Ítem 12: Tiene que alcanzar elementos de trabajo que están muy alejados de su cuerpo
4.3.2. Documentación de los métodos seleccionados
The sample period in this study ranges from 01/1971 to 12/2006. However, the actual
period included in this study is from 01/1971 to 01/2007 because the subsequent realized
monthly returns are used in this dissertation as the proxy of the expected returns.
I retrieve the data of the sample firms from the COMPUSTAT and the CRSP
databases. Firms missing the following information are excluded from the sample: debt in
one year (DATA34) or long-term debt (DATA9) from the COMPUSTAT, price
information, shares outstanding or monthly returns from the CRSP database.
The following data are extracted from the COMPUSTAT North America annual file:
• Common equity (DATA60), which is used to measure book value (BV);
• Debt in current liabilities (DATA34)
• Long-Term Debt (DATA9)
• Long-term debt issuance (COMPUSTAT # 111)
• Change in current debt (COMPUSTAT # 301)
• Capital expenditure (COMPUSTAT # 128)
The financial information above has the report-delay problem. More specifically,
the COMPUSTAT database extracts its financial information disclosed by publicly traded
companies mainly from Forms 10-K and 10-Q, which are the annual and the quarterly
reports required by the SEC. The current deadlines for filing periodic reports are
implemented on Nov. 15th, 2002. These deadlines are reported in Table III (See
Appendix B, Table III for detail).
Prior to the change, a domestic reporting company must file a quarterly report no
later than 45 calendar days after the end of each of its first three fiscal quarters, and an
annual report no later than 90 calendar days after the end of its fiscal year. In the new
ruling, the filers are grouped into three groups - large accelerated filers, accelerated filers
and non-accelerated filers. The deadlines for the non-accelerated filers have not changed.
The deadlines of the 10-Q for the accelerated filers are shorten from 45 to 40 days. The
deadline of the 10-K for the large accelerated filers is reduced to 60 days and for the
accelerated filers to 75 days.
Since there can be a delay of up to 90 days for the 10-K form, I use the annual
financial information at year t four months after its reporting calendar date to calculate
the distance-to-default measure to make sure that all information is available to investors
when the default measure is calculated. For example, the 2005 fiscal year-end data from
the COMPUSTAT database will be used to match the CRSP data from May 1st, 2006 to
value used is from the annual financial information at year t six months after its reporting
calendar date.
The CRSP monthly file is used for the following variables:
• Monthly price
• Holding period returns (including dividend)
• Shares outstanding
• Delisting price
• Delisting returns (dividend included)
• Delisting date
• Shares outstanding when de-listed.
Monthly equity returns used in portfolio and regression analysis are from the CRSP
monthly file, which also contains delisting information.
As the returns of the distressed stocks are directly related to the delisting returns, the
empirical study needs to carefully consider the delisting of stocks. In many cases, the
CRSP monthly file reports delisting dates and delisting returns. This study has 15,937
delisting returns available in the sample, including delisting due to performance-related
reasons (The CRSP delisting code between 400 and 599) and those due to the other
reasons, including mergers and change of exchanges. In the case of the delisting stocks
without the available return information in the CRSP, the last available full-month returns
were used.
The following data are from the CRSP daily file:
• Daily equity price (dividend adjusted)
The price and shares outstanding used to calculate DLI is mostly from the CRSP daily
file. In addition, this study adjusted the delisting returns according to the delisting date
and return information from the CRSP monthly file. The annual equity volatility (σE) is
calculated using the adjusted daily historical data from the CRSP database. More
specifically, the following procedures are used.
Assuming the number of observations is n+1; PE,iVE i, is the stock price at the end of
the ith interval, with i=0,1,…, n+1
And let ) ln( 1 , , − = i E i E i P P r for i=1,2,…,n+1.
The estimation of daily volatility of r is given by i
∑
= − = n i i r r n s 1 2 ) ( 1The volatility per annum can be calculated from the volatility per trading day using
the following formula:
Volatility per annuam Volatility per trading day Number of trading datys per annum s 252
= ×
= ×
Market value (MV) is defined as the product of price at time t and the corresponding
shares outstanding; book-to-market ratio (BM) is defined as book value (BV) divided by
market value (MV). Firms with negative book values are excluded from the sample.
Monthly observations of the one-year Treasury bill rate obtained from the Federal
The historical data of Fama-French three risk factors, including the excess return,
SMB, and HML are from Kenneth French’s website24.
The aggregate default likelihood measure (ADLI) at time t is defined as a simple
average of the default likelihood indicators of all firms in a portfolio.
Table IV summarizes the descriptive statistics of ADLI, size and BM. The table also
includes the summary statistics of the Fama and French factors (HML and SMB). Panel
A reported the summary statistics of the time series. Panel B is the correlation matrix
among the aggregated variables and Fama and French factors. To calculate the values in
Table IV, this thesis first calculates ADLI, size and BM measure each month as the
simple average of all firms in that month to get the time series of these variables. The
time series are then compared with the Fama and French factors (See Appendix B, Table
IV for detail).
The relationships between ADLI, size and BM are quite interesting. The correlation
table shows a significant positive correlation (0.098) between ADLI and BM and a
significant negative correlation (-0.548) between ADLI and size, which are consistent
with our intuition. However, there was a positive correlation (0.211) between the market
value and BM, which is in conflict with my expectation of a negative relationship
because cēterīs paribus, the increase of market value would decrease BM. The significant
positive concurrent correlation between size and BM suggests that the book value may
increase/decrease more than the increase/decrease of the market value. This increasing
book value can be seen as an increasing safety cushion of higher market risks due to high
stock prices. Furthermore, there is no significant correlation between ADLI and the Fama
French factors, which suggests that ADLI may incorporate different information from
24
HML and SMB. The two Fama French factors are significant and negatively correlated
with each other (-0.227), which is what I expected because larger book-to-market ratio
CHAPTER V