چشم انداز میزان بازده در بحران بزرگ مالی: شواهد تست نکته سنجی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|24057||2013||22 صفحه PDF||سفارش دهید||22650 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : International Review of Financial Analysis, Volume 26, January 2013, Pages 18–39
We use a simple partial adjustment econometric framework to investigate the effects of financial crises on the dynamic properties of yield spreads. We find that crises manifest themselves in the form of substantial disruptions revealed by changes in the persistence of the shocks to spreads as much as by in their unconditional mean levels. Formal breakpoint tests confirm that in the U.S. the Great Financial Crisis has been over approximately since the Spring of 2009 and provide a conservative dating centered around the August 2007–June 2009 dates. However, some yield spread series point to an end of the most serious disruptions as early as in December 2008. Some symptoms of an impending crisis re-appear instead in the second half of 2011. We also uncover evidence that the LSAP program implemented by the Fed in the U.S. residential mortgage market has been effective, in the sense that the risk premia in this market have been uniquely shielded from the disruptive effects of the crisis
The financial crisis that has allegedly taken place between 2007 and 2009 in the United States has been viewed as the worst financial disruption since the Great Depression of 1929–1933. Many commentators have in fact taken the habit of referring to it as the Great Financial Crisis (henceforth, GFC). The banking crises of the Great Depression involved runs on banks by depositors, whereas the GFC reflected widespread panic in wholesale funding markets that left banks unable to roll over short-term debt. That has deteriorated to engulf most fixed income (FI) markets, both in the US and internationally where persistently high, often historically abnormal yields and yield spreads between different instruments have been observed. The reaction to the crisis by central banks and governments around the world has been massive. It has involved large-scale interventions in both short- and long-term, in private as well as public segments of international bond markets. Although by the end of 2009, a majority of analysts became willing to admit that the worst of the GFC was over, throughout 2010 and 2011 lingering doubts have persisted as to whether the GFC could be dated as a closed, and yet painful event of the recent financial history. Because a number of such interventions have directly involved the segments of the fixed income (FI) markets more severely affected by the GFC, in this paper we take a perspective that is based on yield spread data. A yield spread is the difference between the yield to maturity of a riskier bond and the yield of a comparatively less risky (or riskless) bond. 1 The dimensions of risk that are measured by yield spreads may be many, but they can be grouped as originating from either their default “intensity” (i.e., probability of default and loss given default) or their liquidity risk. In this paper, we ask four related questions: • How can we date a financial crisis, at least on the basis of the yield spread perspective adopted in this paper? This relates to the general question of what properties of yield spreads are affected by a crisis. • In particular, can we date the GFC? Most researchers have been referring to the crisis as a 2007–2009 phenomenon: is this dating as correct as commonly held and/or can we be more precise about its dating as it is usually required of business cycles? • Were the interventions by the Federal Reserve (more generally, by US policy-makers including the Treasury department) effective in fighting the disruptive effects of the crisis? In particular, were the Large Scale Asset Purchases (LSAP) programs announced in late 2008 and implemented between early 2009 and mid-2010 effective and when? • Do any of these questions admit market-specific answers? For instance, are there FI markets that were never affected by the GFC, or for which the crisis seemed to be over well in advance of mid- to late-2009? Similarly, did the European sovereign debt crisis of 2010–2011 revive fears of either a new financial crisis spreading to US markets or of the GFC itself going through a later, deeper phase? In fact, considerable ambiguity and an intense debate has recently concerned the exact dating of the GFC. The conclusions have often reflect the priors of the different researchers as well as their specific methodological approach. Table 1 offers a synopsis of a few among the papers that have appeared in the literature between 2008 and 2011. Although a simple synopsis cannot claim to be exhaustive, we have systematically searched all papers that have investigated the GFC focusing on the behavior of FI yield spreads. Additionally, many criteria could have been used to sort papers in the table and yet—because most of these manuscripts have been repeatedly revised and updated—we have opted for a simple alphabetical sorting. As a Reader may notice most papers had a distinct policy focus as their objective consisted in drawing a connection between (so-called unconventional) monetary policy liquidity measures and their effects on spreads (the fourth column of this table will be used again in subsequent sections). However, the second column concerning the dating of the GFC immediately reveals that while most papers agreed on early August 2007 as a potential starting date of the GFC, few (or none) of them had traced this claim back to the actual features of spread data.2 In fact, in a few cases (e.g., Furceri and Mourougane, 2009 and Sarkar, 2009) several different stages within the GFC could be isolated and discussed. Only a handful of papers have ventured into establishing an end date for the GFC (see e.g., Aït-Sahalia et al., 2009 and Campbell et al., 2011), although this was generally possible only for papers written or revised after the beginning of 2009: usually this consists in generic claims to the effect that the crisis would have been re-absorbed around mid-2009. Eventually, most of the papers established this dating in a casual fashion, often with explicit reference to only one spread series of immediate interest, and without resorting to the typical techniques that applied time series econometricians would use to locate a break-point