Are corporate bond defaults contagious across sectors?

Ellis, Colin Article Are corporate bond defaults contagious across sectors? International Journal of Financial Studies Provided in Cooperation with: MDPI – Multidisciplinary Digital Publishing Institute, Basel Suggested Citation: Ellis, Colin (2020) : Are corporate bond defaults contagious across sectors?, International Journal of Financial Studies, ISSN 2227-7072, MDPI, Basel, Vol. 8, Iss. 1, pp. 1-17, https://doi.org/10.3390/ijfs8010001 This Version is available at: https://hdl.handle.net/10419/257669 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ International Journal of Financial Studies Article Are Corporate Bond Defaults Contagious across Sectors? Colin Ellis † Department of Economics, University of Birmingham, Birmingham B15 2TT, UK; [email protected] †The views in this paper represent those of the author, and not necessarily any institutions he is (or has been) associated with. Received: 26 July 2019; Accepted: 24 December 2019; Published: 7 January 2020   Abstract: Corporate bond defaults in different sectors often increase suddenly at roughly similar times, although some sectors see default rates jump earlier than others. This could reflect contagion among sectors—specifically, defaults in one sector leading to credit stresses in other sectors of the economy that would not otherwise have seen stresses. To complicate matters, simple correlation-based tests for contagion are often biased, reflecting increased volatility in periods of stress. This paper uses sectoral default data from over 30 sectors to test for signs of contagion over the past 30 years. While jumps in sectoral default rates do often coincide, there is no consistent evidence of contagion across different periods of stress from unbiased test results. Instead, coincident jumps in sectoral default rates are likely to reflect common macroeconomic shocks. Keywords: credit risk; corporate bonds; default contagion JEL Classification: G12; G24; G33 1. Introduction When economic or market conditions deteriorate, we often see simultaneous declines in asset prices, as market participants reassess prospects for revenue, profits, and affordability. In light of these worsening macroeconomic and financial conditions, credit conditions for borrowers often deteriorate rapidly, leading to firms being unable to meet agreed interest payments or being unable to refinance maturing debt. For these reasons, both non-performing loans held by banks and default rates for corporate bonds—covering both intermediated and disintermediated credit provision—tend to rise in recessions. Default rates can jump quite quickly (Figure 1), and these increases are often evident across a range of sectors. In some instances, default rates have increased in one corporate sector before any increase is evident in others. This can give the impression that credit stresses “seep” from one sector to another, which, together with the high correlations in default rates across sectors in recessions, can give the impression of contagion—that is, corporate defaults in one sector having knock-on causal effects that drive up defaults in other sectors. Tests for contagion have been used in a variety of other markets including across equity markets in emerging economies. However, there has been relatively little work to test whether corporate bond defaults actually are contagious in the sense that stress in one sector triggers stress in another sector. This is the fundamental question that this paper addresses. This paper re-examines the empirical question of contagion in corporate bond markets, in particular looking at whether increases in default rates in one corporate sector exhibit contagion effects to default rates in other sectors. The remainder of this paper is structured as follows. Section 2reviews past work on testing for contagion, and the importance of accounting for the broader macroeconomic environment and in Int. J. Financial Stud. 2020,8, 1; doi:10.3390/ijfs8010001 www.mdpi.com/journal/ijfs Int. J. Financial Stud. 2020,8, 1 2 of 17 particular taking explicit account of shifts between highand low-volatility periods. Section 3then presents the data underpinning this analysis, looking at sectoral defaults; Section 4then details the methodological approach taken and presents results from the contagion analysis. Finally, Section 5 concludes the paper. Int. J. Financial Stud. 2020, 8, x; doi: FOR PEER REVIEW www.mdpi.com/journal/ijfs Article Are Corporate Bond Defaults Contagious across Sectors? Colin Ellis † Department of Economics, University of Birmingham, Birmingham B15 2TT, UK; [email protected] † The views in this paper represent those of the author, and not necessarily any institutions he is (or has been) associated with. Received: 26 July 2019; Accepted: 24 December 2019; Published: 8 January 2020 Abstract: Corporate bond defaults in different sectors often increase suddenly at roughly similar times, although some sectors see default rates jump earlier than others. This could reflect contagion among sectors—specifically, defaults in one sector leading to credit stresses in other sectors of the economy that would not otherwise have seen stresses. To complicate matters, simple correlationbased