How to Stop a Herd of Running Bears? Market Response to Policy Initiatives during the Global Financial Crisis
Author:
International Monetary Fund
Search for other papers by International Monetary Fund in
Current site
Google Scholar Close

Contributor Notes

Authors’ E-Mail Addresses: yacine@princeton.edu; jandritzky@imf.org; ajobst@imf.org; snowak@imf.org; and ntamirisa@imf.org
Publication Date:
01 Sep 2009
eISBN:
9781451873511
Language:
English
Keywords:
WP; bank bailout; risk premia; monetary policy; recapitalization program
Download PDF (806.4 KB)

This paper examines the impact of macroeconomic and financial sector policy announcements in the United States, the United Kingdom, the euro area, and Japan during the recent crisis on interbank credit and liquidity risk premia. Announcements of interest rate cuts, liquidity support, liability guarantees, and recapitalization were associated with a reduction of interbank risk premia, albeit to a different degree during the subprime and global phases of the crisis. Decisions not to reduce interest rates and bail out individual banks in an ad hoc manner had adverse repercussions, both domestically and abroad. The results are robust to controlling for the surprise content of announcements and using alternative measures of financial distress.

Abstract

This paper examines the impact of macroeconomic and financial sector policy announcements in the United States, the United Kingdom, the euro area, and Japan during the recent crisis on interbank credit and liquidity risk premia. Announcements of interest rate cuts, liquidity support, liability guarantees, and recapitalization were associated with a reduction of interbank risk premia, albeit to a different degree during the subprime and global phases of the crisis. Decisions not to reduce interest rates and bail out individual banks in an ad hoc manner had adverse repercussions, both domestically and abroad. The results are robust to controlling for the surprise content of announcements and using alternative measures of financial distress.

I. Introduction

When the subprime crisis struck in the United States and especially when it spread to other advanced economies and pushed the global economy into recession, designing an effective policy response to the crisis became the number one priority for policymakers around the globe. The ultimate goal of wide-ranging central bank and government interventions was to address the fragility of banking systems and restore confidence in the financial markets. Achieving these goals required a delicate consideration of the sources of stress and the availability of suitable remedies―all against heightened uncertainty about financial and macroeconomic prospects. Reaching consensus on how quick and aggressive policy actions should be, how much weight should be put on macroeconomic and financial sector policies, and what specific form they should take, particularly given various legal, political and other constraints, has been a challenge both at the national and global levels (Swagel, 2009).

The debate on what policy response would be most effective unfolded in real time, and first econometric analyses, mostly using the event study methodology, also appeared. They largely focused on the effectiveness of the Federal Reserve’s Term Auction Facility (TAF), with conflicting results (contrast, for example, Taylor and Williams, 2009; and McAndrews, Sarkar, and Wang, 2008). Some studies underscored the importance of the U.S. Federal Reserve’s commitment to provide unlimited U.S. dollar swap lines to other central banks in alleviating dislocations in the dollar swap markets (Baba and Packer, 2009; and McAndrews, 2009). Announcements of financial restructuring measures were found to have reduced bank credit default swap (CDS) spreads, including for foreign banks, with the magnitude of the impact correlated with the magnitude of resources pledged (Panetta and others, 2009).2 The literature on the effectiveness of crisis policy response has been growing rapidly, with most analyses focusing on individual countries or specific policy measures.

The contribution of this paper is twofold. First, we construct a detailed database of macroeconomic and financial sector policy initiatives announced during the crisis by four systemically important advanced economies―the United States, the United Kingdom, the euro area, and Japan―between June 1, 2007 and March 31, 2009. The database covers announcements in the area of fiscal policy, monetary policy (interest rate decisions, and quantitative and credit easing), liquidity support (in domestic and foreign currency), as well as financial sector policy (recapitalization, asset purchases, liability guarantees, and deposit insurance). It also contains information about decisions not to act (for example, to leave interest rates unchanged or allow banks to fail) and ad hoc bailouts of individual banks.

Second, using a methodology common in the finance literature―the event study―we assess how successful macroeconomic and financial sector policy initiatives (as well as policy inaction) were in addressing the financial sector distress. Our main indicator of financial distress is a widely monitored measure of credit and liquidity risk premia in the global interbank markets―the change in the spread between London Inter-Bank Offered Rates (Libor) and Overnight Index Swaps (OIS) for the U.S. dollar. We employ a variety of parametric and nonparametric means tests to evaluate whether policy announcements had an economically and statistically significant impact on interbank risk premia.

The event study methodology has a number of advantages. The most important are its simplicity, parsimony and focus on the immediate market response to a policy announcement. Compared to alternative methodologies (for example, the regression analysis used, among others, by Taylor and Williams, 2009; and Baba and Packer, 2009), event studies are better designed to work with the limited sample size. Although the basis of policy evaluation is narrow in an event study, it may be suggestive of policies’ long-term effectiveness as a positive immediate market reaction may be self-fulfilling, laying ground for a sustained policy success.

Nonetheless, an event study should not be interpreted as a comprehensive assessment of policy effectiveness. Such an assessment requires correlating measures of policy intensity and objectives over the entire policy horizon, while controlling for the effects of other policies and changes in market conditions. A comprehensive assessment of policies may reach different conclusions from an event study. For example, some policies which markets initially receive negatively (such as bank failures) may ultimately be welfare-enhancing (for example, by avoiding moral hazard) (Klingebiel and others, 2001). A comprehensive assessment of crisis policies appears too early to be conducted at this time.

The event study needs to be designed carefully to address several issues. The first challenge is to create multiple draws of announcements, which we achieve by classifying announcements by type and pooling them across countries. To ensure that the event window for an announcement is not contaminated by the effects of other announcements, we apply a number of filtering criteria when classifying announcements, exclude overlapping announcements and use a narrow five-day event window. We also explore the robustness of results to using a three-day window. We also examine robustness to controlling for expectations about the Libor-OIS spread and content of monetary and fiscal policy announcements (expectations data for other types of policies were not easily available).

Although an event study cannot fully control for the multitude of macroeconomic and structural factors that may affect market response to news, we attempt to address this issue by splitting the sample period into two―the subprime crisis, prior to the collapse of Lehman Brothers, and the global crisis, after the collapse. We also undertake country-specific analyses to gauge the importance of differences in country conditions and characteristics of announcements. To ensure that we are not using a biased measure of financial distress, we test robustness to alternative system-wide measures of interbank risk premia, such as the recently introduced transaction-based New York Funding rate (NYFR)-OIS spread, the expected Libor-OIS spread, the spread of repo transactions to the risk-free rate, and the spread between the Libor rate and the risk-free rate (the TED spread). We also consider a composite measure of bank-specific default risk (credit default swap (CDS) spread), and measures of market perceptions of macroeconomic prospects and financial market volatility (equity price and volatility indices (VIX), respectively).

Based on the literature on past crises, both macroeconomic and financial sector policy announcements are expected to have a significant calming impact on interbank credit and liquidity risk premia (see Reinhart and Rogoff, 2008; Calomiris, Klingebiel, and Laeven, 2005; Claessens, Klingebiel, and Laeven, 2005; and Furfine, 2002). Financial sector policies aim to restore financial stability, while macroeconomic policies help to avoid the vicious feedback between the financial sector and the broader economy. Consistent with McQueen and Roley (1993), we expect market response to announcements to be state-contingent, i.e., depend not only on the surprise content of announcements but also on the state of the economy and financial markets in which investors interpret them. Owing to a high degree of integration of the global financial system, we would also expect to find evidence of international spillovers from policy announcements by systemically important countries.

The announcement effects are likely to vary across types of policies. For example, although both monetary and fiscal easing aim to support activity, the impact of monetary easing on credit and liquidity risk premia is likely to be stronger as it directly relieves funding pressures and reduces counterparty risk. Unconventional policies are likely to show little impact on the spreads themselves, as their system-wide impact is more difficult to assess, although some institutions are likely to benefit significantly. Announcements of liquidity support, even if reducing funding pressures, may have an ambiguous impact if markets are uncertain whether the liquidity provision could mask underlying solvency problems.

The introduction of government guarantees may have a larger immediate effect on interbank risk premia than asset purchases, because guarantees instantaneously transfer risks from banks’ balance sheets to the sovereign. Principles-based bank recapitalization programs are likely to be welcomed by markets, while decisions to bailout financial institutions in an ad hoc manner may have ambiguous effects. Although intended to allay markets’ fears about the stability of individual institutions, they may increase their concerns about the soundness of the overall financial system, as markets may consider that announcements about ad hoc bank bailouts reveal bad news that financial institutions are in trouble and country authorities are privy. Such announcements may raise uncertainty and information asymmetry about counterparty risk, which have been the main cause behind stress in interbank markets (Heider, Hoerova, and Holthausen, 2009).

The findings of this paper underscore that there was no silver bullet for containing the crisis. Both macroeconomic and financial sector policy announcements were associated with reductions in the Libor-OIS spreads, with market responses to announcements depending on the broader context in which market participants were interpreting the news. Three specific results emerge, which are robust to the changes in specification discussed above:

  1. Decisions to keep interest rates stable or raise them had adverse implications during the subprime crisis, while announcements of interest rate cuts contributed positively during the global phase of the crisis. Announcements of domestic and foreign currency liquidity support were mostly associated with reductions in interbank risk premia throughout the crisis, while fiscal policy announcements consistently had negligible effects.

  2. In the financial sector area, announcements of ad hoc bank bailouts had by far the largest impact, aggravating distress in interbank markets during the global phase of the crisis. This impact was not limited to domestic markets but spread throughout the global financial system. By contrast, systematic, rules-based financial restructuring measures were more likely to be associated with a reduction in interbank risk premia.

  3. In particular, recapitalization announcements had a favorable effect on interbank risk premia during the global crisis, suggesting that markets saw merit in these measures. Liability guarantee announcements had mixed effects: reducing interbank risk premia during the subprime crisis (when they largely reflected the U.K. government’s measures in response to the Northern Rock case) and widening interbank risk premia after the crisis deepened (possibly because most announcements at that time were in response to the Irish move and were associated with concerns about possible regulatory arbitrage and disruptive cross-border flows). Announcements of asset purchases were ineffective throughout the crisis, as markets may have anticipated problems in implementing these measures.

The rest of the paper is organized as follows. Section II discusses the Libor-OIS spread as a measure of credit and liquidity risks premia in interbank markets and describes the specially created database on policy announcements. Section III provides a brief overview of the event study methodology and describes how the event study was designed. Section IV discusses graphical evidence and statistical tests of policy announcement effects for the pooled and country-specific samples as well as confirms the robustness of results to alternative specifications. Section V concludes.

II. Measuring Financial Sector Distress and Policy Initiatives

The analysis of interbank market responses to policy announcements requires daily data on a measure of financial distress in interbank markets and on policy announcements. Such measures are discussed below. The section also describes how the time period is split to control for differences in macroeconomic and financial conditions before and after the collapse of Lehman Brothers.

