, available on the internet at www.imf.org/~/media/Websites/IMF/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09180.ashx, (Washington: IMF).

    • Search Google Scholar
    • Export Citation

  • Jordà, Òscar, Moritz Schularick, and Taylor, Alan, 2011, “Financial Crises, Credit Booms, and External Imbalances: 140 Years of Lessons,” IMF Economic Review, 59(2), Jun, available on the internet at: link.springer.com/article/10.1057/imfer.2011.8, pp. 340–378.

    • Search Google Scholar
    • Export Citation

  • Jordà, Òscar, Bjorn Richter, Moritz Schularick, and Talyor, Alan, 2017, “Bank Capital Redux: Solvency, Liquidity, and Crisis,” NBER Working Paper, No. 23287, available on the internet at: https://www.nber.org/papers/w23287.

    • Search Google Scholar
    • Export Citation

  • Kaminsky, Graciela and Carmen Reinhart, 2001, “Financial Markets in Times of Stress,” NBER Working Paper, 8569, available on the internet at: www.nber.org/papers/w8569.pdf (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Laeven, Luc and Fabian Valencia, 2018, “Systemic Banking Crises Revisited”, IMF Working Paper, WP/18/206, available on the internet at: www.imf.org/~/media/Files/Publications/WP/2018/wp18206.ashx, (Washington: IMF).

    • Search Google Scholar
    • Export Citation

  • Lahnsteiner, M., 2013, “Private Sector Debt in CESEE EU Member States,” Focus on European Economic Integration, Q3, available on the internet at: www.oenb.at/dam/jcr:aadceab2-7495-4676-ab61-15a28d60cedf/feei_2013_q3_studies_lahnsteiner_tcm16-257384.pdf, pp. 30–47, (Vienna: Oesterreichische Nationalbank).

    • Search Google Scholar
    • Export Citation

  • Manasse, Paolo, Roberto Savona, and Vezzoli, Marika, 2013, “Rules of Thumb for Banking Crisis in Emerging Markets,” Working Paper, 872, available on the internet at: papers.ssrn.com/sol3/papers.cfm?abstract_id=2236733, (Universita’ di Bologna: Dipartimento Scienze Economiche).

    • Search Google Scholar
    • Export Citation

  • Manasse, Paolo, Roberto Savona, and Vezzoli, Marika, 2016, “Danger Zones for Banking Crises in Emerging Markets,” International Journal of Finance & Economics, 21, available on the internet at: doi.org/10.1002/ijfe.1550, pp. 360–381.

    • Search Google Scholar
    • Export Citation

  • Montagnoli, Alberto, Konstantinos Mouratidis, and Whyte, Kemar, 2020, “Assessing the cyclical behaviour of bank capital buffers in a finance-augmented macro-economy,” Journal of International Money and Finance, 110, Feb, available on the internet at: doi.org/10.1016/j.jimonfin.2020.

    • Search Google Scholar
    • Export Citation

  • Navajas, Matias and Thegeya, Aaron, 2013, “Financial Soundness Indicators and Banking Crises,” IMF Working Paper, WP/13/263, available via Internet: www.imf.org/~/media/Websites/IMF/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13263.ashx, (Washington, DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • OECD, 2008, Handbook on Constructing Composite Indicators: Methodology and User Guide, available on the internet at: www.oecd.org/sdd/42495745.pdf, (Paris: Organization for Economic Co-operation and Development).

    • Search Google Scholar
    • Export Citation

  • OFR, 2019, 2019 Annual Report to Congress, available on the internet at: www.financialresearch.gov/annual-reports/files/OFR-Annual-Report-2019.pdf, (Washington: Office of Financial Research).

    • Search Google Scholar
    • Export Citation

  • OFR, 2020, Financial System Vulnerabilities Monitor, available on the internet at: www.financialresearch.gov/financial-vulnerabilities/#/, (Washington: Office of Financial Research).

    • Search Google Scholar
    • Export Citation

  • Sachs, Jeffrey, Aaron Tornell, and Andres Velasco, 1995, “Financial Crises in Emerging Markets: The Lessons from 1995,” NBER Working Paper, 5576, available on the internet at: www.nber.org/papers/w5576.pdf, (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • SAS Institute Inc, 2016, Base SAS® 9.4 Procedures Guide: Statistical Procedures, Sixth Edition, available on the internet at: documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.5, (Cary: SAS Institute Inc.).

    • Search Google Scholar
    • Export Citation

  • Schularick, Moritz and Taylor, Alan, 2012, “Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870–2008,” American Economic Review, 102(2), available on the internet at: www.aeaweb.org/articles/pdf/doi/10.1257/aer.102.2.1029, Apr., pp. 1029–61.

    • Search Google Scholar
    • Export Citation

  • Shin, Hyun, 2014, “Procyclicality and the Search of Early Warning Indicators,” Chapter 4 in Stijn Claessens, M. Kose, L. Laeven, and F. Valencia, eds, Balance Sheet Recession is the Reason for Secular Stagnation, available on the internet at: www.elibrary.imf.org/doc/IMF071/20264-9781475543407/20264-9781475543407/Other_formats/Source_PDF/20264-9781484324295.pdf, pp. 157–71, (Washington: IMF).

    • Search Google Scholar
    • Export Citation

  • StataCorp, 2019a, Stata Statistical Software: Release 16, (College Station, TX: StataCorp LLC).

  • StataCorp, 2019b, Multivariate Statistics Reference Manual, Release 16, available on the internet at: https://www.stata.com/manuals/mv/index2.html, pp. 11–19, (College Station, TX: StataCorp LLC).