tests for contagion are often biased, reflecting increased volatility in periods of stress. This paper uses sectoral default data from over 30 sectors to test for signs of contagion over the past 30 years. While jumps in sectoral default rates do often coincide, there is no consistent evidence of contagion across different periods of stress from unbiased test results. Instead, coincident jumps in sectoral default rates are likely to reflect common macroeconomic shocks. Keywords: credit risk; corporate bonds; default contagion JEL codes: G12; G24; G33 1. Introduction When economic or market conditions deteriorate, we often see simultaneous declines in asset prices, as market participants reassess prospects for revenue, profits, and affordability. In light of these worsening macroeconomic and financial conditions, credit conditions for borrowers often deteriorate rapidly, leading to firms being unable to meet agreed interest payments or being unable to refinance maturing debt. For these reasons, both non-performing loans held by banks and default rates for corporate bonds—covering both intermediated and disintermediated credit provision—tend to rise in recessions. Default rates can jump quite quickly (Figure 1), and these increases are often evident across a range of sectors. 0% 2% 4% 6% 8% 10% 12% 14% 16% Jan-70 Jan-74 Jan-78 Jan-82 Jan-86 Jan-90 Jan-94 Jan-98 Jan-02 Jan-06 Jan-10 Jan-14 Jan-18 Figure 1. Global speculative-grade default rate. Source: Moody’s. 2. Past Research: Defining and Testing for Contagion and the Problem of Time-Varying Bias The presence of contagion across asset markets is important for several reasons. For instance, most investment strategies assume that diversification across different markets is an effective mechanism for reducing portfolio risk. This relies on low-risk correlations across different markets; but if contagion occurs after a negative shock, then correlations across markets will increase, limiting the benefit from diversification. On the other hand, if contagion is present across markets and countries, this could lead to a crisis in one sector having a negative impact on the second sector; and, even if the second sector’s economic fundamentals are sound, it could also face a crisis. Given the underlying fundamentals, this could justify intervention by the public authorities to stabilise the sector. A first key issue is to define contagion. This is less straightforward than might otherwise be expected, as different papers adopt different approaches and strategies for identifying shock propagation. In an international context, Dornbush et al. (2000) define contagion as a significant increase in cross-market linkages after a shock, either to an individual country or a group of countries. From this definition, an increase in co-movement across countries or sectors could look like contagion. Countries that are closely tied together will tend to exhibit strong correlations in economic performance, whether positive or negative shocks hit, and so co-movement across those countries may not vary much. But, if two countries are moderately but not strongly correlated, then if a shock to one country affects the other, then the co-movement and correlation could increase, suggesting contagion. This could be misleading. In explaining why, it is useful to draw upon Masson (1997) who defines three mechanisms for this propagation in a cross-country context: aggregate shocks which affect the economic fundamentals of more than one country; country-specific shocks which affect the economic fundamentals of other countries; and shocks which are not explained by fundamentals and are categorized as pure contagion. The distinction matters because pure contagion reflects the transmission of a stress originating in one country or sector to another sector and, in particular, that without the contagion that second country or sector would not otherwise have experienced stress. In the case of aggregate shocks that affect all sectors, we would expect subsequent sectoral movements to be correlated, which could Int. J. Financial Stud. 2020,8, 1 3 of 17 look like contagion even though all sectors have been affected by the same shock, rather than one “infecting” another. The second mechanism, whereby stress in a first sector affects fundamentals in another, represents a direct transmission; Eichengreen et al. (1996) discuss these mechanisms. However, this transmission— and indeed the links it flows through—should be evident in economic and financial linkages such as declining demand for the second sector’s products or increased production problems as suppliers from the first sector are unable to provide intermediate goods that the second sector uses in production. But, even here, the transmission mechanism works through real economic fundamentals rather than via market sentiment or other channels. The third mechanism, whereby co-movements across sectors cannot be explained by changes in fundamentals, is generally thought to represent the formal definition of contagion—the stress in the second sector cannot be explained using the normal economic and financial linkages that are already known to exist. There are several theories that seek to explain the presence of contagion. As