A. The Libor-OIS Spread

We measure the effect of policy announcements on the day-to-day changes in the 3-month U.S. dollar Libor-OIS spread―a proxy for the liquidity and counterparty risk premia in the global interbank markets. The Libor rate comprises the expected risk-free interest rate over a specific term, the term premium, the credit risk premium of unsecured trading with another bank, and the liquidity risk premium of term in lieu of overnight (McAndrews, Sarkar, and Wang, 2008). The OIS rate is a measure of the expected risk-free interest rate over a specific term of secured transactions. Thus, the spread between the Libor and OIS rates over the same term quantifies the premium that banks pay when borrowing funds for a pre-determined period relative to the expected interest cost from repeatedly rolling over funding in the overnight market.

In times of sufficient liquidity and in the absence of market dislocations, the Libor-OIS spread is close to zero. However, when markets are under stress, uncertainty about credit and liquidity risk creates an opportunity cost of term funding, resulting in a positive spread between the Libor and OIS rates. With credit and liquidity risks becoming the major drivers of the increase in the interbank interest rates since the summer of 2007 (Michaud and Upper, 2008), the Libor-OIS spreads turned into a widely monitored indicator of financial distress (Taylor, 2009) and a useful measure of the effectiveness of policy interventions (McCormick, 2007).

When undertaking country-specific analyses, we use the Libor-OIS spreads for the U.S. dollar, pound, euro and yen. Although the Libor-OIS spreads for all major currencies moved closely together during the crisis, they are likely to be reflective of credit and liquidity conditions in interbank markets of the respective countries. Figure 1 plots the daily observations of the levels of and changes in the Libor-OIS spreads from January 1, 2007 to March 31, 2009. Spreads increased sharply in August 2007 (the “black swan” event of Taylor, 2009), remained persistently high through the rest of 2007 and the first half of 2008, and then shot up further in September 2008 after Lehman Brothers’ collapse. The collapse of this systemic institution aggravated distrust in wholesale funding markets, and banks became increasingly reluctant to lend to each other over longer terms. The above periods were associated with the largest day-to-day changes in the spreads.

Figure 1.
Figure 1.

The 3-month Libor-OIS Spreads, in Levels and First Differences, in the United States, United Kingdom, Euro Area, and Japan, January 1, 2007–March 31, 2009

(In basis points)

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Bloomberg and authors’ estimates.Note: Thick lines refer to the level of the Libor-OIS spread and thin lines to first differences. The latter are shown on the right-hand-side axis.

Although the Libor-OIS spread was widely used as a key indicator of financial distress, its reliability can be contested, for example, owing to major dislocations in money markets during the crisis and the role of central banks in displacing financial intermediation in wholesale term markets (Mollenkamp, 2008). Another concern is that the Libor rate tends to be calculated based on banks’ quotes rather than rates used in actual lending transactions.

To confirm that using the Libor rates does not bias the results, we examine robustness to using alternative measures of financial distress (Section IV.B). Instead of the U.S. dollar Libor rate, we consider the recently introduced three-month NYFR, an interdealer-based analog to the London fixing of the unsecured money market in U.S. dollars, which reflects the actual funding cost of banks.3 (Similar data were not easily available for other countries in the sample.) We also consider several alternative system-wide measures of credit and liquidity risks―forward-looking Libor-OIS spreads, repo risk-free spread, and the spread between the Libor rate and the risk-free rate (the TED spread) ―as well as bank-specific measures of default risks (composite bank CDS spreads) and measures of market expectations of macroeconomic prospects and financial stability, such as equity price and volatility indices (VIX).

B. Crisis Timeline

To account for differences in the macroeconomic and financial environment before and after the collapse of Lehman Brothers, we split the crisis period into two subperiods: (i) the subprime crisis from June 1, 2007 to September 14, 2008, which was characterized by a series of predominantly central bank policy measures with a relatively narrow focus on arresting the downward spiral of counterparty confidence; and (ii) the global crisis from September 15, 2008 to March 31, 2009, which witnessed frequent and diverse policy interventions motivated by a sense of heightened urgency about the need to restore financial stability and avoid a global economic depression.

The onset of the crisis is identified as June 1, 2007 based on a Markov-switching vector autoregression of bond market data (see Nowak and others, 2009). The end of the period coincides with the G20 Leaders’ Summit on Financial Markets and the World Economy, held in London on April 2, 2009, when the leaders pledged more than 1 trillion dollars to tackle the global financial crisis by improving international finance and trade, and fostering an economic recovery. The collapse of Lehman Brothers on September 14, 2008 marks the end of the subprime phase and the beginning of the global phase of the crisis.

Differences in the level of the Libor-OIS spread during the precrisis period (January 1 to May 31, 2007) and the two crisis subperiods are apparent from Figure 1. Market responses to policy announcements are likely to have differed during the two subperiods owing to differences in market perceptions of the underlying problems, the nature of the policy effort required to address them, and the policy room country authorities had at their disposal. During the subprime phase, country authorities and markets perceived the crisis as largely limited to the fallout from the collapse of the U.S. subprime mortgage market and its implications for mortgage-backed securities’ markets. Policy priorities largely focused on the unfreezing of credit markets and dealing with weak financial institutions. Interest rate cuts and aggressive provision of liquidity support were seen as ways to address these policy objectives.

The collapse of Lehman Brothers demonstrated the systemic nature of the crisis, at the same time as a sharp deterioration of incoming macroeconomic data pointed to global recession. Aggressive use of monetary and fiscal easing, while appropriate, gradually eroded policy room available to country authorities. Policy priorities during the global phase of the crisis increasingly shifted to restoring market confidence, preventing further systemic bank collapses, and stimulating domestic demand. During that phase of the crisis, interest rates declined sharply, and many central banks shifted to using unconventional monetary policy measures.

C. Policy Announcements

We compile data on major policy initiatives announced by country authorities in the United States, the United Kingdom, the euro area, and Japan in response to the financial crisis. Data for the euro area include policy announcements by the European Central Bank (ECB) and national authorities from Austria, Belgium, France, Germany, Ireland, Italy, the Netherlands, and Spain. Dates of policy announcements are identified based on official press releases, major newspapers and news search engines, and are double-checked against similar compilations of crisis events by central banks, investment banks, international organizations and individual researchers (for example, Federal Reserve Bank of St. Louis 2009; Furceri and Mourougane, 2009; Global Financial Association, 2009; Guillén, 2009; and IMF, 2009a and b).

To minimize noise and the number of overlapping events (which may bias statistical tests), we focus on watershed policy events. Such events are distinguished by the prominence of their media coverage. For the United States and the United Kingdom, we identify major policy events by their appearance as front-page news in the Financial Times and/or the Wall Street Journal within the symmetric four-day window around the official announcement. Given a greater diversity of the economic and financial press in the euro area and Japan, for these economies, we double-check the robustness of the front-page criteria to using alternative news sources, including Bloomberg and Associated Press, and the coverage in Federal Reserve Bank of New York (2009).

Policy initiatives are dated as of their official announcement. For a few measures that involved a multi-stage decision-making process (for example, the adoption of a fiscal stimulus package), the consecutive stages of the process are recorded to the degree the authorities made public announcements at each stage.

We distinguish ten categories of announcements in four policy areas: (i) fiscal measures; (ii) monetary policy; (iii) liquidity support; (iv) financial sector measures; and (v) policy inaction and ad hoc bank bailouts (Table 1):

Table 1.

Classification of Policy Measures

article image
Source: Authors.

Fiscal measures include all policy actions that aim at stimulating domestic demand, through increases in expenditures or reductions in taxes, unless classified in other categories. Since fiscal measures typically require legislative approval, the political decision process is usually protracted with a series of announcements making headlines.

Monetary policy measures include interest rate decisions and quantitative and credit easing, aimed at easing monetary conditions. Quantitative easing involves the central bank’s purchasing government securities, while credit easing consists of purchases of private sector debt in primary or secondary markets, including mortgage-backed securities.

Liquidity support is the provision of domestic currency liquidity through broadened access to central bank refinancing, extended collateral framework, more frequent auctions, or longer maturities, as well as the provision of foreign currency liquidity through swap agreements between central banks and central bank funding facilities for foreign currency liquidity.

Financial sector policies include the tools commonly utilized to resolve systemic banking crises, and are further broken down according to their implications for bank balance sheets:

  • Asset purchase programs use public funds to buy risky assets from banks to shield them from losses. Banks profit from asset purchase programs to the extent that credit risk is removed from their balance sheets, and also because the purchases may put a floor on market prices in banks’ trading books. The category also includes ring-fencing of bad assets, which may be conducted either off-balance sheet through a special purpose vehicle absorbing assets, or on the balance sheet through asset guarantees. Asset purchases usually involve signing a loss-sharing agreement between a public institution providing funds and the bank receiving them. The measure can either be adopted for a single institution or as a system-wide facility for a given asset class.

  • Liability guarantees are system-wide guarantees for newly issued or existing wholesale financing, and the enhancement of deposit protection schemes. It also embraces the lender-of-last resort funding to individual banks and other ways to grant financial institutions access to alternative funding sources, such as the charting of U.S. investment banks as bank holding companies to allow them to tap retail funds.

  • Finally, recapitalization includes the direct injection of capital partially or fully originating from public funds, including the announcement of system-wide recapitalization programs, like the U.K. Bank Recapitalization Fund, and nationalization, which includes the assumption of a controlling stake in a bank.

A special category—policy inaction and ad hoc bailouts—comprises decisions not to enact comprehensive measures to contain the financial crisis. The most prevalent policy announcement in this category is decisions to maintain interest rates unchanged. Interest rate increases, which were rare, are also included. The category also covers decisions to allow banks to fail—Lehman Brothers, IndyMac, and NetBank—and decisions to bailout individual troubled institutions. In contrast to the financial sector measures discussed above, a bank bailout is an ad hoc, unsystematic rescue package or nationalization plan aimed to prevent the failure of an individual financial institution. We use the adoption of a comprehensive recapitalization program or system-wide principles for bank rescues as the criterion for distinguishing between ad hoc bailouts and systematic restructuring measures.

For announcements consisting of several measures, we identify the main measure based on the degree of prominence of front-page coverage for the measures included in the package, which also helps to reduce the number of overlapping announcements. For example, the FOMC’s vote to maintain the interest rate corridor on March 18, 2009, is considered less significant than the same day’s release that the Fed would purchase agency debt and treasury securities for more than one trillion U.S. dollars. In a few cases where several equally important policy initiatives were announced on the same day, they are included as separate entries in the database.

All in all, the database includes 196 front-page announcements (Table 2). Financial sector initiatives accounted for the largest share of front-page announcements (38 percent), followed by policy inaction and ad hoc bank bailouts (23 percent), liquidity support (19 percent), and monetary policy announcements (15 percent). Fiscal policy announcements were just 5 percent of the total. The largest number of front-page announcements covered the policy measures undertaken by the United States (43 percent) and the ECB and the euro area governments (37 percent). The U.K. front-page announcements accounted for 16 percent of the total number of front-page announcements, and Japanese, for only 4 percent, possibly reflecting a smaller direct impact of the crisis on this country with correspondingly fewer crisis-response measures taken.

Table 2.

Number of Front-Page Policy Announcements, June 1, 2007–March 31, 2009

article image
Source: Authors.