    • Search Google Scholar
    • Export Citation

  • Stock, James and Watson, Mark, 1999, “Business Cycle Fluctuations in U.S. Macroeconomic Time Series,” Chapter 1 in JB. Taylor and M. Woodford, eds, Handbook of Macroeconomics, 1(A), available on the internet at: www.sciencedirect.com/science/article/pii/S1574004899010046, pp. 3–64.

    • Search Google Scholar
    • Export Citation

Appendix I. Definitions of Macrofinancial Indicators used in Analysis

article image
article image
Notes:BIS TCS – Bank for International Settlements Total Credit Statistics;BIS PP – Bank for International Settlements Property Prices Statistics;BIS-IMF-WB – Joint BIS-IMF-OECD World Bank External Debt Statistics;CRI – Credit Research Initiative (CRI) database of the National University of Singapore’s Risk Management Institute;Eurostat – Eurostat Institutional Sector Accounts database;GPG – Global Property Guide;IMF FSI- IMF Financial Soundness Indicators database;IMF IFS- IMF International Financial Statistics database;IMF WEO – IMF World Economic Outlook database;OECD- Organization for Economic Cooperation and Development Sectoral Accounts database.

Appendix II. Definitions of Macrofinancial Indicators Derived from Sectoral Accounts

The 2008 System of National Accounts (SNA) contains standards for data collection at the level of sectors of the economy. It classifies all agents in the domestic economy in four institutional sectors—households, non-financial corporations, financial corporations (including the central bank) and general government. “The sectoral accounts present the accounts of institutional sectors in a coherent and integrated way, linking – similar to the way in which profit and loss, cash flows and balance sheet statements are linked in business accounting– uses/expenditure, resources/revenue, financial flows and their accumulation into balance sheets from one period to the next. … Accordingly, the sectoral accounts present the data with three constraints: each sector must be in balance vertically (e.g. the excess of expenditure on revenue must be equal to financing); all sectors must add up horizontally (e.g. all wages paid by sectors must be earned by households); and transactions in assets/ liabilities plus holding gains/losses and other changes in the volume of assets/liabilities must be consistent with changes in balance sheets (stock-flow consistency).” (ECB, 2011, p. 103).

The debt stock includes the outstanding amounts of loans and debt securities on the liability side of sectoral balance sheets. In line with the approach taken by the European Commission under the Macroeconomic Imbalance Procedure (MIP), the definition excludes financial derivatives, trade credit, and other accounts payable (EC, 2012a). Data are unconsolidated within each sector (i.e., transactions between constituents of the same sector are recorded in gross terms, rather than netting them out), except in the case of the general government, for which only consolidated data are published. We rely on unconsolidated financial accounts data for the private sector, as this information is used in ratios that involve non-consolidated data from non-financial accounts for all domestic sectors, except the general government.

The two liquidity risk metrics are constructed as the ratios of the stock of debt and interest payments to the augmented gross disposable income of the various sectors. The resulting two metrics are the debt-to-income and interest payments-to-income ratios. In constructing the former, the stock of debt is used as a proxy for the relative size of principal debt repayments (data on which are not available) over time, as well as across countries under the implicit assumption of similar maturity structures. In SNA, the interest payments made by borrowers are split between “SNA interest” and a financial intermediation service charge indirectly measured (FISIM). Only “SNA interest” is recorded as interest revenues and expenditures in the non-financial accounts, whereas FISIM is classified mostly under final consumption—in the case of households and government—and intermediate consumption, in the case of financial and non-financial corporations.20 In order to reconstruct the total interest payments made by sectors, we calculate interest payments as the sum of “SNA interest” and FISIM (see also Lahnsteiner, 2013). We proxy the debt servicing capacity of each sector of the economy by its gross disposable income (GDI) taken before interest payments and, in the case of banks and corporates, also before payments to shareholders (i.e., reinvested earning on FDI and distributed income of corporations) (augmented gross disposable income).21,22 The adjustments are required because: (1) money spent on interest payments is part of sectors’ debt servicing capacity; and (2) bond holders have priority over shareholders in the distribution of profits.

The solvency risk metric (or leverage ratio) is constructed as the ratio of the stock of debt to firms’ capital/households’ net worth.23 The capital of non-financial and financial corporations and households’ net worth are defined as the difference between the respective sector’s assets (both financial and non-financial) and liabilities other than equity.24 However, many countries do not publish data on non-financial assets, which prevents us from constructing precise measures of firms’ capital/households’ net worth comparable across countries. Instead, we follow the existing literature by proxying households’ net worth by the difference between financial assets and liabilities; and financial and non-financial corporations’ capital—by the value of “Shares and Other Equity”. In the case of firms, our metric will be a close approximation of firms’ capital, if the “Tobin’s Q” is equal to one.25

Sectoral interest rate – income growth differentials are used as proxies for mispricing risk. The implicit interest rate is calculated as the ratio of interest payments (including both “SNA interest” and FISIM) over the average of beginning and end-period stock of debt of each sector, except for financial corporations. The calculation cannot be performed for financial corporations, because their interest payments also include the payment of interest on deposits, which are not included in the definition of debt in the denominator of the ratio. The definition of the augmented gross disposable income of all sectors is the same as the one used in the construction of the liquidity risk metric.

Appendix III. Literature Review of Empirical Studies of Financial Crises

article image
Legend: Confidence levels: ••• – 99 percent •• – 95 percent • – 90 percent