noted earlier, Masson (1997) presents a theory of multiple equilibria which shows that a crisis in one country can drive a shift in investor expectations (rather than real linkages) that then impacts on another country. Valdes (1996) argues that episodic disruptions in capital market liquidity can lead to fire sales in some asset classes, potentially affecting prices of assets in countries not affected by the initial shock. Mullainathan (2002) posits that investors do not accurately recall past events, and memories of past crises cause investors to re-evaluate their assumptions, in particular the probability of bad outcomes. Drazen (1998) asserts that political economy considerations can drive co-movements, citing the European devaluations of 1993 as an example of “bunching” in economic policy shifts. As Forbes and Rigobon (1999) note, despite the different mechanisms these theories represent, they all essentially treat contagion as the unexplainable component—the residual in any stylized or empirical analysis. This raises the obvious problem that, if we do not accurately capture real spillovers among sectors or economies—if our models are misspecified—then we might erroneously think that contagion exists when in fact it does not. Essentially, contagion is defined as the observation that cross-market linkages during a crisis are different than during relatively stable periods. Consistent with this view, economists have developed a variety of tests for contagion but which often have a common underlying component. This is best illustrated by one simple approach that compares the correlation (or covariance) among two markets or sectors during a relatively stable period with the correlation during a period of turmoil, often directly after a shock occurs. Contagion is then identified wherever there is a significant increase in the cross-market correlation during the period of turmoil, relative to the average correlation across periods of both calm and turmoil. A common problem with this, however, is that correlation coefficients are calculated incorrectly. Forbes and Rigobon (2002) describe this issue in detail in their seminal work. To summarise, unless we control for the higher variance that naturally exists in stressed periods, the correlation coefficients that are calculated to evaluation contagion will be biased and misleading, leading to contagion being identified where in fact none exists. Forbes and Rigobon (2002) demonstrated this effect when testing for contagion across equity markets and found that, while interdependence was present, actual contagion was far less prevalent than often assumed, and in particular was much less prevalent than suggested by misspecified correlation tests. This bias arises from the unadjusted correlation coefficient not taking account of different market conditions during periods of stress versus periods of calm. Over a long enough sample, summary statistics such as correlation coefficients may “average out” periods of relatively high and relatively low volatility. But the formal correlation-based test for contagion that is typically used is exactly that correlations among sectors or markets increase once market turmoil takes hold; and those increases can occur anyway if system-wide shocks occur without formal contagion actually being present. Essentially, because the data sample is split into “high” and “low” volatility periods, and we are comparing the high volatility period to an average of the high and low periods, the correlation test will naturally tend to signal contagion if the bias that is naturally present in the test is not accounted for. Int. J. Financial Stud. 2020,8, 1 4 of 17 Forbes and Rigobon (1999) provide a formal proof for this theorem, building on the previously work of Ronn (1998) who addresses bias in the estimation of intra-market correlations for stocks and bonds. The estimated—but unadjusted—correlation among two variables will increase when the variance of one of them increases, even if the actual correlation does not change. Happily, it is relatively straightforward to correct for this bias and, as such, calculate adjusted correlation coefficients that take this tendency into account. Loretan and English (2000) also note this effect. The original discussion in Forbes and Rigobon (1999) demonstrates how correlation coefficients can be biased when the variances of the underlying variables change, but for completeness this is revisited here. Suppose that xand yare stochastic variables, and they are related as follows: yt=α+βxt+t(1) where E[ εt ]=0, E[ εt2 ]< ∞ ,E[x t , εt ]=0 and also (for simplicity) that | β | < 1. Now suppose that the variance of x t changes over the sample period so that it is lower in one part of the sample (l) and higher in the second part of the sample (h). Since by assumption the error and explanatory (xt) variables are orthogonal, the consistent and efficient ordinary least squares (OLS) estimator will return the same parameter estimates across samples: βh = βl .By assumption, we also know that σhxx > σlxx which, when combined with the standard definition of β , implies that σhxy > σlxy : the covariance between x and ywill be higher in the second period. Since the residual variance is constant by