Although countries’ approaches to stabilize the financial sector and support domestic growth were broadly similar, the exact timing and characteristics of measures varied depending on authorities’ perceptions of the extent and timing of crisis impact as well as local institutional, structural and political factors. Figure 2 shows the evolution of the front-page policy announcements over time, pointing to a considerable variation in countries’ policy responses to the crisis. With the onset of the subprime crisis, most countries stepped up the provision of liquidity support to financial institutions, but only the United States (and to a lesser degree the United Kingdom) aggressively cut interest rates during that period. The United States also initiated the first fiscal stimulus early on, in January 2008, long before the crisis took on its global dimension, while other countries announced fiscal stimulus packages much later, in the last quarter of 2008.

Figure 2.
Figure 2.

Cumulative Number of Front-Page Policy Announcements June 1, 2007–March 31, 2009

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Authors’ estimates.

Several countries resorted to ad hoc interventions to bail out troubled financial institutions during the subprime crisis, such as the bailout of Bear Stearns in the United States, guarantees to Northern Rock in the United Kingdom, and the rescue of IKB and two Landesbanken in Germany. The United States was the only country that employed a diverse set of financial sector measures early on, ranging from asset purchases to liability guarantees and recapitalization. The U.K. early response to the crisis concentrated on the provision of liability guarantees and changes in deposit insurance schemes, motivated by the need to address shortcomings in the latter. The euro area responded with a large number of recapitalization and liability guarantee measures only after the collapse of Lehman Brothers. In Japan, major financial sector policy announcements focused on recapitalizations.

In addition to dates and types of policy announcements, we record information about the expected and officially announced intensity of some announcements. Data on officially announced intensity are available for fiscal, monetary, and financial sector interventions in the United States and the United Kingdom. For fiscal stimulus packages or asset purchases that imply (quasi-) fiscal outlays, we record the officially announced size of the package or special-purpose budget allocation. Expectations data are available only for fiscal and monetary interventions in these two countries. For fiscal stimuli, market expectations of the magnitude of intervention are determined using news searches in the Financial Times and the Wall Street Journal within one week prior to the official release, and checked against the IMF’s internal data. For policy rate decisions, we use median expectations from Bloomberg‘s surveys of market analysts.

III. Event Study Methodology

We evaluate the real-time impact of policy announcements on interbank credit and liquidity risk premia using the event study methodology. The methodology is well established, especially in the finance literature (see Campbell, Lo, and MacKinlay, 1997; and Kothari and Warner, 2007). The event study methodology is generally well suited to assessing the short-run effects of policy and other announcements. Its main strengths are simplicity and parsimony. Compared to regression-based analyses, event studies are less affected by the problems associated with joint hypothesis testing―the difficulty of distinguishing whether insignificant test results are due to a failure of the efficient market hypothesis or a genuinely insignificant impact of announcements (Jensen, 1978; and Fama, 1991). The event study methodology is also designed to work with the limited samples of announcements we have.

An event study needs to be designed carefully to address several issues. To create multiple draws of similar events, we classify announcements into several types (as discussed in Section II.C) and pool them across countries. (We examine consistency of country-specific analyses with those based on the pooled sample.) It is also important to ensure that the results for a given type of policy announcement are not driven by other events. Applying the front-page and main-event criteria for classifying announcements (see Section II.C) helps to reduce the number of overlapping events. Furthermore, when undertaking the analysis on the pooled sample, exclude domestic and foreign announcements of the same type which fall within five days from each other, making an exception for announcements of support packages where identifying the main event was difficult.4 Country-specific event studies exclude overlapping domestic announcements and tests of bilateral and total spillovers exclude overlapping announcements for the foreign country in question or all foreign countries, respectively.

Limiting the size of the event window helps to avoid contaminating the analysis of given announcement effects with those of preceding and subsequent announcements in an environment where such announcements were made in relatively short succession. We use a narrow event window―one day before and three days after an announcement―while examining the robustness of results to using a symmetric one-day window. A longer post-announcement window allows for a more protracted-than-usual absorption of news, which appears appropriate as many crisis policy initiatives were unprecedented and/or complex, without any apparent benchmarks for evaluating their effects. This interpretation is supported by the graphical analysis (Figure 3), which shows a pronounced accumulation of announcement effects. Besides reducing the contamination of the event window with the effects of other events, limiting the event window also helps to reduce potential endogeneity between policy announcements and market responses.

Figure 3.
Figure 3.

Impact of Policy Announcements on the Libor-OIS Spread, June 1, 2007–March 31, 2009

(In basis points)

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Authors’ estimates.Note: The variable plotted on the vertical axis is the average cumulative abnormal differences in basis points within the event window of one day before the event and three days after the event. The horizontal axis shows days within the event window, with “0” corresponding to the day of the announcement. The subprime phase of the crisis covers June 1, 2007 to September 14, 2008, and the global phase of the crisis covers September 15, 2008 to March 31, 2009.

Another aspect of the identification problem is that policy announcements may affect markets before the event window because they were anticipated. In this case, the policy measures would be priced out before the announcement, rendering insignificant estimates of the announcement effects. To account for such a possibility, we collect measures of the surprise component of interest rate cuts and fiscal stimulus packages in the United States and the United Kingdom (see Section II.C for details) and examine the robustness of results to using only the surprise content of announcements. Expectations data for other countries or types of announcements were not easily available.

The last challenge, which event studies cannot address directly (by design), is controlling for the multitude of factors that may have bearing on market response to announcements. The event study focuses on the credibility of these announcements as well as market participants’ beliefs about their appropriateness, timeliness, feasibility and ultimate implementation. Market perceptions of policy announcements are likely to be state-contingent, depending on how markets perceive the underlying problem that needs to be addressed through policy measures and whether the announced policy measure is timely, appropriate, sufficient, and credible to address this problem. These considerations are likely to have differed considerably before and after the collapse of Lehman Brothers, and the splitting of the crisis period into the two subperiods (see Section II.B for details) helps to control for differences in the macroeconomic and financial conditions to some extent.

Tests of the robustness of results to using alternative measures of financial distress also help to confirm that the main results are not biased. We consider alternative system-wide measures of interbank risk premia such as the recently introduced transaction-based NYFR-OIS rate, expectations about the Libor-OIS spread, the spread of repo transactions to the risk-free rate, and the TED spread. We also consider composite measures of bank-specific default risk (credit default swap spreads), and measures of market perceptions of macroeconomic prospects and financial market volatility, such as equity price and volatility indices (VIX), respectively.

We analyze the impact of policy announcements on the changes, not the level, of the Libor-OIS spread, and capture the cumulative impact of policy announcements over a few days. The event study methodology requires aggregating the abnormal differences in the market indicator of interest within each event window to construct cumulative abnormal differences, under an assumption that no other factors moved the stress indicators during the event window. These differences are then averaged across types of policy to calculate average cumulative abnormal differences (ACAD). Focusing the analysis on short-term changes in the Libor-OIS spread avoids the need to model the time-varying properties of its level, including trends, structural breaks, nonlinearities and nonstationarity.

We define abnormal differences as actual daily changes in the Libor-OIS spread. During both phases of the crises, daily changes of the Libor-OIS spread were stationary and mean-reverting. The means were not statistically different from zero, i.e., expectations of positive and negative changes in the Libor-OIS spread were cancelling each other out. Statistical tests based on an alternative definition of abnormal differences—the difference between the actual daily change on each day of the event window and the expected daily change measured as the average daily change over the previous 20 working days—point to the same conclusions as the baseline results. This is not surprising as the two measures of abnormal differences are highly correlated with each other.

For other financial market indicators, which we consider as part of robustness analysis, the assumption of zero mean reversion is not valid. Abnormal differences are computed as a difference between the expected daily change of the market indicator and its actual daily change. The expected daily change of the market indicator, is estimated as the average daily change over the previous 20 working days, and is subtracted from the actual daily change on each day of the event window to obtain abnormal differences.

We apply parametric and nonparametric tests of means before and after announcements to abnormal differences to ascertain whether the announcement induces a statistically significant effect on interbank risk premia. Parametric tests attribute an equal chance to both positive and negative deviations from expectations (in addition to allowing for asymmetric tail behavior of abnormal changes) while nonparametric tests do so without distributional assumptions. A small number of observations weakens the power of statistical tests, pointing to the need to consider both the economic and statistical significance of results. Statistical tests are specified in line with the literature, for example, Patell (1976); Brown and Warner (1985); Boehmer, Masumeci, and Poulsen (1991); Campbell, Lo and McKinlay (1997); and McKinlay (1997). For a detailed description of test statistics, see Appendix I.

IV. Impact of Policy Announcements on Interbank Credit and Liquidity Risk

To gauge the response of the Libor-OIS spreads to announcements of policy interventions, we first plot changes in the Libor-OIS spreads during the event window and then test the statistical significance of differences in the behavior of the Libor-OIS spreads prior to and after announcements. We start with the graphical analysis and statistical tests on a pooled sample of announcements and the dollar Libor-OIS spread and then proceed to the robustness analysis and country-specific analyses on the Libor-OIS spreads for all major currencies.

A. Analysis on a Pooled Sample

Graphical Analysis

Announcements of policy initiatives had a small calming effect on the distressed financial markets, with the Libor-OIS spread falling by about 3 basis points on average during both crisis periods (Figure 3, top panel). In contrast, policy inaction (defined here as interest rate increases, decisions to keep interest rates stable, ad hoc bank bailouts, and bank failures) were associated with a dramatic increase of the financial market stress during the global phase, yielding an increase of more than 40 basis points on average over the event window for the respective announcements during that phase.

This finding carries through to monetary and financial sector policy measures. The middle panel of Figure 3 underscores the adverse effects of maintaining or increasing policy rates, especially during the global phase. The Libor-OIS spreads widened by about 25 basis points on average over the event window during that phase. Announcements of interest rate cuts are found to be associated with a decline of the Libor-OIS spread. Such announcements were more effective in calming financial markets than announcements about liquidity support, especially when the crisis worsened.

Among the measures targeted directly at banks, announcements concerning recapitalization were associated with the largest reductions in the Libor-OIS spread during the global phase (by about 20 basis points on average over the event window). On the opposite, liability guarantee announcements were even followed by wider spreads during the global phase, as were asset purchases. Ad hoc bank bailouts and bank failures yielded a much worse response than any other policy initiative, with the Libor-OIS spread widening by almost 50 basis points on average over the event window.

Few policy announcements were associated with reductions in interbank risk premia (Figure 4). Although the frequency plots for policy actions and policy inaction and ad hoc bank bailouts are centered at zero (i.e., announcements tended to have a negligible effect on the Libor-OIS spreads), the frequency distribution of policy actions is wider and more skewed to the left. This implies that policy actions reduced the Libor-OIS spreads more often than policy inaction and ad hoc bank bailouts did. A similar analysis by category of policy announcement shows that the frequency distribution for fiscal policy announcements is skewed to the right, i.e., such announcements were rarely associated with an outright reduction of spreads. While a large mass of financial sector announcements is on the right-hand-side of the distribution, the frequency plot has a long left tail, pointing to a few financial sector announcements that did reduce spreads. The left-hand-side tail of the frequency distribution for monetary policy and liquidity support also contains some high-impact events.

Figure 4.
Figure 4.

