Measuring changes in credit risk: The case of CDS event studies

This paper examines the size and power of test statistics designed to detect abnormal changes in credit risk as measured by credit default swap (CDS) spreads. We follow a simulation approach to examine the statistical properties of normal and abnormal CDS spread changes and assess the performance of normal return models and test statistics. Using daily CDS data, we find parametric test statistics to be generally inferior to non-parametric tests, with the rank test performing best. A CDS factor model based on factors identified in the empirical literature is generally well specified and more powerful in detecting abnormal performance than some of the classical normal return models. Finally, we examine abnormal CDS announcement spread changes around issuer's rating downgrades to demonstrate the effect of different CDS spread change measures and normal return models on event study results.     

Bond market event study methods

The procedures used in corporate bond event studies to date fail to control for heteroskedasticity due to differences in return volatility by term-to-maturity, rating, and other factors resulting in low test power. Bond return standardization yields considerably more powerful tests. Also, due to infrequent trading, use of bond transaction price observations over several days before and after an event, while giving more weight to returns calculated from transactions closer to the event, yields considerably more powerful tests than returns based solely on transactions the day before and the day after the event. Exploring the test bias caused by overlapping event dates, we find that, adjusted for rating and maturity, the correlation among standardized abnormal bond returns is small but that even fairly small correlations can result in biased test statistics. A bond market modification of the Kolari and Pynnönen (2010) procedure corrects this bias.     

Nonparametric rank tests for event studies

Because stock prices are not normally distributed, the power of nonparametric rank tests dominate parametric tests in event study analyses of abnormal returns on a single day. However, problems arise in the application of nonparametric tests to multiple day analyses of cumulative abnormal returns (CARs) that have caused researchers to normally rely upon parametric tests. In an effort to overcome this shortfall, this paper proposes a generalized rank (GRANK) testing procedure that can be used on both single day and cumulative abnormal returns. Asymptotic distributions of the associated test statistics are derived, and their empirical properties are studied with simulations of CRSP returns. The results show that the proposed GRANK procedure outperforms previous rank tests of CARs and is robust to abnormal return serial correlation and event-induced volatility. Moreover, the GRANK procedure exhibits superior empirical power relative to popular parametric tests.     

CONSISTENT ESTIMATION OF RESIDUAL VARIANCE IN REGULATORY EVENT STUDIES

This study presents new evidence on alternative methods used to test for abnormal returns in regulatory event studies where cross-sectional correlation in residuals is significant. Results contradict earlier studies that find no advantages to using joint generalized least squares (JGLS) methods over ordinary least squares (OLS). We find that in an actual regulatory event study cross-correlation is significant, and that failing to correct for this correlation results in substantially higher calculated F-statistics. In Monte Carlo simulations we find that OLS test statistics are not well specified when residuals exhibit cross-sectional correlation at levels that are reasonable to expect in daily return data, while JGLS test statistics are well specified. The study includes tests of the effective power of the OLS and JGLS statistics.     

DETECTING ABNORMAL RETURNS USING THE MARKET MODEL WITH PRE-TESTED DATA

The current literature suggests various alternative procedures for increasing the power of tests to detect abnormal returns in event studies. Using randomly constructed portfolios, we simulate events and compare the results of tests using three alternative procedures: traditional, cross-sectional, and cross-sectional with standardized residuals. For each test, we compare results when all observations are included with results when the observations with high trading volume are omitted from the estimation period. The simulation results indicate that both the traditional approach with omitted observations and the cross-sectional approach using standardized residuals with all observations yield approximately the correct test sizes and significantly improve the power of tests to detect abnormal returns. However, the cross-sectional approach using standardized residuals is clearly dominant among the three procedures.     

事件模拟与非正常收益模型的检验力——基于中国A股市场的经验检验

本文分别以1990年12月至2003年12月两市所有A股公司为抽样总体,检验了均值调整模型、市场调整模型和市场模型为基础的多种检验方法的检验力。研究发现,无论事件研究中各公司事件是否相近或重叠,都应采用市场模型为基础的非参数秩检验法。而累积非正常收益的检验也应以市场模型为计算基础。若样本公司事件日相近或重叠,统计量的设置要考虑累积非正常收益截面数据的相关性。均值调整模型在本文所定义的各种检验方法中,均无明显优势。经敏感性测试,本文结论不变。     

Clustering financial time series with variance ratio statistics

This study introduces a new distance measure for clustering financial time series based on variance ratio test statistics. The proposed metric attempts to assess the level of interdependence of time series from the point of view of return predictability. Simulation results show that this metric aggregates time series according to their serial dependence structure better than a metric based on the sample autocorrelations. An empirical application of this approach to international stock market returns is presented. The results suggest that this metric discriminates stock markets reasonably well according to size and the level of development. Furthermore, despite the substantial evolution of individual variance ratio statistics, the clustering pattern remains fairly stable across different time periods.     