construction and |β| < 1, we know that: σxx σyy !h > σxx σyy !l (2) Which, substituting into the standard definition of the correlation coefficient: ρ=σxy σxσy =βσx σy(3) implies that: ρh> ρl(4) In other words, the estimated correlation between xand ywill be higher when the variance of xincreases, even if the true correlation between xand ydoes not change. The standard correlation coefficient is conditional on the variance of x, and the bias this implies can be quantified as: ρu t=ρts1+δt 1+δtρ2 t (5) where ρu t is the unadjusted (or conditional) correlation coefficient and ρt is the actual (or unconditional) correlation coefficient. In this expression, δt is the relative increase in the variance of x. As such, during periods of high volatility and turmoil, estimated correlation coefficients will be biased upwards. Forbes and Rigobon (1999) also note that it is straightforward to adjust for this bias, solving Equation (5) for the unconditional correlation: ρt=ρu t q1+δt[1−ρu t2] (6) Using this simple approach, they demonstrate that the estimated contagion among stock markets during the late 1990s was largely a statistical artefact. That is because the common technique to test for contagion was to estimate whether there is a significant increase in correlation coefficients during periods of market turmoil, and while the biased estimated correlation coefficients did indicate that Int. J. Financial Stud. 2020,8, 1 5 of 17 contagion was present, the true unbiased correlation coefficients were much lower and typically did not suggest contagion. In other words, the test statistic upon which the contagion hypothesis was evaluated was biased, leading to misleading results and inference. Apart from Forbes and Rigobon (2002), other papers have also looked at different markets and episodes to test for contagion. Forbes and Rigobon (2001) themselves provide a broad survey of conceptual and empirical issues around measuring contagion. Since their original paper examining bias in correlation coefficients, methods for estimating contagion have become more sophisticated than just testing for changes in (biased) correlation coefficients. For instance, rather than using correlation tests over discrete samples, an alternative would be to estimate models with time-varying parameters to account for possible shifts (see for instance Ellis et al. (2014)). However, there have also been recent attempts to model contagion and network effects more explicitly in a time series format. One particular approach has been the emergence of correlation network models. As Giudici and Parisi (2018) note, while bivariate analysis and causal models can investigate whether the risk represented by a particular institution is affected by market crises or exogenous risks, these correlation network models try to explain whether that risk depends on endogenous contagion effects. These network models can take the form of contagion models that combine financial networks with price-based contagion models, such as those employed by Billio et al. (2012), who propose different statistical measures of connectedness, finding that different financial sectors have become more interrelated over time. Diebold and Yılmaz (2014) also propose a variety of connectedness measures based on variance decompositions that can be used to track time-varying connectedness of stock return volatilities. Ahelegbey et al. (2016) present an innovative Bayesian graphical approach to identification in vector autoregressive (VAR) models, finding strong unidirectional linkage from financial to non-financial sectors during the recent financial crisis and bidirectional linkages during the European sovereign debt crisis. However, it is unclear whether these linkages represent contagion per se, or more normal shock transmission. Das (2016) defines a new score of systemic risk that depends on individual risk and interconnectedness across financial institutions, where the network contributions implied by the latter could be interpreted as contagion. Interestingly, however, Das’s work on spillover risk finds that splitting up too-big-to-fail banks does not lower systemic risk, consistent with contagion effects from large firms not being the primary driver of systemic risk. More recently, Giudici and Parisi (2018) used a vector autoregressive (VAR) approach to model credit default swap (CDS) spreads, using cross-sectional dependency to model the contagion transmission mechanism. They then used the VAR to distinguish between contagion and idiosyncratic risk. Herculano (2018) used a Bayesian spatial autoregressive model to identify bank defaults and found that spillover effects among peers were positive and significant, although also very heterogeneous, with some banks being more important than others. However, spillovers within the banking sector may not apply in other sectors. In some instances, the definition of contagion is rather different than the fundamental approach proposed by Forbes and Rigobon. For instance, Ahrend and Goujard (2014) examined bilateral financial and trade linkages and equity and bond price movements between 2002 and 2011, finding that financial turmoil was transmitted through bilateral debt integration and common bank lenders. However, these