Frequency Plots for the Pooled Sample of Announcements

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: AuthorsNote: Kernel density estimation with Epanechnikov kernel and linear binning. “Policy action” includes all types of policy announcements, while “policy inaction and fail” covers policy inaction and ad hoc bank bailouts. “Monetary policy” covers monetary policy and liquidity support announcements.

The effect of policy announcements on the Libor-OIS spread varied over the event window during both phases of the crisis (Figure 5). Although some fiscal and financial sector policy announcements were followed by lower Libor-OIS spreads during the subprime crisis, after the collapse of Lehman all types of policy announcements were accompanied by wider Libor-OIS spreads. The adverse impact of policy announcements on the Libor-OIS spreads during that period was rising over the event window, suggesting that markets were not just concerned about the ability of the announced policies to contain the crisis but that these concerns were worsening in the days immediately following policy announcements.

Figure 5.
Figure 5.

Time Profile of the Response of the Libor-OIS Spread to Policy Announcements

(In basis points)

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Authors’ estimates.Note: Average cumulative abnormal differences within the event window of one day before the announcement and three days after the announcement date and aggregated for each quarter over the sample time period (2007 Q3-2009 Q4). Analysis is performed on a pooled sample of all sample country announcements and the average 3-month Libor-OIS spread. Data for fiscal and monetary policy surprises cover the United States and the United Kingdom only. Other data on policy announcements cover, in addition, the euro area and Japan. The red screens indicate the quarter within which the collapse of Lehman Bros. occurred.

Statistical Analysis

Statistical tests confirm the message of the frequency plots that out of a large variety of macroeconomic and financial sector announcements, most were not associated with a significant impact on interbank credit and liquidity risk premia (Table 3 and Figure 6). This finding is in line with the notion that there was no silver bullet for containing the crisis. Like the dynamic plots, statistical tests show that the effects of announcements varied considerably across different types of policies and phases of the crisis.

Figure 6.
Figure 6.

Magnitude and Statistical Significance of the Libor-OIS Spread Response to Policy Announcements (Pooled Sample)

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Authors’ estimates.Note: In basis points. The first bar refers to the subprime crisis and the second to the global crisis. Average cumulative abnormal differences within the event window of 1 day before the announcement and 3 days after the announcement. Statistical tests are performed on a pooled sample of all country announcements and the U.S. dollar 3-month Libor-OIS spread. The bars show the contribution of individual policy announcements to the overall impact of announcements in the respective policy category, as a scaled measure relative to the combined impact of all policy types in the respective group of measures. Asterisks indicate statistical significance at the 10 percent level or less. Data for fiscal and monetary policy surprises cover the United States and the United Kingdom only.
Table 3.

Statistical Tests on a Pooled Sample for Alternative Event Windows and Measures of Surprise

article image
Source: Authors’ estimates. Note: Statistical tests are performed on a pooled sample of policy announcements by all countries and the composite 3-month Libor-OIS spread for all four sample countries. ***,**,* indicate statistical significance (two-tailed, standard normal distribution) at the 1,5 and 10 percent level, respectively. The peak-to-trough change in the Libor-OIS spread was 79.17 basis points (with a daily average change of 0.18 basis points and a standard deviation of 1.02 basis points) during the subprime phase and 141.19 basis points (with daily average change of 0.15 basis points and a standard deviation of 2.80 basis points) during the global phase of the financial crisis. AAD stands for Average Abnormal Difference; ACAD for Average Cumulative Abnormal Difference, and ASIPE for Average Standardized Interval Prediction Error.

Monetary Policy. Interest rate cuts led to significant declines in the Libor-OIS spreads (by about 3½ basis points) during the global phase of the crisis. More countries implemented interest rate cuts during that period than in the early stages of the crisis, the magnitude of the cuts was larger, and on one occasion interest rate cuts were coordinated by major central banks. The decline in the Libor-OIS spreads following interest rate cuts may have reflected markets’ expectation that lower interest rates would increase liquidity in the financial system, thereby reducing liquidity risk in interbank markets.

We do not find evidence that announcements about unconventional monetary policy helped to reduce interbank credit and liquidity risk premia, possibly because the likely effects of these policies were more difficult to discern than those of interest rate decisions. Although some institutions may have been expected to benefit significantly from unconventional policies, the system-wide impact of these measures was uncertain and difficult to assess.

Liquidity Support. Announcements of liquidity support were associated with reductions in interbank risk premia during both phases of the crisis. During the early stage of the crisis, announcements about the provision of U.S. dollar liquidity through swap agreements between the Federal Reserve and other major central banks (which often coincided with announcements about the provision of domestic currency liquidity support) were accompanied by statistically significant but small reductions in the Libor-OIS spreads, in line with the findings by Baba and Packer (2009) and McAndrews (2009). During the global crisis, the Libor-OIS spread declined by about 1 basis point on average following announcements of domestic currency liquidity support, and this effect was statistically significant.

These findings are consistent with the recent studies (for example, Artuç and Demiralp, 2009; McAndrews, Sarkar, and Wang, 2008; Deutsche Bank, 2009; and Christensen, 2009) confirming the effectiveness of the Federal Reserve’s liquidity support (and in contrast to papers by Taylor and Williams, 2009; and Brunetti, Di Filippo, and Harris, 2009). Available evidence seems to suggest that the provision of liquidity support helped to reduce funding pressures for those institutions that had access to liquidity facilities, contributing to lower liquidity risk premia in the interbank markets.

Fiscal Easing. Announcements concerning fiscal stimulus packages were not associated with significant reductions in interbank credit and liquidity risk premia, possibly because fiscal policy measures were seen as mainly targeting other objectives, for example, increasing domestic demand, and as likely to have uncertain and delayed effects on interbank risks.

Financial Restructuring. During the subprime phase of the crisis, announcements of liability guarantees were accompanied by large and statistically significant reductions in interbank credit and liquidity risk premia (by about 5¾ basis points on average). This effect is likely to reflect the large guarantees extended by the U.K. government to depositors of a mortgage lender Northern Rock. The authorities provided stronger support for depositors than the latter anticipated (as manifested by the run on the bank) and also guaranteed wholesale creditors. The support of Northern Rock may have been viewed as crucial for reducing risk premia in global interbank markets, given the role of the United Kingdom as a major money center. Other countries used few liability guarantee measures during the subprime period and these measures were smaller in magnitude (Table 5).

Table 4.

Statistical Tests on a Pooled Sample for Alternative Measures of Financial Risk

article image
Note: “↑” (“↓”) indicates an increase (decrease) of the corresponding market indicator. “―” denotes that no observations were available due to the absence of such policy measure or the late sample starting date (for example, for the New York Funding Rate and the Libor-OIS futures rates). ***,**,* indicated, standard normal distribution) at the 1,5 and 10 percent level based on both parametric and nonparametric tests.

In contrast to LIBOR, which represents a quote of the benchmark rate of interest at which banks can expect to lend funds to each other in the London interbank market, the New York Funding Rate (NYFR) is the representative transaction rate at which an institution would be likely to obtain funding in the market. The NYFR was launched on June 11, 2008 by the inter-dealer brokerage ICAP in response to market concerns about the accuracy of the LIBOR fixing and the panel composition of contributing banks.

The difference between the current Libor-OIS spread and the forward contract on the Libor-OIS spread, 3-month contract. A positive difference between the two indicates elevated liquidity and credit risks in the future compared to the present.

Controlling for expectations over the 20 day pre-event window.

Based on the 3-month government bond yield.

The TED spread is defined as the difference between the LIBOR rate and the risk free rate at a maturity term of 3 months.

VIX is a Volatility Index, created by the Chicago Board Options Exchange as a measure of equity market volatility. The computation of VIX is based on the implied volatility of eight option series on the S&P 100 index, or OEX. VIX is quoted in percent per annum.

See IMF (2009b) for the further details on the CDS and equity composites.

8/ Data on market expectations regarding financial sector policies were insufficient for comparable analysis.
Table 5.

Characteristics and Impact of Financial Sector Policy Measures

article image
source: Authors’ estimates.

Includes guarantees on bond issuance, interbank lending, and other wholesale liabilities (In USD billions) according to the announced package unless specified otherwise. Guarantee on assets of financial institutions are excluded (such as in the case of Japan). For the United States, these cover bank guarantees under the FDIC Temporary Liquidity Guarantee Program (TLGP) for new bank debt and certain checking and other non-interest bearing accounts (up to US$1.45 trillion), the FDIC guarantee to Citigroup (US$10 billion), as well as other guarantees to non-banks and non-financial institutions. However, the Hope for Homeowners Program (up to US$1 billion) and the Temporary Guarantee for Money Market Funds (up to US$3 trillion) are not considered.

Refers to the actual capital committed under announced packages (In USD billions). The estimates for the United States do not include the cost of support measures outside the Troubled Asset Relief Program (TARP) unless specified. So far, US$127 billion of TARP funds (US$700 billion) have been allocated for capital injections. Other capital injections considered include the Capital Purchase Program (up to US$218 billion), the Systemically Significant Failing Institutions Program (US$70 billion),the Targeted Investment Program (US$40 billion) and the conservatorship of Fannie Mae and Freddie Mac (up to US$200 billion). All crisis liquidity facilities and direct lending are excluded: TSLF (up to US$200 bilion), AMLF (US$23.6 billion), TAF (up to US$600 billion), TALF (up to US$1 trillion), CPFF (up to US$1.8 trillion), MMIFF (up to US$540 billion), loan commitments underthe Home Affordable Modification Program (US$50 billion), Unlocking SBA Lending Markets (US$15 billion), Automotive Industry Financing Program (US$30 billion), programs of the National Credit Union Administration (up to US$41 billion), as wellas loans of up to US$60 billion to AIG, US$234 billion to Citigroup, and US$87 billion to Bank of America.

This excludes general asset purchases in capital markets, such as the purchase of bank stocks (Japan). For the United States, US$100 billion of TARP funds have been allocated for asset protection under the Private-Public Investment Program (PPIP). In addition, the following measures are included: GSE MBS and debt purchases (up to US$1.45 trillion), AIG assets (up to US$53 billion), Bear Stearns’ assets (US$29 billion), as well as US$314 billion under the GSE MBS Purchase Program.

“Subprime” (06/01/07-09/13/08), “Global” (09/14/08-3/31/09). The number below the crisis sub-period label show the ratio between the cumul. average abnormal change of the respective type of financial sector policy and the amplitude of the LIBOR-OIS spread over the subperiod. A bolded number indicates statistical significance (parametric and non-parametric) of at least 10 percent during the respective period.

Herfindahl-Hirschman Index (HHI): S(market share of firm)^2 based on the market capitalization of all financial sector firms in each country as of June 7, 2009.

Fitch Ratings (2009) and Moody’s (2009a and 2009b).

These numbers were obtained from the Bank of England (2009).

8/ GDP-weighted composite of Austria, France, Germany, Greece, Ireland, Italy, Netherlands, and pain. Note: Percentage of GDP calculated using WEO projections of GDP in 2008 in US dollars. “—” indicates that the measure was not used. “n.a.” indicates that data is not available.