in time.While the question of precisely dating the GFC remains relevant, it is also important for us to explain the logic of an empirical approach to these issues based on yield spreads, i.e., on bond market-driven estimates of measures of risk premia. There are a number of reasons that can be invoked. A few of them are generally applicable to all research that has focused on FI yield spreads, and others specific to the GFC. First, to filter a financial crisis through the lenses of spread data is implicitly a way to relate financial events to business cycle developments. A feature of U.S. post-WWII business cycle experience that has been widely documented (see e.g., Gilchist et al., 2009 and Guha and Hiris, 2002) is the tendency of a number of FI spreads to widen shortly before the onset of recessions and to narrow again before recoveries. One interpretation of these results is that these credit risk spreads measure the default risk on private (relatively risky) debt (see Stock & Watson, 2003). For instance, Philippon (2009) has proposed a model in which the predictive content of corporate bond spreads for economic activity reflects a decline in economic fundamentals stemming from a reduction in the expected present value of corporate cash flows prior to a downturn. Rising credit spreads may also reflect disruptions in the supply of credit resulting from the worsening in the quality of corporate balance sheets or from the deterioration in the health of financial intermediaries—the financial accelerator mechanism emphasized by Bernanke, Gertler, and Gilchrist (1999). Therefore, the information contained in yield spreads is important because it may be indicative of an important channel through which financial prices affect the real side of the economy. More generally, a better understanding of the dynamics of credit and liquidity risk premia incorporated in the prices of FI products (like term deposits and bonds)—specifically, the asymmetric adjustment process that characterizes turbulent crisis periods from more normal states—has a number of practical implications for investors. When market participants perceive an increase in default risk, they will re-allocate to safer assets and the default risk premium will widen. Our analysis of the changing dynamic properties of a number of commonly reported yield spread series confirms the (possibly obvious) claim that the GFC has been over since the late Spring of 2009 or the Summer of the same, at the latest. Although there is considerable uncertainty as to when exactly the crisis ceased producing its disruptive effects, there is no doubt that after mid-2009 most FI markets have reverted to a normal, pre-crisis state. The financial crisis can be conservatively dated as an August 2007–June 2009 phenomenon, although some yield spread series seem to point to an end of the most serious disruptions as early as December 2008. The LSAP programs implemented by the Fed in the US (agency-supported) residential mortgage market seems to have been considerably effective in the sense that risk premia in this market have been uniquely shielded from the adverse effects of the crisis. Interestingly, this has not occurred in the commercial mortgage market, at least insofar as the private label market for which we have collected data. This in spite of the fact that some of the interventions under the LSAP programs have also specifically targeted the commercial mortgage segment. This may imply that while selective portions of LSAP have produced the desired effects, it may not have been the case across the board. Further bivariate tests reveal that the financial crisis may be characterized as a period in which the yields defining most of the spreads investigated stopped reacting to departures from their (common) “attractor” level in the way they usually did under normal circumstances, always increasing even when the past spread exceeds the long-run attractor yield. On the contrary, in the non-crisis periods and especially in the aftermath of the Great Crisis, we observe that for most spreads, yields tend to adjust in directions—upwards for yields on high (low) default (liquidity) risk bonds, and downward for yields on high (low) default (liquidity) risk bonds—that are compatible with mean-reversion and stationarity of the spreads. Finally, a number of our tests reveal that tensions have recently re-surfaced in a few short-term FI markets between the late Summer of 2011 and the Fall of the same year. For instance, this occurs in the LIBOR and financial commercial paper markets. In the paper we speculate on the causes of these renewed state of turmoil and link it to well-known fears of a possible propagation to the international banking system (therefore also involving US intermediaries) sparked by the re-financing difficulties encountered by a number of European sovereign issuers during the second half of 2011. Two literatures are related to our goals in this paper. One recent literature—partially summarized in Table 1—has debated whether the liquidity facilities and LSAP program implemented by the Federal Reserve have been as effective as the policy-makers had hoped for. On the one hand, several papers have argued that the short-term liquidity programs implemented by the Fed between 2007 and 2008 have been successful. For instance, Adrian, Kimbrough, and Marchioni (2010) have concluded that the Commercial Paper Funding Facility has been successful and that its declining volumes during 2009 were simply caused by its self-liquidating nature. Christensen, Lopez, and Rudebusch (2009) have assessed the effects of the establishment of the liquidity facilities—in particular, of the Term Auction Facility—on the interbank lending market and, in particular, on term LIBOR spreads over Treasury yields. Their multifactor arbitrage-free model of the term structure of interest rates and bank credit risk reveals that the central bank liquidity facilities established in December 2007 helped lower LIBOR rates. Gagnon, Raskin, Remache, and Sack (2011) have used an event study to argue that the LSAP did reduce U.S. long-term yields. On the other hand, several papers have reached opposite conclusions with reference to the credit facilities and the LSAP. For instance, Taylor and Williams (2009) have reported that the TAF was ineffective in significantly influencing the spread between LIBOR rates and overnight lending rates. Thornton (2009) has stressed that until mid-September 2008, the Fed offset the effect of its TAF lending through the liquidity programs on the total supply of credit through open market operations thus reducing their ability to affect financial markets. However, as already discussed, most of these papers have, also in the light of the fast pace of the events between 2008 and 2009, often escaped the task of documenting the exact starting and—if appropriate—end dates of the GFC, even when their focus is predominantly directed at the behavior of interest rate spreads. Using modern break point econometric tools, our paper takes up such a challenge. A second literature has proposed increasingly sophisticated models of the dynamics in yield spreads. For instance, Davies (2008) has analyzed the determinants of US credit spreads over an extensive 85 year sample that covers several business cycles. He finds that econometric models are capable of explaining up to one fifth of the movement in the various spreads considered. This explanatory power derives from autoregressive-type models augmented by relatively small groups of lagged explanatory variables such as changes in riskless interest rates and returns on firms' equities or assets, as in Longstaff and Schwartz (1995).3Papageorgiou and Skinner (2006) have studied corporate credit spreads and the Treasury term structure focussing on the evidence of breakpoints in such relationships. Their results suggest that these relations are not constant but change slowly through time. Compared to this literature, our approach is specifically geared towards our opening questions and therefore based on the simplest available set of econometric tools adequate to develop break tests, i.e., univariate partial adjustment time series models. These models are useful to simultaneously estimate the persistence of the dynamic spread process (i.e., the implied half-life of a shock) and the long-run spread, thus disregarding the relationship between risk spread curves and the risk-free term structure of interest rates. Moreover, our partial adjustment model can be interpreted as a special, restricted AR(2) process and hence it belongs to the simple class of ARMA models. This has the advantage of allowing us to implement a few well-known breakpoint test methodologies such as Chow's (1960) and Andrews (1993). The paper has the following structure. Section 2.1 reviews the unfolding of the GFC and proposes a short list of key episodes used in our commentary. Section 2.2 examines how the yield spreads in seven different bond markets have reacted to these key events. Section 3 presents our econometric methodology. Section 4 contains our main empirical findings. It shows that yield spreads can be described as covariance stationary series, that the parameter estimates of a simple partial adjustment model are subject to considerable instability over time, and formally tests for and finds breakpoints that we interpret as data-driven markers of the onset and the end of the GFC. Section 5 concludes.
نتیجه گیری انگلیسی
This paper has employed simple breakpoint tests applied to univariate and bivariate partial correction models of individual, weekly yield spread series to ask how does a financial crisis affect bond risk (both liquidity and credit) premia and whether it is possible to “date” a financial crisis. Two insights are important and would probably deserve further investigation. First, although most commentaries during the crisis have insisted upon drawing our attention to level of yield spreads as indicators of market disruption, our empirical results show that the crisis has had the power to affect the persistence structure—more precisely, the typical average duration of shocks—of the dynamic process followed by the spreads. In a policy perspective, this means that not only do (bond) risk premia increase during a financial crisis—as everybody would expect—but also that any shock that may cause these premia to depart from their normal levels, is destined to produce long-lived effects. Second, we have uncovered evidence that while one market—the prime (agency-sponsored) fixed-rate residential mortgage market—seems to have escaped the crisis altogether, in a few other FI segments the crisis is not only over, but the dynamics of spreads seems to have already completed reverted to the patterns that have characterized the pre-crisis periods. The finding that some markets may have “under”-shot to the bubble-like conditions of the pre-crisis period may instead provide reason for concern. Yet, other FI markets were found (e.g., the LIBOR, OIS or the 3-month financial commercial paper markets) to affected by renewed tensions starting from the late Summer of 2011. Although our analysis has lacked of sufficient to draw firm conclusions on whether these signals may mark a new (potentially, third) stage of the GFC and not a new, possibly European sovereign-driven crisis, our econometric methodology has thus proven flexible enough to provide a real-time tool to monitor the existence of tensions in US bond markets. Of course, this paper could be extended also in ways that do not involve its methods but instead require additional data. First, a number of papers (e.g., Gilchrist et al., 2009) have not used standard index data to build yield spread series but instead carefully constructed credit and liquidity spread indices for different sectors of economic activity, rating categories, and alternative maturities directly from raw data sets that include individual (corporate) bond prices. It would be important to pursue our question of whether and how a crisis affects the persistence structure of spread dynamics across different risk (rating) classes and over the entire term structure of spreads (see e.g., Ahn, Dieckmann, & Perez, 2009). Second, even if one limits herself to spreads commonly reported in the literature, a number of additional spreads could have been examined in addition to the seven series used in this paper, such as swap rate spreads (vs. Treasuries) or medium-term REFCO liquidity (vs. Treasuries) spreads as in Longstaff et al. (2005), short-term spreads (vs. Treasuries) for adjustable mortgage rates which are popular in the real estate literature (see e.g., Lehnert, Passmore, & Sherlund, 2008), and corporate default spreads that also involve non-quality grade bonds (e.g., a Aa-Bbb junk spread), as in Joutz and Maxwell (2002).