identified transmission mechanisms were actually not indicative of formal contagion, which by definition should not be accounted for by increases in debt integration or common counterparties. Like many of the other related papers in this area, Mezei and Sarlin (2018) also focused on systemic risk rather than contagion per se. They promote a mechanism for aggregating individual risk and interconnectedness but seem more concerned with cross-border linkages than contagion as defined earlier (that is, the transmission of a shock from one entity to another other than via a common source or established transmission mechanisms). Int. J. Financial Stud. 2020,8, 1 6 of 17 Aamir and Shah (2018) investigated co-movements among Asian stock markets and Adad and Chulia (2013) examined the impact of monetary policy surprises on European government bond markets. Das et al. (2007) explored correlations of individual corporate defaults rather than across sectors, finding “excess” default clustering can be matched by assuming some “extra” correlation, and macroeconomic variables can account for some of this. However, individual corporate defaults can often be correlated within sectors due to the common exposures to exogenous developments and shocks—the spate of US shale defaults after the 2014/2015 decline in the oil price is a clear example of this. As such, these within-sector defaults may likely represent common exposures rather than genuine contagion that instead may be easier to identify across different sectors. There is some past evidence that the default of one company can increase the measured probability of another company defaulting (see Azizpour et al. 2018). Here, the authors used latent variable techniques to account for unobservable “frailty”, and contagion is assumed to follow the specific form proposed by Hawkes (1971). The results of this approach suggest that the impact of one default on the default rate has a half-life of only three months. However, it is questionable whether the modelling correctly identifies changes in underlying macroeconomic conditions (which, as noted above, can bias measures of contagion) from non-stationary shifts in the latent “frailty” factor. It is also striking that more sophisticated computational techniques employed to examine contagion over the past two decades are far from uniform in their assessments. For instance, Leschinski and Bertram (2017) use a “rolling window analysis” to identify contagion effects during the euro crisis, finding pure contagion from Italy and Spain but not for Greece, Ireland or Portugal. Yet, at the same time, Caporin et al. (2018) analysed sovereign risk contagion using quantile regressions focusing on the euro area and found that the risk spillover was not affected by the sign or size of the shock, implying that contagion has remained subdued. At best, these different results demonstrate that more complex modelling approaches still result in ambiguity about the role and impact of contagion. However, probably the most important paper for gauging the effectiveness of newer methods of testing for contagion is Rigobon (2016) which reviews the empirical literature. As with the simple correlation analysis discussed earlier, Rigobon notes that the econometric issues arising from endogeneity and omitted variables produce time-varying biases. All the empirical approaches examined, including VARs, event studies, autoregressive conditional heteroskedasticity models, non-linear regressions, suffer to a greater or lesser extent—put simply, Rigobon finds that there is no single technique that can completely address this issue without formally correcting for these biases. This implies that, while the techniques mentioned above are, to varying degrees, more complex than the statistical analysis implicit in correlation coefficients, the underlying structural data conditions that the original paper by Forbes and Rigobon (1999) speak to still matter. Whenever the variance of explanatory variables varies over the data sample, it is critical to adequately correct for this in the estimated variance–covariance statistics that underpin the coefficient estimates and other features of the approaches and models outlined above. Simply using a Granger causality test or VAR will not adequately adjust for the fact that statistical bias will be present in any estimated coefficients where the variances of the underlying data vary over time. As such, while the approach originally presented by Forbes and Rigobon (1999) is simple, compared with some of the complex alternatives employed today, it still serves a useful purpose: any model estimates will be biased unless the underlying biases arising from statistical analysis are accounted for. Sometimes, stripping away statistical sophistication to focus on the underlying issue can shed more light on an issue than resorting to greater computational complexity. As such, armed with the original approach set out by Forbes and Rigobon (1999)—and conscious that more complex approaches will suffer from the same concerns about bias that they illustrate—this analysis will now examine sectoral default rates for evidence of contagion. To the best of my knowledge, there has been little exploration of whether this type of