In contrast, during the global phase, announcements about liability guarantees were not successful in lowering liquidity and credit risk premia in interbank markets. Spreads widened in response to such announcements, possibly owing to the fact that many announcements were implemented in response to the Irish government’s decision to introduce blanket guarantees on assets and were accompanied by concerns that such decisions would result in regulatory arbitrage and outflows of deposits to countries with more generous guarantees and deposit protection. In addition, unlike in the subprime phase, guarantees on impaired assets may have no longer been able to allay markets’ heightened concerns about the underlying soundness of financial institutions.

Announcements concerning recapitalization are associated with a favorable effect on the Libor-OIS spread during the global phase of the crisis (with the spread declining by about 1 basis point on average), although this effect was statistically insignificant. This result supports the argument of Eichengreen and Baldwin (2008), made in early October 2008, that an urgently implemented, globally coordinated recapitalization was the key policy action needed to stabilize the banking system at that time.

News about asset purchase programs were followed by higher credit and liquidity risk premia throughout the crisis, although these effects were statistically insignificant. One possible reason for the more negative impact of announcements about asset purchases than those about liability guarantees and recapitalization is market’s concerns about the potential ineffectiveness of asset purchases―owing to banks’ unwillingness to realize implicit losses out of fear that participation in asset purchase programs would be interpreted as a negative signal about their soundness. Hoshi and Kashyap (2008) point to banks’ stigma derived from participating in the asset purchase program and their reluctance to sell the distressed assets at current market prices (below their fundamental value).

Policy Inaction and Ad Hoc Bank Bailouts. The impact of ad hoc bank bailouts of individual institutions by far outweighed the impact of policy actions during the global phase. Bailouts were accompanied by a 25⅓ basis point increase in the Libor-OIS spread on average in that period. Bailouts tended to send shockwaves through markets as they suggested the possibility of failure of a systemic institution (see, for example, Hoshi and Kashyap, 2008). In most cases, bailouts were carried out with constructive ambiguity about the government’s willingness to provide further support to banks—an intentional difference compared to system-wide recapitalization and liability guarantee programs. Banks may have interpreted bailout announcements as evidence of forthcoming bad news about the soundness of other financial institutions amid increased uncertainty about the government’s intentions of supporting them. As a result, bailouts may have not been able to mitigate asymmetric information and uncertainty about counterparty risk―the main causes of stress in interbank markets (Heider, Hoerova, and Holthausen, 2009).

During the subprime phase of the crisis, the effects of ad hoc bank bailouts were smaller and statistically insignificant, possibly because concerns about counterparty risk were less acute and related more to idiosyncratic reasons such as weak bank management rather than the overall strength of bank balance sheets. However, decisions not to change interest rates or raise them resulted in a small but statistically significant widening of the Libor-OIS spread (by about 1¼ basis points) during the subprime crisis.

B. Robustness Checks

To evaluate the robustness of results, we first consider a narrower event window (three days instead of five). This is a less conservative benchmark for assessing the announcement effects as it gives a higher weight to transitory effects. The narrowing of the event window does not affect the main conclusions, with the results showing a variation typical of event studies (Table 3). As before, liquidity support measures are found to have been associated with lower spreads during both phases of the crisis. Announcements of liquidity guarantees are found to have reduced the Libor-OIS spreads during the subprime phase but not during the global phase. Announcements about policy inaction and ad hoc bank bailouts had large and statistically significant adverse effects on spreads during both phases the crisis, reflecting decisions to maintain stable rates or raise them during the subprime crisis and ad hoc bank bailouts during the global phase. However, the earlier finding interest rate cuts had favorable effects during the global crisis is not longer statistically significant.

The results are also robust to controlling for the surprise content of announcements (see the last panel of Figure 6). Consistent with the earlier findings, surprises in interest rate cuts had a large and statistically effect on interbank risk premia during the global phase of the crisis, while fiscal policy surprises had insignificant effects. Although constrains on the availability and quality of expectations data prevent us from extending the analysis of surprises to other types of announcements, the above analysis for monetary and fiscal policy surprises provides some comfort that using dummy variables does not severely bias the results. Using dummy variables also seems to be appropriate as during a crisis both the expected and unexpected components of policy announcements are likely to have material effects owing to high uncertainty surrounding expectations and rapidly changing policy environment. As McQueen and Roley (1993) show, the effect of macroeconomic releases depends not only on their new informational content but also the state of the economy and financial markets in which investors interpret them. In normal times, policy announcements, for example, interest rate decisions, may be better anticipated (Bernanke and Kuttner, 2005).

Lastly, we confirm that the main results hold across alternative measures of financial distress (Table 4). All system-wide measures of interbank risk premia―the NYFR-OIS rate, the expected Libor-OIS spread, the spread of repo transactions to the risk-free rate, and the TED spread―registered an improvement after announcements of liquidity support, recapitalization programs and interest rate cuts (whether measured as dummy variables or surprises, with the exception of the repo-risk free spread) during the global crisis. The mixed results for liability guarantee are also confirmed by analyses on the above measures of system-wide measures of interbank risk premia. These measures also deteriorated following announcements of decisions to keep interest rates unchanged and bailout individual banks in an ad hoc manner during both phases. The effects of policy announcements on the measures of macroeconomic prospects and financial volatility (the equity price index and VIX) and bank-specific default risks (CDS spreads) were more mixed than those on the system-wide interbank risks, reflecting differences in the nature of these measures. The latter findings are consistent with those by Baba and Parker (2009).

C. Country-Specific Results

Country-specific analyses shed light on the drivers of the pooled results. In this section, we discuss the results of the graphical and statistical analyses on the Libor-OIS spreads for individual currencies and the respective domestic policy announcements. We also examine bilateral and total spillovers from foreign policy announcements.

Graphical Analysis

Plotting responses of the Libor-OIS spread to policy announcements shows a diversity of market responses to crisis-related policy initiatives across countries. Policy efforts to address the crisis rarely reduced the Libor-OIS spread and in some cases even led to a widening of spreads. However, policy inaction and ad hoc bank bailouts had a much worse impact on the Libor-OIS spread (the left panel of Figure 7). This effect was particularly pronounced during the global stage of the crisis and most significant for the United States.

Figure 7.
Figure 7.

Impact of Domestic and Foreign Policy Announcements on the Libor-OIS Spread, June 1, 2007–March 31, 2009

(In basis points)

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Authors’ estimates.Note: The variable plotted on the vertical axis is the average cumulative abnormal differences in basis points within the event window of one day before the event and three days after the event. The horizontal axis shows days within the event window, with “0” corresponding to the day of the announcement. The subprime phase of the crisis covers June 1, 2007 to September 14, 2008, and the global phase of the crisis covers September 15, 2008 to March 31, 2009.

Given the global nature of the crisis and close integration of financial markets, crisis policy initiatives taken by systemic countries are likely to have bearing on market conditions in other countries. Similarly, policy inaction and ad hoc bank bailouts can affect financial risks in other countries. The graphical analysis of spillover effects on the other countries supports this hypothesis (the right panel of Figure 7). The spillover effect was less prevalent during the subprime stage of the crisis, given the small impact of foreign policy initiatives on the U.S. and U.K. Libor-OIS spreads. However, during the global stage of the crisis, international policy spillovers intensified. While foreign policy initiatives generally had few effects beyond the national borders, foreign policy inaction and ad hoc bank bailouts sent shock waves around the globe. The result underscores the strong policy interdependence between major systemic economies particular with regard to monetary inaction and ad hoc or unsystematic rescue packages aimed at single individual financial institutions which might have systemic relevance beyond national borders.

Statistical Analysis

Figures 8–9 and Table 6 report the main findings of statistical tests of domestic and foreign policy announcements on the Libor-OIS spreads for the respective currencies.

Figure 8.
Figure 8.

Magnitude and Statistical Significance of the Libor-OIS Spread Response to Domestic and Foreign Policy Announcements (by Country)

(In basis points)

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Authors’ estimates.Note: Average cumulative abnormal differences within the event window of 1 day before the announcement and 3 days after the announcement. Bars show the contribution of individual policy announcement to the overall impact of the respective policy category. Asterisks on the bars indicate statistical significance at the 10 percent level or less.
Figure 9.
Figure 9.

Magnitude and Statistical Significance of the Libor-OIS Spread Response to Domestic and Foreign Policy Inaction and Ad Hoc Bank Bailouts (by Country)

(In basis points)

Citation: IMF Working Papers 2009, 204; 10.5089/9781451873511.001.A001

Source: Authors’ estimates.Note: Average cumulative abnormal differences within the event window of 1 day before the announcement and 3 days after the announcement. Bars show the contribution of individual policy announcement to the overall impact of the respective policy category. Asterisks on the bars indicate statistical significance at the 10 percent level or less.
Table 6.

Statistical Significance of Foreign Policy Announcements on the Libor-OIS Spreads
Source: Authors’ estimates. Note: Abbreviated country names ("US" for the United States, "UK"for the United Kingdom, "EU" for the euro area, and "JP" for Japan) indicate statistically significant spillover effects of domestic policies on the corresponding country, based on bilateral country analyses. The column headers are the countries where spillover effects from other countries were measured, and the country abbreviations are the countries where these spillovers originated. Italics indicate an increase in the Libor-OIS spreads and the regular font means the opposite. Borders show statistically significant spillover effects from all foreign announcements. Dashed-line borders indicates a reduction of the Libor-OIS spreads, while solid line-borders mean an increase. "o" indicates that spillovers were statistically insignificant. "—" means that spillovers were not feasible, i.e., foreign announcements in the respective policy category did not occur or did not appear on the front pages of major financial newspapers.

Monetary Policy. Aggressive interest rate cuts by the United States during the subprime crisis and by the United Kingdom during the global phase of the crisis were associated with significant declines in the respective Libor-OIS spreads (by about 3 and 5½ basis points, respectively, see Figure 8). Other countries either did not cut interest rates (the euro area and Japan during the subprime crisis) or did so much more gradually, possibly owing to concerns about reaching the zero-interest-rate floor (for example, the United States and Japan during the global crisis) and concerns about price stability and the functioning of the money market (for example, the ECB during the global phase). International spillovers from interest rate cuts were significant during the global phase, with all countries benefiting from the interest rate reductions undertaken by their peers (Figure 8 and Table 6). The Libor-OIS spreads declined most significantly in the United Kingdom in response to foreign interest rate cuts during the global crisis (by about 5½ basis points).

Only U.K. announcements about the use of unconventional monetary policy were associated with reductions in interbank credit and liquidity risk premia domestically and had some small (albeit statistically significant) spillovers to Japan and the euro area (Figure 8 and Table 6). These results may reflect the large scale of unconventional monetary operations pursued by the Bank of England and their narrow focus on purchases of government paper. The Federal Reserve’s unconventional monetary operations, while also sizeable, covered a broad range of instruments, aiming in part to support asset prices, for example, those of mortgage-backed securities. (The Bank of Japan announced limited purchases of equities from banks, while the ECB set up a small program to buy covered bonds.) A broad focus may have made it more difficult for markets to discern the effects of the Federal Reserve’s unconventional policies on interbank liquidity and counterparty risk premia.

