Appendix In this Appendix, we present the construction of variables, data source, and some empirical procedures. A.1. Variable Definition and Data Source Variable B/M CAPX/A Cash/A Cash flow volatility Change in analyst recommendation Completion rate Debt/Assets EBIDA/Assets Freq. of secondary share issuance Hedge fund ownership (HF) Information flow manipulation Insider net sales Institutional ownership (IO) Interest coverage ratio Liquidity ratio Log(M/B) Definition and data source (in the parenthesis) Book-to-market equity ratio, where book value of equity is as of the fiscal year end prior to event announcement and market of equity is as of the month end prior to event announcement (CRSP). he ratio of capital expenditure to book value of total assets as of the prior fiscal year end. he ratio of cash and marketable securities to book value of total assets as of the prior fiscal year end. he standard deviation of cash flows (denominated by book value of total assets) during the past five years. he change of the average analyst recommendation score (1 for strong buy, 2 for buy, 3 for hold, 4 for underperform, and 5 for sell) from 12th month before to the month before the event announcement. (I/B/E/S) he ratio of actual repurchases made in the two years following the repurchase announcement to the announced repurchase program size. Actual repurchases are defined, following Grullon and Michaely (2002), as total expenditures on the purchase of common and preferred stocks minus any reduction in the value of the net number of preferred stocks outstanding. (SDC and Compustat) he ratio of long-term debt to book value of total assets as of the prior fiscal year end. he ratio of EBIDA to book value of total assets as of the prior fiscal year end. he proportion of SEO firms that offer both primary and secondary shares. (SDC) he percentage of the event firm shares held by hedge funds in aggregate (homson-reuters Institutional Holdings (13F)) relative to the total number of shares outstanding in the quarter (CRSP). he three-day (-1, 1) cumulative abnormal returns (CAR) of management forecasts of annual earnings-per-share in the month prior to event announcement. he CARs are estimated using the conventional market model with the CRSP value-weighted return as the market portfolio. he parameters of the market model, alpha and beta, are estimated based on the daily returns from day -365 to day -60 relative to the forecast date. he proportion of good news refers to the fraction of forecasts with positive CARs. (First call) Monthly average insider sales minus purchases, denominated by shares outstanding, during the 12 months prior to event announcement, adjusted by the expected average net sales estimated from 48 months before to 13 months before the announcement. (homson-reuters Insiders Data) he percentage of event firm shares held by 13F filing institutions in aggregate (homson- Reuters Institutional Holdings (13F)) relative to the total number of shares outstanding in the quarter. Cash flow divided by interest expense. As in Amihud (2002), the illiquidity for stock i in month t, ILLIQ it = (1 D it ) r itd d=1 DVol itd where D it is the number of trading days in month t, r itd is the daily return of stock i on day d of month t, DVol itd is the dollar trading volume (in millions) of stock i on day d. (CRSP) he natural logarithm of market-to-book equity ratio, where market of equity is as of the month end prior to event announcement (CRSP) and book value of equity is as of the fiscal year end prior to event announcement. D it 1
Log(ME) Non-operating income Number of analysts following OIBDP/A OIBDP/Sales Prior return Proceeds/ME Profit margin Proportion of repurchase in payout RD/A Repurchase frequency Return volatility ROA Size of secondary share issuance Variance ratio he natural logarithm of market capitalization. Non-operating income denominated by book value of total assets as of the prior fiscal year end. he number of analysts who follow the event stock as of the event announcement month. (I/B/E/S) he ratio of operating income before depreciation to book value of total assets as of the prior fiscal year end. he ratio of operating income before depreciation to sales. Stock return in the 12 months prior to the announcement. (CRSP) Announced repurchase/issuance transaction value (SDC) scaled by market capitalization in the month prior to the event announcement (CRSP). he ratio of net income to sales. By firm number, it is the proportion of repurchase firms in the total number of firms that have positive payout. By dollar value, it is the