cross-sectoral contagion exists in corporate bond markets which is the key contribution of this paper. Int. J. Financial Stud. 2020,8, 1 7 of 17 3. Sectoral Bond Defaults: Inspecting the Data Ultimately, this analysis regards the question of whether corporate bond distress is contagious across sectors as an empirical one. In order to investigate this, we need to examine corporate default rates across sectors to see if there is evidence of contagion from one sector to another. To be clear, this analysis examines empirical data for signs of contagion among different sectoral default rates. Each sectoral default rate was calculated as the number of default events, relative to the universe of rated entities within that sector, over a specified time period. The default data here were collected and published by Moody’s, a global credit rating agency. Moody’s has collated data on defaults for over a century and publishes default rates back to 1920 on some bases. In this instance, the data sample covered defaults by non-financial corporate issuers as a percentage of Moody’s rated universe. All the data used in this analysis were default rates (percentages) and were not absolute numbers of default events. The data are categorized according to one of 33 corporate sectors and default rates are published on an issuer-weighted basis. The entire data sample covered the period from the start of 1983 to the start of 2017; default rates were reported on a monthly frequency. Overall, the dataset comprised 13,530 observations across the 33 sectors considered here. Towards the start of the sample, the data were more heavily concentrated on US issuers, reflecting Moody’s geographical coverage; but, as Moody’s expanded into other markets, over time, more issuers from other parts of the world—particularly the UK, Germany, and France—were included in the data. More detail on Moody’s default data are presented by Moody’s (2019a). It is important to consider what constitutes a default event in these data. Moody’s defines defaults on the basis of one (or more) of four events (see Moody’s 2019b). These include: a missed or delayed disbursement of a contractually obligated interest or principal payment; a bankruptcy filing or legal receivership by the debt issuer or obligor that will likely cause a miss or delay in future debt service payments; and a distressed exchange where an issuer offers creditors a new or restructured debt (or assets) that amount to a diminished value relative to the debt obligation’s original promise, with the effect of allowing the issuer to avoid a likely eventual default. Finally, a change in the payment terms imposed by the sovereign that results in a diminished financial obligation also constitutes a default. Many defaults represent one of the first two instances, although distressed exchanges do periodically account for a sizeable number of total defaults. Simply from inspecting the data, it is readily evident that default rates typically jump at similar times across different sectors which, in principle at least, could be consistent with contagion across sectors (Figure 2). In the data sample examined here, there are three particular episodes where default rates jumped across a wide range of sectors. These three periods corresponded to macroeconomic and financial economic downturns that started in the late 1980s, the early 2000s, and the August 2007 Financial Crisis. These episodes are also clearly evident in Figure 1as spikes in the default rate. It is clear from Figure 2that a spike in the default rate in one sector often occurs at the same time that default rates jump in other corporate sectors. This is also confirmed by simple correlation calculations among different sectoral default rates across the whole data sample. These are presented in Figure 3, where darker shading indicates stronger correlation. However, as noted in the earlier discussion, a coincident jump in default rates could reflect the macroeconomic environment deteriorating and simultaneously impacting on different sectors; this would reflect a common macro shock rather than “contagion” from one sector to another. It is very important to differentiate between co-movement and genuine contagion as discussed previously. This is what the remainder of this paper investigates. Int. J. Financial Stud. 2020,8, 1 8 of 17 Int. J. Financial Stud. 2020, 8, x FOR PEER REVIEW 7 of 17 The default data here were collected and published by Moody’s, a global credit rating agency. Moody’s has collated data on defaults for over a century and publishes default rates back to 1920 on some bases. In this instance, the data sample covered defaults by non-financial corporate issuers as a percentage of Moody’s rated universe. All the data used in this analysis were default rates (percentages) and were not absolute numbers of default events. The data are categorized according to one of 33 corporate sectors and default rates are published on an issuer-weighted basis. The entire data sample covered the period from the start of 1983 to the start of 2017; default rates were reported on a monthly frequency. Overall, the dataset comprised 13,530 observations across the 33 sectors considered here. Towards the start of the sample, the data were