Liquidity Support. Announcements of domestic currency liquidity support in the United Kingdom during the subprime crisis and the euro area during the global crisis were associated with significant reductions in interbank credit and liquidity risk (Figure 8). The provision of liquidity support may have helped to reduce funding pressures for those institutions that had access to liquidity facilities, and it may have lowered the liquidity risk premia in the interbank markets (Section IV.A). In other countries, however, at the time of announcement, liquidity support measures did not translate into significantly lower Libor-OIS spreads for the financial system, possibly because markets expected that benefiting institutions would hoard cash or did not consider liquidity problems to be the main reason for bank distress. Further, central bank borrowing carries a stigma that significantly influences banks’ decision to use the liquidity support (see Furfine, 2003).

During the subprime crisis, announcements about the provision of U.S. dollar liquidity through swap agreements between the Federal Reserve and other major central banks in response to global U.S. dollar shortages are found to have resulted in a significant reduction in the Libor-OIS spreads for the United States and the United Kingdom. The U.S. dollar Libor-OIS spread also benefited from spillovers from the euro area’s announcements about forex swaps. Although the finding of strong announcement effects for forex swaps is consistent with other studies (for example, Baba and Packer, 2009; and McAndrews, 2009), it is possible that the results partly reflect the effects of domestic currency liquidity support, as the respective announcements often took place in together (for example, the Federal Reserve’s announcements on December 12, 2007).

Fiscal Easing. Although fiscal policy measures were not targeting reductions in interbank spreads, they may have affected them to the extent that they were interpreted as evidence of political will to support ailing financial institutions. Statistical tests suggest that U.S. announcements of fiscal stimulus during the subprime crisis were not associated with significantly lower Libor-OIS spreads for the U.S. dollar, but had beneficial spillovers to the United Kingdom and the euro area (Table 6).

During the global phase, announcements of fiscal stimulus packages by the euro area helped reduce the Libor-OIS spreads. (The euro area governments announced fiscal stimulus measures later than their peers as they expected automatic stabilizers to play a more significant role during the crisis. Coordination difficulties and the constraints imposed by the European Union’s Stability and Growth Pact also contributed.)

Countries had mutually beneficial spillovers from foreign fiscal policy announcements. Stronger effects of foreign fiscal policy announcements compared to domestic announcements may reflect asymmetric information about the content and effects of domestic and foreign stimulus packages and more optimistic expectations about the success of foreign packages.

Financial Restructuring. Approaches to financial restructuring varied considerably among the four economies in question, reflecting differing perceptions of the degree of bank undercapitalization, potential spillovers from other countries and feasibility and effectiveness of policies in the local institutional and political environment. During the subprime phase, liability guarantees introduced by the United Kingdom helped to reduce credit and liquidity risk premia in that country (by almost 30 basis points on average, a large effect both in absolute terms and in comparison to other policy measures). As pointed out earlier, this effect can be traced to the case of Northern Rock, in which the authorities showed a strong commitment towards depositors. The United States and, to a lesser extent, the euro area benefited from this stance, with interbank spreads for their currencies declining by about 13½ basis points and 5 basis points, respectively (Figure 8 and Table 6). In contrast, during the global phase, announcements about liability guarantees and deposit insurance were not successful in lowering liquidity and credit risk premia in interbank markets (Figure 8).

Announcements concerning recapitalization were not successful in reducing the Libor-OIS spreads during the subprime crisis, possibly because they mostly addressed shortfalls at individual institutions. During the global phase, however, such announcements were associated with reductions in the Libor-OIS spreads in the United States and Japan. This finding underscores the high confidence-building effect of the recapitalization measures undertaken by the United States and Japan, especially considering their relatively small scale (Table 5). This effect may owe to the targeted nature of the recapitalization programs. For example, nationalization of the U.S. quasi-public housing government-sponsored enterprises, which were believed to be at the core of crisis risks, and the focus on banks’ equity holdings in Japan, where banks’ exposure to toxic assets was negligible and the main impact on their balance sheets stemmed from the declining values of equity holdings. The insignificant impact of recapitalization measures in the United Kingdom and the euro area may have reflected greater concerns about the rigor of stress tests used to determine banks’ capital needs in these economies. The euro area, Japan, and the United Kingdom benefited from U.S. announcements of recapitalizations (Table 6). Announcements of asset purchase programs failed to lower credit and liquidity risk premia, both domestically and overseas.

Policy Inaction and Ad Hoc Bank Bailouts. Although policy initiatives had a mixed impact on the Libor-OIS spreads, the announcements of policy inaction and ad hoc bank bailouts were triggering wider spreads, particularly during the global phase of the crisis (Figure 9). Announcements of ad hoc bank bailouts led to the largest widening of the Libor-OIS spreads during the global phase of the crisis, both in domestic markets (by 40½ basis points for the U.S. dollar Libor-OIS spread, 23¼ basis points for the euro spreads, and 22½ basis points for the British pound spreads) and overseas (by about 47 basis points for the U.S. dollar Libor-OIS spread, about 14 basis points for the pound and euro spread and 6 basis points for the yen spread). In particular, U.S. announcements of ad hoc bank bailouts led to significant adverse spillovers into the U.K. and euro area banking systems.

During the global phase of the crisis, international spillovers from announcements of policy inaction and ad hoc bailouts were often occurring in both directions. For example, such announcements by the United States led to a significant widening of the pound and yen Libor-OIS spreads (Table 6). Likewise, announcements of policy inaction and ad hoc bank bailouts by the euro area resulted in statistically significant adverse spillovers on the Libor-OIS spreads for the U.S. dollar.

During the subprime phase of the crisis, the effects of policy inaction and ad hoc bank bailouts stemmed from decisions not to reduce interest rates. These effects were small and caused few spillovers, possibly because concerns about the counterparty risk were less acute during the initial stage of the crisis.

Having reviewed the effects of announcements by policy category, we now summarize them by country, again drawing on Figures 8–9 and Table 6.

United States. Being the country where the crisis originated, the United States front-loaded its policy response to the crisis. It also employed a variety of policy tools early on, announcing policy measures in most categories during the subprime crisis, before it became apparent that U.S. housing problems would significantly affect the financial sectors of other advanced economies. However, with the exception of interest rate cuts and forex swaps, these measures were unsuccessful in reducing short-term credit and liquidity risk premia in the financial system. The U.S. dollar Libor-OIS spreads narrowed upon U.K. announcements of liability guarantee measures and euro area announcements of forex swaps.

During the global phase, the United States continued to announce major policy initiatives in nearly all macroeconomic and financial sector areas. However, only announcements of recapitalization measures were associated with reductions in the U.S. dollar Libor-OIS spread. The U.S. dollar Libor-OIS spread also benefited from foreign announcements of recapitalizations, particularly in the euro area. The effect of bank bailouts domestically and abroad, largely in the euro area, was most notable.

United Kingdom. Like the United States, the United Kingdom introduced a broad range of policy initiatives early on, given its close financial integration with the United States. Announcements of liquidity support, forex swaps and liquidity guarantees were associated with a narrowing of the Libor-OIS spread. The impact of liability guarantees was particularly large compared to that of other measures, owing to their large scale (Table 5). Markets may have favored the use of liability guarantees because, given the large size of the financial sector relative to the economy, recapitalization on a comparable scale would have had a direct impact on the U.K. fiscal position, while liability guarantees represented a contingent public liability. The only beneficial spillover from foreign policy announcements for the United Kingdom was from announcements of fiscal easing and liability guarantees in the United States.

During the global phase the U. K. authorities announced additional major initiatives in fiscal policy, conventional and unconventional monetary policy, liquidity support, and financial restructuring measures (recapitalization, asset purchases, and liability guarantees). Tests suggest that of these, only interest rate cuts lead to a significant outright reduction of credit and liquidity risk premia. Announcements about quantitative easing and asset purchases were also associated with favorable effects. The United Kingdom benefited from interest rate cuts by the euro area and Japan and, like before, announcements of liability guarantees by the United States.

Euro area. Policy response by the euro area was largely back-loaded to the global phase of the crisis. During the subprime phase policy announcements did not result in significant reductions in the Libor-OIS spreads. The euro area benefited, however, from the U.S. announcements of fiscal stimulus and the U.K. announcements of liability guarantees and recapitalization (Table 6). During the global phase, domestic announcements of fiscal easing were associated with reductions in interbank credit and liquidity risk premia.

While foreign bank bailouts had some uneven and, in comparison to other countries, smaller effects, bank bailouts within the euro area increased spreads, albeit less so than similar events in the United States. Announcements of other financial sector measures, particularly recapitalization programs and liability guarantees, were followed by wider spreads, although this result is not statistically significant. Although such a response was found for the United Kingdom too, for the euro area it may reflect a limited integration of the crisis response, with most recapitalization and liability guarantee measures targeted at selected national banks. Insufficient coordination may explain the detrimental effect of the respective announcements on the Libor-OIS spread, triggering a response that is qualitatively similar (albeit smaller) than that of ad hoc bank bailouts.

Japan. None of the policy measures used by Japan during the subprime crisis were front-page events, and foreign announcements of policy initiatives did not have a material impact on the yen Libor-OIS spread during that period. During the global phase of the crisis, announcements of fiscal easing and recapitalization helped reduce credit and liquidity risk premia in the financial system. Japan benefited from foreign policy initiatives in several areas, for example, interest rate cuts and quantitative easing by the United Kingdom, as well as U.S. fiscal easing and recapitalization.

V. Conclusion

An immediate positive market response to announcements of policy initiatives during a financial crisis may be self-fulfilling and indicative of whether these initiatives would help to restore confidence. This paper examined the global crisis through the lens of interbank markets’ immediate response to announcements of macroeconomic and financial sector policy initiatives in systemically important advanced economies. The analysis drew on a unique database of policy announcements and a standard event study methodology, with the findings supported through the gamut of robustness checks. The following broad policy implications emerge from the study:

  1. A systemic financial crisis requires an integrated macro-financial policy response. The paper does not find strong evidence that either macroeconomic or financial policies had an advantage in calming interbank markets. The study does not point to a policy silver bullet that could have been expected to contain the crisis.

  2. A coordinated policy response is required to restore market confidence during a global crisis. The paper finds that foreign policy initiatives had significant bearing on credit and liquidity risk premia in domestic interbank markets. International spillovers of policy announcements intensified as the crisis deepened.

  3. Some policy decisions, for example, decisions not to reduce interest rate, to allow banks to fail, or ad hoc bank bailouts tend to increase credit and liquidity risk premia. This is one of the strongest and most consistent results in the paper. It underscores the importance of a comprehensive approach to financial crisis resolution, as policy actions that are perceived as ad hoc or targeted at individual systemic institutions tend to exacerbate market fears.

Results need to be taken with caution as they do not provide a comprehensive assessment of policy effectiveness. The study focuses on an immediate market response to policy announcements, which tends to have bearing on but may not be always indicative of the long-term effectiveness of policies. It emphasizes the effectiveness of policy announcements in breaking adverse trends in global interbank markets during the financial crisis, which was without doubt one of the most important goals of crisis containment.

References

  • Artuç, Erhan, and Selva Demiralp, 2009, “Provision of Liquidity through the Primary Credit Facility during the Financial Crisis: A Structural Analysis,” FRBNY Economic Policy Review.