proportion of dollar amount used for repurchases in the total dollar amount that is paid out. he ratio of research and development expenses to book value of total assets as of the prior fiscal year end. he fraction of years during the past five years in which a sample repurchase firm has announced repurchases. (SDC) Standard deviation of daily returns in the prior 12 months. (CRSP) he ratio of net income to book value of total assets as of the prior fiscal year end. he fraction of secondary shares in the total offering shares. (SDC) As proposed in French and Roll (1986), the ratio of per hour open-to-close stock return variance to close-to-open stock return variance during the 12 month prior to the event announcement. (CRSP) A.2. Estimation of Long-Run Abnormal Returns We employ three methods to estimate long-run abnormal returns: (1) the calendar-time portfolio approach (as proposed in Mitchell and Stafford, 2000), (2) Ibbotson s (1975) returns across time and securities (IRAS), and (3) the buy-and-hold abnormal returns (BHAR) relative to control firms matched on size, book-to-market equity ratio, and return momentum (as illustrated in Barber, Lyon, and sai, 1999) In the calendar-time portfolio approach, in each month we construct an event portfolio of firms that have announced stock repurchases or issuances in the past months, where is the holding period. he portfolio is rebalanced monthly to add the event firms that have just announced a repurchase or an SEO and to drop the firms that have reached the end of the holding period. We compute the monthly equal- and value-weighted portfolio excess returns and run a time-series regression of the portfolio excess returns on Fama and French s (1993) three factors: R p,t r f,t = a p + b p (R m,t r f,t ) + s p SMB t + h p HML t + e p,t. (A1) 2
R p,t is the event-portfolio raw return in calendar month t. Regression intercepts a p measure monthly average abnormal return for event-portfolio p. o compute the cumulative abnormal return during the holding period (for example, to be comparable with the abnormal returns estimated from the following two other methods), we can multiply a p by. Our second method follows Ibbotson (1975), in which he proposes a method to estimate the post- IPO aftermarket returns across time and securities. he method is frequently quoted as IRAS (the acronym for Ibbotson s Returns across ime and Securities). Under this method, we run the following regression in each event month τ: R i,t r f,t = a τ + b τ (R m,t r f,t ) + s τ SMB t + h τ HML t + ε τ,t, (A2) where R i,t is the return on stock i in calendar month t that corresponds to event month τ, τ [1, ], with τ = 0 being the month of event announcement. Reported abnormal returns during the holding period are sums of the regression intercepts, i.e., τ=1 α τ. he standard error used to compute the t-statistic is the square root of the sum of the squares of the monthly standard errors. Barber and Lyon (1997) show the efficacy of a control firm approach for detecting long-run abnormal stock returns. his approach yields well-specified test statistics by alleviating the new listing, rebalancing, and skewness biases that typically confound long-run studies. In a subsequent study, Barber, Lyon, and sai (1999) provide a practical guide on how to select a control firm for each sample firm in estimating the buy-and-hold abnormal returns. Our third method follows their suggestion. First, at the end of each June, we construct size deciles for the following 12 months based on the June market capitalization of NYSE stocks, then also assign AMEX and NASDAQ stocks into the appropriate NYSE size deciles based on their market capitalization in June. We further cut the smallest size decile into quintiles of equal numbers of stocks. As a result, we obtain 14 size portfolios in total. In the meantime, we sort all firms at the end of June into deciles based on book-to-market equity ratio (B/M), again using the break points of NYSE firms. Book-to-market equity ratio is defined as the previous fiscal year end book value of equity divided by the market capitalization at the previous calendar year end. his independent double-sorting results in 140 size and B/M portfolios. hen, for each event firm, we select a control firm from the same size and B/M portfolio in the month before announcement that has the closest match on the prior 6-month return and has not been involved in the same type of event (repurchase or SEO) in the prior months. If the control firm is delisted before the end of the holding period, we use the second closest match firm as the replacement 3
