Chapter 5: Avenues For Future Research And Conclusions
A.2. Variable Construction
A.2.1. Description of Explanatory Variables
The variable name and descriptions are outlined below in Table A.2. Variables are calculated based on each stock i and quarter q.
Table A.2: Description of Explanatory Variables
The table shows the description and data source of the explanatory variables. The variables are institutional ownership (Instii,q); number of announcements released on same day (Distract_Ui,q); number of analysts issuing
forecasts within the 90 days prior to the earnings announcement (Analysti,q); volatility of daily BHAR in the 40
trading days prior to earnings announcement (Volatilityi,q); arbitrage risk (ArbRiski,q); hedgeable risk (ExpRiski,q);
average of Amihud illiquidity measured across 40 trading days prior to earnings announcement (Illiqi,q); average
of quoted spread at closing across 40 trading days prior to earnings announcement (Spreadi,q); stock price at the
end of quarter, q (Pricei,q); average daily turnover (in millions of dollars) measured from 271 trading days to 22
trading days prior to earnings announcement (Turni,q); market capitalisation (Mcapi,q); book-to-market ratio
(BMi,q); BHAR across the past 40 trading days prior to earnings announcement (Momi,q). BHAR are calculated
using 5x5 size quintile and BM quintile matched portfolios. Data source refers to datasets obtained from the Wharton Research Data Service except FF which refers to the Fama-French factors obtained from Professor Kenneth French’s website.
Variable Description Data Source
Instii,q Fraction of shares outstanding held by institutions filing Form for the
quarter prior to the earnings announcement (% multiplied by 0.01).
CDA Spectrum Distracti,q Number of earnings announcements for each trading day. I/B/E/S
Analysti,q Number of analysts reporting quarterly earnings forecasts to I/B/E/S in
the 90 days prior to earnings announcement. Only the most recent forecast per unique analyst is kept.
I/B/E/S
Volatilityi,q Standard deviation of daily abnormal returns across 40 days prior to
earnings announcement.
CRSP, FF ArbRiski,q Residual variance from market model regression (using returns from the
S&P 500 index (dividends included)) estimated over 48 months ending 1 month prior to earnings announcement.
CRSP
ExpRiski,q Explained variance from market model regression (using returns from the
S&P 500 index (dividends included)) estimated over 48 months ending 1 month prior to earnings announcement.
CRSP
Spreadi,q The average of bid-ask spread at close across past 40 trading days prior
to earnings announcement (% multiplied by 0.01).
CRSP Illiquidityi,q Illiquidity measure following Amihud (2002). The average of absolute
value of daily return divided by the daily dollar share turnover. Measured across the 40 trading days prior to earnings announcement.
CRSP
Pricei,q Stock price at the end of the announcement quarter (adjusted for stock
splits)
Compustat Turn,q Average daily share turnover between 271 and 22 trading days prior to
earnings announcement ($million).
CRSP Mcapi,q Number of shares outstanding multiplied by price as measured in the most
recent June ($million).
Compustat BMi,q Book-to-market ratio as measured in the most recent December. Compustat
171
A.2.2. Investor Sophistication (Instii,q)
Investor sophistication is proxied by the fraction of shares outstanding held by institutions. To calculate percentage of institutional ownership I follow Campbell et al. (2009) and obtain a
firm’s quarterly 13F filings.112 The 13F filing is required by all institutions in the US with greater than $100 million of securities under discretionary management regardless of whether they are regulated by the SEC. All holdings greater than 10 000 shares or larger than $200,000 are disclosed. For each firm-CUSIP and date I sum up the shares held by all institutions to arrive at the total institutional ownership by the end of quarter. I obtain for each firm-quarter the total institutional ownership from the 13F filings and divide by the firm market capitalization calculated from the CRSP-Compustat merged database by multiplying the total shares outstanding with the stock price (both adjusted for stock splits). I follow Campbell et al. (2009) and remove any firm-quarter observation where the level of institutional ownership as a percentage of shares outstanding is greater than 100%.
A.2.3. Analyst Following (Analysti,q)
When dealing with analyst forecast data, I note that analysts may release multiple forecasts for quarterly earnings and therefore I exclusively select the latest forecast for each analyst per firm- quarter per firm from I/B/E/S. I remove any forecasts more than 90 calendar days old.
A.2.4. Arbitrage Risk (ArbRiski,q) and Hedgeable Risk (ExpRiski,q )
Arbitrage risk is defined as the idiosyncratic portion of a stock’s volatility that arbitrageurs
cannot diversify away by holding offsetting positions in index funds. The implicit assumption is that arbitrageurs hold large, but few, positions at any one time and therefore are not diversified investors.
172 Arbitrage risk is computed by the method of Wurgler and Zhuravskaya (2002) which runs a regression, per firm-quarter observation, of risk-free rate adjusted return on the risk-free rate adjusted return of the S&P 500. The residual variance of the regression proxies arbitrage risk while the explained variables proxy the hedgeable risk. My regression is estimated over 48 months and ends 1 month prior to the earnings announcement for each firm-quarter observation.
A.2.5. Illiquidity (Illiquidityi,q) and Transaction Cost (Spreadi,q)
The illiquidity measure follows Amihud (2002) and is calculated by taking the daily absolute return multiplied by 1 million divided by the daily turnover. For proxy for transaction cost I follow Vega (2006) and take the average of the closing spread from 41 trading days to 2 trading days before earnings announcements.
173