    • Search Google Scholar
    • Export Citation

  • Baba, Naohiko, and Frank Packer, 2009, “From Turmoil to Crisis: Dislocations in the FX Swap Market Before and After the Failure of Lehman Brothers,” BIS Working Paper No. 285, Monetary and Economic Department (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Bank of England, 2009, Financial Stability Report, Table 1.C Size of Financial System Support Measures, Issue 25 (June), pp. 20f.

  • Bernanke, Ben S. and Kenneth N. Kuttner, 2005, “What Explains the Stock Market’s Reaction to Federal Reserve Policy?” Journal of Finance, Vol. 60 (June), pp. 1221–57.

    • Search Google Scholar
    • Export Citation

  • Boehmer, Ekkehart, Jim Masumeci, and Annette B. Poulsen, 1991, “Event Study Methodology under Conditions of Event Induced Variance,” Journal of Financial Economics, Vol. 30 (December), pp. 253–72.

    • Search Google Scholar
    • Export Citation

  • Brown, Stephen. J., and Jerold. B. Warner, 1980, “Measuring Security Price Performance,” Journal of Financial Economics, Vol. 8 (March), pp. 3–31.

    • Search Google Scholar
    • Export Citation

  • Brown, Stephen. J., and Jerold. B. Warner, 1985, “Using Daily Stock Returns: The Case of Event Studies,” Journal of Financial Economics, Vol. 14 (September), pp. 205–58.

    • Search Google Scholar
    • Export Citation

  • Brunetti, Celso, Mario Di Filippo, and Jeffrey H. Harris, 2009, “Effects of Central Bank Intervention on the Interbank Market During the Sub-Prime Crisis,” Available via the Internet: http://ssrn.com/abstract=1428662.

    • Search Google Scholar
    • Export Citation

  • Calomiris, Charles, Daniela Klingebiel, and Luc Laeven, 2005, “Financial Crisis Policies and Resolution Mechanisms: A Taxonomy from Cross-Country Experience,” in Systemic Financial Distress: Containment and Resolution, ed. by Patrick Honohan and Luc Laeven (Cambridge: Cambridge University Press).

    • Search Google Scholar
    • Export Citation

  • Campbell, John Y., Andrew W. Lo, and A. Craig McKinlay, 1997, The Econometrics of Financial Markets (Princeton: Princeton University Press).

    • Search Google Scholar
    • Export Citation

  • Christensen, Jens, 2009, “Have the Fed Liquidity Facilities Had an Effect on Libor?” FRBSF Economic Letter, No. 25, August 10.

  • Claessens, Stijn, Daniela Klingebiel, and Luc Laeven, 2005, “Crisis Resolution, Policies, and Institutions: Empirical Evidence,” in Systemic Financial Distress: Containment and Resolution, ed. by Patrick Honohan and Luc Laeven (Cambridge: Cambridge University Press).

    • Search Google Scholar
    • Export Citation

  • Corrado, Charles, J., and Terry L. Zivney, 1992, “The Specification and Power of the Sign Test in Event Study Hypothesis Tests Using Daily Stock Returns,” Journal of Financial and Quantitative Analysis, Vol. 27 (September), pp. 465–78.

    • Search Google Scholar
    • Export Citation

  • Deutsche Bank Securities Inc. (Deutsche Bank), 2009, “Are the Fed’s Programs Working?” Deutsche Bank Global Markets Research, Global Economic Perspectives, March 4.

    • Search Google Scholar
    • Export Citation

  • Eichengreen, Barry, and Richard Baldwin, eds., 2008, Rescuing Our Jobs and Savings: What G7/8 Leaders Can Do to Solve the Global Credit Crisis (London: VoxEU.org ate Centre for Economic Policy Research). Available via the Internet:http://www.voxeu.org/index.php?q=node/2545.

    • Search Google Scholar
    • Export Citation

  • Fama, Eugene F. (1991), “Efficient Capital Markets: II,” Journal of Finance, 46, pp.1575–1617.

  • Federal Reserve Bank of New York, 2009, “Timelines of Policy Responses to the Global Financial Crisis.” Available via the Internet:http://www.ny.frb.org/research/global_economy/policyresponses.html.

    • Search Google Scholar
    • Export Citation

  • Federal Reserve Bank of St. Louis, 2009, “The Financial Crisis: A Timeline of Events and Policy Actions.” Available via the Internet: http://timeline.stlouisfed.org.

    • Search Google Scholar
    • Export Citation

  • Fitch Ratings, 2009, High-Grade Sovereigns and the Global Financial Crisis. Special Report, March (New York).

  • Frank, Nathaniel, and Heiko Hesse, forthcoming, “The Effectiveness of Central Bank Interventions during the First Phase of the Subprime Crisis,” forthcoming IMF Working Paper (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Furceri, Davide, and Annabelle Mourougane, 2009, “Financial Crises: Past Lessons and Policy Implications,” OECD Economics Department Working Paper No. 668 (Paris: Organization for Economic Co-Operation and Development).

    • Search Google Scholar
    • Export Citation

  • Furfine, Craig, 2002, “The Interbank Market during a Crisis,” European Economic Review, Vol. 46 (May), pp. 809–20.

  • Furfine, Craig, 2003, “Standing Facilities and Interbank Borrowing: Evidence from the Federal Reserve’s New Discount Window,” International Finance, Vol. 6 (November), pp. 329–47.

    • Search Google Scholar
    • Export Citation

  • Global Financial Association, 2009, “Global Financial Crisis Timeline.” Available via the Internet: http://www.glofin.org/blog.

  • Guillén, Mauro, 2009, “The Global Economic and Financial Crisis: A Timeline.” Available via the Internet: http://lauder.wharton.upenn.edu.

    • Search Google Scholar
    • Export Citation

  • Guo, Lin, and Timothy S. Mech, 2000, “Conditional Event Studies, Anticipation, and Asymmetric Information: the Case of Seasoned Equity Issues and Pre-issue Information Releases,” Journal of Empirical Finance, Vol. 7 (August), pp. 113–41.

    • Search Google Scholar
    • Export Citation

  • Heider, Florian, Marie Hoerova, and Cornelia Holthausen, 2009, “Liquidity Hoarding and Interbank Market Spreads: The Role of Counterparty Risk,” mimeo, European Central Bank, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1362113.

    • Search Google Scholar
    • Export Citation

  • Hoshi, Takeo, and Anil K. Kashyap, 2008, “Will the U.S. Bank Recapitalization Succeed? Lessons From Japan,” NBER Working Paper No. 14401 (Cambridge: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • International Monetary Fund, 2009a, “Note by the Staff of the International Monetary Fund on Stocktaking of the G-20 Responses to the Global Banking Crisis,” Group of Twenty—March 13–14 (Washington).

    • Search Google Scholar
    • Export Citation

  • International Monetary Fund, 2009b, “Market Interventions During the Financial Crisis: How Effective and How to Disengage,” in Global Financial Stability Report, World Economic and Financial Surveys, October (Washington).

    • Search Google Scholar
    • Export Citation

  • Jensen, Michael C., 1978, “Some Anomalous Evidence Regarding Market Efficiency,” Journal of Financial Economics, 6, pp.95-101.

  • Kothari, S.P., and Jerold B. Warner, 2007, “Econometrics of Event Studies,” in Handbook of Corporate Finance, Vol. 1, ed. by B. E. Eckbo (Amsterdam: Elsevier/North-Holland).

    • Search Google Scholar
    • Export Citation

  • Klingebiel, Daniela, Randy Kroszner, Luc Laeven, and Pieter van Oijen, 2001, “Stock Market Responses to Bank Restructuring Policies during the East Asian Crisis,” World Bank Policy Research Working Paper No. 2571 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

  • McAndrews, James, Asani Sarkar, and Zhenyu Wang, 2008, “The Effect of the Term Auction Facility on the London Inter-Bank Offered Rate,” Staff Report No. 335 (New York: Federal Reserve Bank of New York).

    • Search Google Scholar
    • Export Citation

  • McAndrews, James, 2009, “Segmentation in the U.S. Dollar Money Markets during the Financial Crisis,” a paper presented at the International Conference of the Bank of Japan’s Institute for Monetary and Economic Studies, Tokyo, Japan, May.

    • Search Google Scholar
    • Export Citation

  • McCormick, Liz C., 2007, “Interest-Rate Contracts May Be Best Gauge of Fed Plan (Update2)”, Bloomberg, December 17. Available via the Internet: http://bloomberg.com.

    • Search Google Scholar
    • Export Citation

  • McKinlay, A. Craig, 1997, “Event Studies in Economics and Finance,” Journal of Economic Literature, Vol. 35 (March), pp. 13–39.

  • McQueen, Grant, and Vance V. Roley, 1993, “Stock Prices, News, and Business Conditions,” Review of Financial Studies, Vol. 6 (September), pp. 683–707.

    • Search Google Scholar
    • Export Citation

  • Meier, André, 2009, “Panacea, Curse, or Nonevent? Unconventional Monetary Policy in the United Kingdom,” IMF Working Paper No. 163/09 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Michaud, Francois-Louis, and Christian Upper, 2008, “What Drives Interest Rates? Evidence from the Libor Panel,” BIS Quarterly Review (March), pp. 47–58.

    • Search Google Scholar
    • Export Citation

  • Mikkelson, Wayne H., and Megan M. Partch, 1986, “Valuation Effects of Security Offerings and the Issuance Process,” Journal of Financial Economics, Vol. 15 (January/February), pp. 31–60.

    • Search Google Scholar
    • Export Citation

  • Mollenkamp, Carrick, 2008, “Bankers Cast Doubt on Key Rate Amid Crisis,” Wall Street Journal, April 15.

  • Moody’s, 2009a, “How Far Can Aaa Governments Stretch Their Balance Sheets? Stress-Testing Governments’ Debt Affordability Through the Crisis,” Investor Services (February).

    • Search Google Scholar
    • Export Citation

  • Moody’s, 2009b, “High-Grade Sovereigns and the Global Financial Crisis,” Investor Services (March).

  • Nowak, Sylwia, Jochen R. Andritzky, Andreas Jobst, and Natalia T. Tamirisa, 2009, “Macroeconomic Fundamentals, Price Discovery and Volatility Dynamics in Emerging Markets,” IMF Working Paper No.09/147 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Panetta, Fabio, Thomas Faeh, Giuseppe Grande, Corrinne Ho, Michael King, Aviram Levy, Federico M Signoretti, Marco Taboga, and Andrea Zaghini, 2009, “An Assessment of Financial Sector Rescue Programmes,” BIS Paper No. 48, Monetary and Economic Department (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

  • Patell, James M., 1976, “Corporate Forecasts of Earnings per Share and Stock Price Behavior: Empirical Tests,” Journal of Accounting Research, Vol. 14 (Autumn), pp. 246–76.

    • Search Google Scholar
    • Export Citation

  • Reinhart, Carmen M., and Kenneth Rogoff, 2008, This Time is Different: Eight Centuries of Financial Folly (Princeton: Princeton University Press).

    • Search Google Scholar
    • Export Citation

  • Swagel, Phillip, 2009, “The Financial Crisis: An Inside View,” Brookings Papers on Economic Activity, Spring Conference draft.