(and the third closest if the second closest is also delisted later). he long-run abnormal return is the difference in the buy-and-hold return between the event and the control firm: BHAR i = τ=1(1 + R i,τ ) τ=1 (1 + R control,τ ). (A3) A.3. Rhodes Kropf, Robinson, and Viswanathan (2005) Decomposition of the M/B Equity Ratio Rhodes Kropf, Robinson, and Viswanathan (2005) propose a way to decompose the M/B equity ratio into a component of long-run growth opportunities and another component of (firm-level and industrywide) mispricing. In particular, a firm s log market-to-book equity ratio can be decomposed into two items, Log ( M ) = Log B (M) + Log V (V ), where M is the market value of equity and B is the book value of B equity; both are observable. V stands for the intrinsic value of equity, which is unobservable. Based on the accounting residual income model, Rhodes Kropf, Robinson, and Viswanathan (2005) assume that V is a linear function of the firm s book value of equity, net income (i.e., the growth of book value of equity), and leverage, or empirically, the fitted value of the following regression, Log(M it ) = α 0jt + α 1jt Log(B it ) + α 2jt Log( NI it ) + α 3jt I Log( NI it ) + α 4jt LEV it + ε it. (A4) he parameters are allowed to vary over time (t) and across industries (j) to capture variation in investment opportunities across times and industries. he parameters also capture differences in discount rates among firms. NI it stands for the absolute value of net income of firm i at time t. I is an indicator variable that equals one for firm-years with negative net income and zero otherwise. LEV is market leverage ratio. ε it captures the deviation of intrinsic value from the observed market value of equity and, therefore, is a natural proxy for mispricing. Like Rhodes Kropf, Robinson, and Viswanathan (2005), we run the above cross-sectional regressions for each industry in each year to estimate the parameters α jt. he three variables, i.e., book value of equity, net income, and leverage, are able to explain the within-industry cross-sectional variations of market value of equity well, generating regression R-squared over 80% for most industries. We then take the time series average of α jt, the estimated α jt, from the above regression, to compute the long-run average parameters α j = 1 t α jt. he final measure of mispricing, after purging growth opportunities, is Log ( M ) = Log(M V it) [α 0 + α it j jlog(b 1 it ) + α jlog( NI 2 it ) + α ji 3 Log( NI it ) + α jlev 4 it ]. (A5) 4
A.4. Estimation of Earning Management (Abnormal Accruals) Our estimation of earnings management (abnormal accruals) follows Gong, Louis, and Sun (2008). Specifically, for each calendar quarter and two-digit SIC-code industry, we estimate the following model using all firms with necessary data available in Compustat: 4 A i = j=1 γ j 1 Q j,i + γ 4 Sales i + γ 5 PPE i + γ 6 LA i + γ 7 Assets i + ε i. (A6) A is total accruals, measured as CA CL CASH + SD DEP, where CA is change in current assets, CL is change in current liabilities, CASH is change in cash and marketable securities, SD is change in debt in current liabilities, and DEP is depreciation and amortization expense. Q j is a dummy variable that equals one for fiscal quarter j and zero otherwise; Sales is the quarterly change in sales; PPE is property, plant, and equipment at the beginning of the quarter; LA is the lag of total accruals; Assets is total assets at the beginning of the quarter; and ε, the regression residual, is the abnormal accrual. All the variables are scaled by total assets at the beginning of the quarter. o mitigate the effects of outliers and errors in the data, we delete the top and bottom 1% of the variables in the estimation for each quarter. We also require at least 20 observations for each estimation. We then adjust the estimated abnormal accruals (i.e., the regression residuals) for performance. For each quarter and each industry (two-digit SIC code), we create five portfolios with at least four firms each by sorting the data into quintiles based on return-on-assets (ROA) from the same quarter in the previous year. he performance-matched abnormal accruals for a sample firm are the firm-specific abnormal accruals minus the median abnormal accruals for its respective industry-performance-matched portfolio. In addition to controlling for performance, the portfolio benchmarking approach controls for random effects arising from other events that may affect accruals (Korthari, Leone, and Wasley, 2005, Journal of Accounting and Economics). 5