more heavily concentrated on US issuers, reflecting Moody’s geographical coverage; but, as Moody’s expanded into other markets, over time, more issuers from other parts of the world— particularly the UK, Germany, and France—were included in the data. More detail on Moody’s default data are presented by Moody’s (2019a). It is important to consider what constitutes a default event in these data. Moody’s defines defaults on the basis of one (or more) of four events (see Moody’s 2019b). These include: a missed or delayed disbursement of a contractually obligated interest or principal payment; a bankruptcy filing or legal receivership by the debt issuer or obligor that will likely cause a miss or delay in future debt service payments; and a distressed exchange where an issuer offers creditors a new or restructured debt (or assets) that amount to a diminished value relative to the debt obligation’s original promise, with the effect of allowing the issuer to avoid a likely eventual default. Finally, a change in the payment terms imposed by the sovereign that results in a diminished financial obligation also constitutes a default. Many defaults represent one of the first two instances, although distressed exchanges do periodically account for a sizeable number of total defaults. Simply from inspecting the data, it is readily evident that default rates typically jump at similar times across different sectors which, in principle at least, could be consistent with contagion across sectors (Figure 2). In the data sample examined here, there are three particular episodes where default rates jumped across a wide range of sectors. These three periods corresponded to macroeconomic and financial economic downturns that started in the late 1980s, the early 2000s, and the August 2007 Financial Crisis. These episodes are also clearly evident in Figure 1 as spikes in the default rate. Figure 2. Corporate default rates across different sectors. Note that the exhibit presents sectoral default rates for 33 non-financial corporate sectors; each sector is shown in a different colour. The vertical axis was deliberately limited to 50% to illustrate volatility within and across sectors relative to lower default rates in some time periods. Source: Moody’s. 0% 5% 10% 15% 20% 25% 30% 1983 1984 1986 1987 1989 1990 1992 1993 1995 1996 1998 1999 2001 2002 2004 2005 2007 2008 2010 2011 2013 2014 2016 Figure 2. Corporate default rates across different sectors. Note that the exhibit presents sectoral default rates for 33 non-financial corporate sectors; each sector is shown in a different colour. The vertical axis was deliberately limited to 50% to illustrate volatility within and across sectors relative to lower default rates in some time periods. Source: Moody’s. Int. J. Financial Stud. 2020,8, 1 15 of 17 omitted variables and time-varying biases. As such, while it is important to note that the correlation analysis presented herein was not as sophisticated as some contagion estimation techniques—which is arguably a limitation of this research—the same issues will also pervade those more complex statistical approaches, unless the bias effects illustrated herein are sufficiently controlled for. Overall, these results suggest that, while default rates often rise across a range of sectors in the face of an economic downturn, reflecting a common underlying shock, corporate bond markets do not present a significant risk in terms of contagion spreading from one sector to another. As such, while investors should be vigilant for macroeconomic shocks that can drive credit distress across a range of sectors, there is little sign that idiosyncratic credit stresses in one sector will drive widespread credit stresses in other sectors. Funding: This research received no external funding. Conflicts of Interest: The author declares no conflict of interest. Appendix A Numerical Listing of Corporate Sectors The main body of this paper uses default rates from 33 different non-financial corporate sectors to test for evidence of contagion. Given the difficulties in presenting 33 different sectoral titles in chart format, for simplicity, each sector is assigned a number in the analysis. Table A1 below presents the sectoral key, listing the individual sectors against the number they are assigned (for instance in Figure 3). Table A1. List and numbering of non-financial corporate sectors. Source: Moody’s and author. S01 Aerospace and Defense S02 Automotive S03 Beverage, Food and Tobacco S04 Capital Equipment S05 Chemicals, Plastics and Rubber S06 Construction and Building S07 Consumer goods: Durable S08 Consumer goods: Non-durable S09 Containers, Packaging and Glass S10 Energy: Electricity S11 Energy: Oil and Gas S12 Environmental Industries S13 FIRE: Finance S14 FIRE: Insurance S15 FIRE: Real Estate S16 Forest Products and Paper S17 Healthcare and Pharmaceuticals S18 High Tech Industries S19 Hotel, Gaming and Leisure S20 Media: Advertising, Printing and Publishing S21 Media: Broadcasting and Subscription S22 Media: Diversified and Production S23 Metals and Mining S24 Retail S25 Services: Business S26 Services: Consumer S27 Telecommunications S28 Transportation: Cargo S29 Transportation: Consumer S30 Utilities: Electric S31 Utilities: Oil and Gas S32 Utilities: Water S33 Wholesale Int. J. Financial Stud. 2020,8, 1 16 of 17 References Aamir, Muhammad, and Syed Ali Shah. 2018. Determinants of Stock Market Co-Movements between Pakistan and Asian Emerging Economies. Journal of Risk and Financial Management 11: 32. [CrossRef] Adad, Pilar, and Helena Chulia. 