A Complete Nonparametric Event Study Approach

Event studies have been used to examine the direction, magnitude, and speed of security price reactions to various phenomenon. Concerns over the lack of normality in stock return distributions motivated the introduction of nonparametric test statistics in the event study literature. A parametric procedure (OLS), however, has been extensively employed in the estimation of parameters for the market model. This paper, in contrast, applies Theil's nonparametric regression in the estimation of abnormal returns; an approach which is distribution free and provides a complete nonparametric approach for the detection of abnormal performance. Simulation results indicate Theil's estimation procedure offers a slight improvement in power in the detection of abnormal performance over the traditionally employed methodology. The results suggest employing Theil's nonparametric estimation procedure combined with the rank statistic. This complete nonparametric combination offers similar power with fewer underlying assumptions.     

Statistical inference in multiperiod event studies

We examine several event-study test statistics that can be used to detect abnormal performance during amultiperiod event window. We demonstrate that one of the most commonly used test statistics does not, under the assumptions made, have the distribution claimed (standard normal), and thus tests using it will be biased. The magnitude of that bias is shown to increase with the length of the event window and can generally be expected to lead to excessive rejection of the null hypothesis. We also compare the relative power of alternative test statistics that are normally distributed and are straightforward to apply.     

New low-frequency spread measures

I develop new spread proxies that pick up on three attributes of the low-frequency (daily) data: (1) price clustering, (2) serial price covariance accounting for midpoint prices on no-trade days, and (3) the quoted spread that is available on no-trade days. I develop and empirically test two different approaches: an integrated model and combined models. I test both new and existing low-frequency spread measures relative to two high-frequency benchmarks (percent effective spread and percent quoted spread) on three performance dimensions: (1) higher individual firm correlation with the benchmarks, (2) higher portfolio correlation with the benchmarks, and (3) lower distance relative to the benchmarks. I find that on all three performance dimensions the new integrated model and the new combined model do significantly better than existing low-frequency spread proxies.     

FURTHER INSIGHT INTO THE STANDARDIZED UNEXPECTED EARNINGS ANOMALY: SIZE AND SERIAL CORRELATION EFFECTS

Studying size and serial correlation effects, the authors examine why portfolios selected on the basis of standardized unexpected earnings (SUEs) exhibit excess returns. Results of the study indicate that the SUE effect and the size effect are independent anomalies; however, a more plausible explanation of the SUE effect may involve serial correlation of SUEs. The authors demonstrate that more than half of the post-announcement responses to current quarter earnings may actually be pre-announcement adjustments to next quarter's earnings.     

Improved Methods for Tests of Long-Run Abnormal Stock Returns

We analyze tests for long-run abnormal returns and document that two approaches yield well-specified test statistics in random samples. The first uses a traditional event study framework and buy-and-hold abnormal returns calculated using carefully constructed reference portfolios. Inference is based on either a skewness-adjusted t -statistic or the empirically generated distribution of long-run abnormal returns. The second approach is based on calculation of mean monthly abnormal returns using calendar-time portfolios and a time-series t -statistic. Though both approaches perform well in random samples, misspecification in nonrandom samples is pervasive. Thus, analysis of long-run abnormal returns is treacherous.     

Discriminating between wealth and information effects in event studies in accounting and finance research

This article examines the power of tests of given size to detect and distinguish between wealth (i.e., mean) and information (i.e., variance) effects in event studies. We find that an Estimated Generalized Least Squares (EGLS) mean-effects test is consistently more powerful than the test based upon the average standardized residual and is as powerful as a nonparametric rank test. Unlike the test based upon the average standardized residual and the rank test, the EGLS test is well specified even when the event affects the variances of the prediction errors. We also find that conventional parametric tests to detect changes in the variance of the event-day average abnormal return are misspecified when the null of no change is true. We analyze the reasons this occurs and suggest a rank procedure that produces tests of the correct size under the null. Our evidence suggests that the critical factors allowing researchers to distinguish between wealth and information effects are an estimation procedure incorporating the heteroskedasticity inherent in market model prediction errors and an explicit test for event-day variance changes.     

AN IMPLICIT MEASURE OF THE EFFECTIVE BID-ASK SPREAD: A NOTE

We extend Roll's study of the effective bid-ask spread in an efficient market environment by allowing for serially correlated order arrival and quote behavior. This extension results in a more general effective bid-ask spread measure, which precludes imaginary spreads and includes Roll's measure as a special case when the serial correlation is zero. This new measure is related to the length of the measurement interval due to the serial correlation, and thus has the potential to explain the previously observed differential between weekly and daily derived spreads.     