  • Taylor, John B., 2009, Getting Off Track: How Government Actions and Interventions Caused, Prolonged, and Worsened the Financial Crisis (Stanford: Hoover Press).

    • Search Google Scholar
    • Export Citation

  • Taylor, John B., and John C. Williams, 2009, “A Black Swan in the Money Market,” American Economic Journal: Macroeconomics, Vol. 1 (January), pp. 58–83.

    • Search Google Scholar
    • Export Citation

Appendix I. Statistical Tests

A. Parametric Tests

The intuition underlying the design of parametric tests is as follows. We define a parametric test statistic as the ratio between average cumulative abnormal differences (ACAD) and three different measures of historical volatility of the Libor-OIS spread over an estimation period of 20 days before the event window: (i) the square root of the sum of squared differences; (ii) the standard deviation of simple differences; and (iii) the expected prediction error (derived from a simple autoregressive process) adjusted by the ratio between the volatility during both the estimation and event periods. As opposed to the first two (conventional) measures of volatility (Mikkelson and Partch 1986), the alternative specification of volatility in (iii) accounts for changes in volatility on a day-to-day basis relative to the empirical experience within a short event window. The last specification is particularly relevant for cases when a policy measure was anticipated by markets and had an effect on the Libor-OIS spread before the event window.

The standardized abnormal changes derived from these parametric measures are tested for convergence to both the standard normal and an extreme value distribution calibrated to the higher moments of daily changes during the estimation window. The latter null hypothesis of expected changes of abnormal returns restricts statistical significance only to those observations that are truly exceptional even at a time of large changes in the market indicator. More precisely, the parametric ACAD test statistic, which shows the average daily response of the tested market indicator, is defined as follows:

ACAD i , τ , m σ i , l , m ( ACAD ) = 1 N m T ∑ i ∈ N m ∑ τ ∈ T ( x i , τ − x i , τ − 1 ) 1 N m 2 ∑ i ∈ N m ∑ l ∈ L ( x i , l − x i , l − 1 ) 2 L − 2 = 1 N m T ∑ i ∈ N m ∑ τ ∈ T AD i , τ 1 N m 2 ∑ i ∈ N m ∑ l ∈ L AD i , l 2 L − 2 ∼ { Φ ( 0 , 1 ) G ( 0 , σ x , ξ x ) , ( 1 )

for event window of length T days, event time indicator τ ∈ [−1,3], sample observations N of policy event category m, a pre-event (estimation) window L of 20 days (and date indicator l∈[τ−20,τ−1]), and convergence to the standard normal distribution Φ(.).5 Abnormal differences are generally defined as the daily changes ADi,τ,k = xi,τ,k−xi,τ−1,k of an indicator variable x in response to event i of category m on event time day τ.6 Abnormal differences are averaged across types of policy measures to obtain average cumulative abnormal differences for each country or region during pre-crisis, subprime and global crisis periods.

We also consider an alternative specification of volatility in (1):

ACAD i , τ , m σ i , τ , m ( ACAD ) = 1 N m ∑ i ∈ N m ∑ τ ∈ T AD i , τ T N m ∑ i ∈ N m ∑ l ∈ L ( AD i , l − AD ¯ i , l ) 2 L = 1 T ∑ i ∈ N m ∑ τ ∈ T AD i , τ ∑ i ∈ N m ∑ l ∈ L ( AD i , l − D ¯ i , l ) 2 L ∼ { Φ ( 0 , 1 ) G ( 0 , σ x , ξ x ) ( 2 )

by calculating a rescaled measure of the average standard deviation in order to determine the statistical significance of the cumulative daily price response over the event window conditional on a more balanced specification of volatility during the estimation window. Lastly, we condition the individual abnormal differences on a measure of adjusted standard deviation subject to the relative change of volatility during both the estimation and event periods. We first derive the measure of standard deviation

σ i , τ , l , m = var ( ε i , l ) ( 1 + 1 L + ( AD i , τ − AD ¯ i , l ) 2 ∑ l ∈ L ( AD i , l − AD ¯ i , l ) 2 ) ∼ student − t ( 0 , L L − 2 ) ( 3 )

from the ordinary prediction errors εi,j, of the AR(1) process of the selected market variable (the Libor-OIS spread) at first differences subject to the ratio of the squared abnormal change of the market variable at event time τ and the volatility of abnormal differences over the 20-day estimation window. Following Brown and Warner (1985), we obtain the (daily) standardized prediction error (SPE)

SPE i , τ , l , m = 1 T + 1 ∑ τ ∈ T AD i , τ σ i , τ , m ( 4 )

and the average standardized interval prediction error (ASIPE)

N m N m ∑ i ∈ N m SPE i , τ , m ∼ { Φ ( 0 , 1 ) G ( 0 , σ x , ξ x ) ( 5 )

as a test statistic.

B. Nonparametric Tests

Given the weak power of parametric tests in small samples of events with diverse effects on market indicators, we use nonparametric tests to remove priors about the distribution of abnormal differences from the measure of statistical significance. Based on Corrado and Zivney (1992), a standard sign test determines whether the incidence of positive or negative responses to a particular type of policy measure is statistically significant under normality:

( N m + N m − 0.5 ) N m 0.5 ∼ Φ ( 0 , 1 ) , ( 6 )

where N+ denotes the number of cumulative abnormal differences (CAD) that exhibit the desired direction of market response every time a policy of type m is announced. However, such a test statistic captures only the dominant direction of market response and ignores the relative magnitude of market response in either direction.

To control for such relative market response, we introduce the sign-size test. It provides for a high degree of asymmetry in the magnitude of individual market responses to different types of announcements. In this test a given announcement type would be deemed statistically significant only if there is a positive market response (in the sense of a decline in the Libor-OIS spread), which, on average, is larger than any negative response to the same type of policy measure over the sample time period.

We control for the size of standardized CAD^, which transforms equation (6) into the test statistic of the sign-size test:

( 1 N m + ∑ i ∈ N m + ( k m − CAD ^ i + ) 1 N m − ∑ i ∈ N m − ( CAD ^ i − − k m ) + 1 N m + ∑ i ∈ N m + ( k m − CAD ^ i + ) N m + N m − 0.25 ) N m 0.5 = ( ∑ i ∈ N m + ( k m − CAD ^ i + ) N m + N m N m − ∑ i ∈ N m − ( CAD ^ i − − k m ) + N m ∑ i ∈ N m + ( k m − CAD ^ i + ) − 0.25 ) N m 0.5 ∼ Φ ( 0 , 1 ) , ( 7 )

where

CAD ^ i = | CAD | − min N ( CAD i ) | max N ( CAD i ) | + | min N ( CAD i ) | ∈ [ 0 , 1 ] ( 8 )

for the standardized CAD^i+ and CAD^i− of all individual counts of positive (i.e., Libor-OIS spread reducing) and negative (i.e., Libor-OIS spread increasing) market responses to a particular policy measure m, and the threshold value

k m = { | min N m ( CAD i ) | | max N m ( CAD i ) | + | min N m ( CAD i ) | if min N m ( CAD i ) < 0 < max N m ( CAD i ) 1 if min N m ( CA D i ) < max N m ( CAD i ) < 0 , else 0 . ( 9 )

Which determines at which point the standardized CAD^ indicates a postivie of negative market response.

We apply the above statistical tests to individual types and categories of policy announcements. In addition, we examine the contribution of each type to the overall impact of the respective broad policy category M, such as monetary policy, financial sector support or policy inaction and ad hoc bank bailouts. For example, we calculate contribution of recapitalization announcements to the overall impact of announcements about financial sector measures. Since the tests are performed separately for each type and category of policy announcement, the analysis of contributions requires scaling the impact of individual types of announcements, especially if the constituent policy measures triggers opposite market responses on average. The scaled absolute contribution of an individual type of policy announcement m to the impact of the respective policy category is defined as

ACAD m ⌢ = N m N M ACAD m ∑ M ACAD m ∑ M ACAD m + i f ∑ M ACAD m > 0 , ( 10 )

where ∑MDACAD+ and ∑MACADm− denote the sum of all positive and negative ACADs associated with the individual types of policy measures that make up the broader policy category M.
1

Corresponding author: Natalia Tamirisa, ntamirisa@imf.org. We are grateful to Prakash Loungani for encouragement and helpful discussions on the project, and to Tam Bayoumi, Andy Berg, Olivier Blanchard, Stijn Claessens, Giovanni Dell’Arricia, Atish R. Ghosh, Thomas Harjes, Laura Kodres, Manfred Kremer, Luc Laeven, Andre Meier, Gian Maria Milesi-Ferretti, Jonathan D. Ostry, Alex Popov, Effie Psalida, David Romer, Emil Stavrev, Mark Stone, and participants in the Research Department’s seminars for insightful comments. We thank Wouter Elsenburg for sharing the CDS spreads dataset and Sumit Aneja, Tiffany Coln, Sarah Mogab, Carmela Pedicini, Jair Rodriguez, Sheila Tomilloso Igcasenza, and David Velazquez-Romero for providing excellent research assistance.

2

Other studies assessing the impact of policy interventions during the current crisis include Artuç and Demiralp (2009), Meier (2009), Frank and Hesse (forthcoming), and IMF (2009a and b).

3

In contrast to LIBOR, which represents a quote of the benchmark rate of interest at which banks can expect to lend funds to each other in the London interbank market, the NYFR is the representative transaction rate at which an institution would be likely to obtain funding in the market. The NYFR was launched on June 11, 2008 by the inter-dealer brokerage ICAP in response to market concerns about the accuracy of the LIBOR fixing and the panel composition of contributing banks.

4

They include 12 events in the United States (on 12/19/2007, 7/13/2008, 9/15–17/2008, 10/03/2008, 10/14/2008, 11/10/2008, 11/25/2008, 1/16/2009, and 1/28/2009), 6 events in the United Kingdom (on 7/05/2007, 10/4/2007, 11/08/2007, 10/03/2008, 01/19/2009, and 02/26/2009), 7 events in the euro area (on 2/13/2008, 9/29/2008, 10/07/2008, 10/12–15/2008, and 10/20/2008), and 2 events in Japan (on 7/15/2008 and 10/31/2008).

5

For parametric tests, we also examine statistical significance based on an alternative distributional assumption that accounts for excess skewness and kurtosis of abnormal differences. We calibrate a generalized extreme value distribution (GEV) as a means to capture asymptotic tail behavior at percentiles far removed from central observations. GEV prescribes that there exists a choice of constants an > 0 and bn, such that the probability of a n-sequence of normalized maxima Fn[anx+bn](x) converges to the non-degenerate limit distribution G(x)=exp(−(1+ξ(x−μ)/σ)+−1/ξ) as n →∞ and x∈ℝ, after adjustment of scale parameter σ>0, location parameter μ and the shape parameter ξ, which indicates the velocity of asymptotic tail decay of a given limit distribution (“tail thickness”). We calibrate G(x) for μ=0

6

As discussed in the preceding section, averages during the pre-event window are zero for changes in the Libor-OIS spread and nonzero for other indicators.

Same Series

Other IMF Content

Other Publishers