2013. European Government Bond Markets and Monetary Policy Surprises: Returns, Volatility and Integration. Research Institute of Applied Economics Working Paper No. 2013/25. Barcelona: University of Barcelona. Ahelegbey, Daniel Flix, Monica Billio, and Roberto Casarin. 2016. Bayesian Graphical Models for Structural Vector Autoregressive Processes. Journal of Applied Econometrics 31: 357–86. [CrossRef] Ahrend, Rudiger, and Antoine Goujard. 2014. Are all forms of financial integration equally risky? Asset price contagion during the global financial crisis. Journal of Financial Stability 14: 35–53. [CrossRef] Azizpour, Shahriar, Kay Giesecke, and Gastavo Schwenkler. 2018. Exploring the sources of default clustering. Journal of Financial Economics 129: 154–83. [CrossRef] Billio, Monica, Mila Getmansky, Andrew Lo, and Loriana Pelizzon. 2012. Econometric measures of connectedness and systemic risk in the finance and insurance sectors. Journal of Financial Economics 104: 535–59. [CrossRef] Caporin, Massimiliano, Loriana Pelizzon, Francesco Ravazzolo, and Roberto Rigobon. 2018. Measuring sovereign contagion in Europe. Journal of Financial Stability 34: 150–81. [CrossRef] Das, Sanjiv Ranjan. 2016. Matrix Metrics: Network-Based Systemic Risk Scoring. Journal of Alternative Investments 18: 33–51. [CrossRef] Das, Sanjiv, Darrell Duffie, Nikunj Kapadia, and Leandro Saita. 2007. Common failings: How corporate defaults are correlated. Journal of Finance 62: 93–117. [CrossRef] Diebold, Francis, and Kamil Yılmaz. 2014. On the network topology of variance decompositions: Measuring the connectedness of financial firms. Journal of Econometrics 182: 119–34. [CrossRef] Dornbush, Rudiger, Yung Park, and Stijn Claessens. 2000. Contagion: understanding how it spreads. The World Bank Research Observer 15: 177–97. [CrossRef] Drazen, Allan. 1998. Political Contagion in Currency Crisis. College Park: University of Maryland Mimeo. Eichengreen, Barry, Andrew Rose, and Charles Wypolsz. 1996. Contagious Currency Crises. National Bureau of Economic Research Working Paper No. 5681, July. Cambridge: National Bureau of Economic Research. Ellis, Colin, Haroon Mumtaz, and Pawel Zabczyk. 2014. What Lies Beneath? A Time-varying FAVAR Model for the UK Transmission Mechanism. The Economic Journal 124: 668–99. [CrossRef] Forbes, Kristin, and Roberto Rigobon. 1999. No Contagion, Only Interdependence: Measuring Stock Market Co-Movements. National Bureau of Economic Research Working Paper No. 7267, July. Cambridge: National Bureau of Economic Research. Forbes, Kristin, and Roberto Rigobon. 2001. Measuring Contagion: Conceptual and Empirical Issues. In International Financial Contagion. Edited by Claessens Stijn and Forbes Kirsten. New York: Springer. Forbes, Krinstin, and Roberto Rigobon. 2002. No Contagion, Only Interdependence: Measuring Stock Market Comovements. Journal of Finance 57: 2223–61. [CrossRef] Giudici, Paolo, and Laura Parisi. 2018. CoRisk: Credit Risk Contagion with Correlation Network Models. Risks 6: 95. [CrossRef] Hawkes, Alan. 1971. Spectra of some self-exciting and mutually exciting point processes. Biometrika 58: 83–90. [CrossRef] Herculano, Miguel C. 2018. The Role of Contagion in the Transmission of Financial Stress. European Systemic Risk Board Working Paper No. 81, August. Frankfurt: European Systemic Risk Board. Leschinski, Christian, and Philip Bertram. 2017. Time varying contagion in EMU government bond spreads. Journal of Financial Stability 29: 72–91. [CrossRef] Loretan, Mico, and William English. 2000. ‘Evaluating “correlation breakdowns” during periods of market volatility’, in Bank for International Settlements: International Financial Markets and the Implications for Monetary and Financial Stability. Switzerland: Bank for International Settlements. Masson, Paul. 1997. Monsoonal Effects, Spillovers and Contagion. Washington, DC: IMF Mimeo. Mezei, J ó zsef, and Peter Sarlin. 2018. RiskRank: Measuring interconnected risk. Economic Modelling 68: 41–50. [CrossRef] Moody’s. 2019a. Annual Default Study: Defaults Will Rise Modestly in 2019 Amid Higher Volatility. February. Available online: www.moodys.com (accessed on 23 December 2019). Int. J. Financial Stud. 2020,8, 1 17 of 17 Moody’s. 2019b. Rating Symbols and Definitions. February. Available online: www.moodys.com (accessed on 23 December 2019). Mullainathan, Sendhil. 2002. A Memory-Based Model of Bounded Rationality. The Quarterly Journal of Economics 117: 735–74. [CrossRef] Rigobon, Rpberto. 2016. Contagion, Spillover and Interdependence. European Central Bank Working Paper No. 1975, November. Frankfurt: European Central Bank. Ronn, Ehud. 1998. The Impact of Large Changes in Asset Prices on Intra-Market Correlations in the Stock and Bond Markets. New York: Mimeo. Valdes, R. 1996. Emerging Market Contagion: Evidence and Theory. Cambridge: Massachusetts Institute of Technology Mimeo. © 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).