Do corporate insiders circumvent insider trading regulations? The case of stock repurchases

We use calculated values of standardized abnormal insider trading activity to investigate for patterns of unusual insider activity around fixed-price and Dutch auction repurchase announcements. Firms are classified according to whether the repurchase is signaling information about future cash flows, about the distribution of excess free cash flows, or about management's attempts to maintain control in the presence of a takeover. We find below normal levels of sales well before the event and above normal levels of sales after the event. This tendency is strongest for fixed-price offers and for firm's conveying information about future cash flows, and is absent for firms involved in takeovers. No evidence exists of abnormal levels of purchases before or after the event. We interpret the evidence as consistent with insiders successfully circumventing policies and regulations designed to prevent the exploitation of private information by timing the pattern of their security sales.     

Financial disclosure,investor protection and stock market behavior: an international comparison

The main purpose of this study is to examine the relationship between financial opacity, investor protection and stock market behavior for sixteen countries. We use the 1995 CIFAR corporate disclosure ratings and the 2006 World Bank investor protection index to measure a country’s relative level of financial transparency and legal protection for investors. The return behavior of each country is examined using numerous time series tests such as serial correlation, Markov chain, runs, duration dependence and variance ratio tests. We found that the results show no significant differences between high and low disclosure countries. However, high disclosure countries appear to be associated with a lower level of stock market volatility. Cox proportional hazard test results indicate that extreme returns (positive and negative) are more likely in low disclosure countries.     

Testing for Serial Correlation in the Presence of Stochastic Volatility

This paper develops a regression-based testing procedure for serial correlation in the presence of stochastic volatility. The asymptotic distribution of the test is derived, and the finite sample properties are investigated. Monte Carlo results shows that the test is reliable in terms of both size and power performances, when the underlying process is a log-linear stochastic volatility. Moreover, the test is superior to Woolridge's (1991) robust LM tests in terms of size in finite sample. Serial correlation tests were conducted for nominal returns of ten exchange rates, and indicated that there is a strong evidence of serial correlation for Yen/Dollar exchange rates.     

EXTENSIONS OF THE STANDARDIZED CROSS‐SECTIONAL APPROACH TO SHORT‐HORIZON EVENT STUDIES

Strong evidence indicates that short‐horizon event‐induced abnormal returns and volatility vary significantly over event days. Event‐study methods that assume constant event‐induced abnormal returns and volatility over event days have potentially inflated Type I error rates and poor test power. Our simple extensions of the Boehmer, Musumeci, and Poulsen (1991) approach scale abnormal returns with conditional variance, which is estimated with GARCH(1,1) and an indicator of the event in a two‐stage estimation. Our method improves the Boehmer, Musumeci, and Poulsen approach on model specification and test power, even under challenging event‐induced mean and volatility structures, and could standardize short‐horizon event studies.     

WINNERS AND LOSERS ON NYSE: A RE-EXAMINATION USING DAILY CLOSING BID-ASK SPREADS

We study a sample of NYSE stocks that experienced a large one-day price change during 1992 and were reported as daily largest percentage gainers and largest percentage losers in the Wall Street Journal. The sample indicates significant reversals during the immediate post-announcement period. We test for market efficiency by using bid-ask spreads obtained from the transactions data for the days immediately after the announcement. The overall results indicate that the returns during the reversal period are less than the average bid-ask spread during the same time. We also find that major losers, firms with ?20 percent to ?50 percent event-date abnormal returns, experience price reversals generating returns that are significantly greater than the average bid-ask spread during that period. We interpret this result as consistent with the overreaction hypothesis. A test of a trading rule to exploit this overreaction is not profitable, providing support for weak-form market efficiency.     

EMPIRICAL EVIDENCE ON THE IMPACT OF THE BID-ASK SPREAD ON THE CHARACTERISTICS OF CRSP DAILY RETURNS

It is widely recognized that Center for Research in Security Prices (CRSP) returns may differ from “true” returns because of the bid-ask effect. Using a large sample of New York Stock Exchange and American Stock Exchange securities, I confirm a discernible bid-ask effect, the magnitude and importance of which decrease with the security's price level (increase with the spread). I find volatility estimates using CRSP returns to be greater than those based on quote returns. However, market model properties, such as β and R², are generally unaffected. Bid-ask effects are clearly apparent in event studies, but because of certain offsetting effects commonly used test statistics remain unaffected. Low-priced stocks (below $2.00) do not conform to these patterns. Finally, the evidence raises the possibility that the existing literature on filter rule tests may underestimate the bid-ask spread component of transaction costs.