Transparency and the Pricing of Market Timing Xin Chang Nanyang Technological University Zhihong Chen City University of Hong Kong Gilles Hilary INSEAD Research Questions • Can managers lower the cost of equity by actively timing the market when they issue external capital? • What is the role played by corporate transparency in this process? Can Firm Time The Market? • The “market timing theory” relies on the idea that managers know more about the fundamental value of their firm than outside investors, – Managers detect temporary mispricings. • Managers can try to take advantage of the mispricing by issuing or buying back capital. Can Firm Time The Market? • If new investors take the issuance of capital as a signal that a firm is overvalued, the price should adjust. (e.g., Myers and Majluf, 1984). – If true, managers and current shareholders cannot take advantage of mispricings. • If quasi-rational investors buy the new capital at the inflated price, managers can transfer wealth from these new investors to current shareholders. Can Firms Time The Market? • There is a positive correlation between good market conditions and equity issuances (e.g., Loughran et al. 1994, Graham and Harvey 2001). • It is not clear if this reflecting managers’ private information or something else such as time-varying investment opportunities (Schultz 2003, Baker, et al. 2007). Pricing of Market Timing • If current investors believe this is true, these issuance gains should be reflected in the firm valuation. • The price of successful market timers should be higher for a given level of expected earnings. – Equivalently, holding profitability and risk constant, the discount rate implied by a price and a given stream of expected earnings is lower. Market Timing Pricing • H1: Firms that are expected to time the market when they issued or repurchased capital should have a lower expected cost of equity capital. Are SEOs overvalued on average? • Firm level (Ritter 2003) and aggregate level (Baker and Wurgler 2000) evidence on abnormal performance after SEOs. • But, all studies use ex-post returns and complex procedures to address associated econometric problems. • Ritter (2003) indicates that “the conclusions regarding abnormal performance are hotly debated and sensitive to the methodology employed and the sample used”. Broader View • We rely on the aggregate amount of capital issued by firms, rather than focusing on special and rare events such as IPOs or SEOs. • Takeuchi (2008) reports that – firms making SEOs represent only 6% of firms with net increases in equity – 18% of firms with net increases in equity of more than 10% of assets in a given year. Transparency • Transparency affects financial policy. – Poor accounting quality is associated with higher SEO issuance costs (Lee and Masulis (2009)). – Transparent firms have more flexibility to issue equity (rather than debt), have a greater control over the issuance size and are less influenced by market conditions (Chang et al. (2006, 2009)). Transparency • Transparent firms obtain a fair price when they issue equity in periods of low sentiment and capture excess value in periods of high sentiment. • Opaque firms “break-even” in periods of high sentiment and abstain from issuing equity in periods of low sentiment. • H2: The effect of past market timing on the expected cost of capital should be stronger for transparent firms than for opaque firms. Main Specification Ri*,t R f ,t 1MTCovi,[0,t ] C ei,t • Measure of market timing: MTCov = Cov(EF,MB) / Assets where EF: sum of net debt and equity issues for a given year. MB: market to book ratio for the year. Estimate Implied Cost of Equity • Four implied cost of equity models • All based on dividends discount model but make different assumptions on future earnings growth. • Use the average of the four estimates to mitigate model-specific measurement errors. Why not ex post returns? • Market timing ability relies on the existence of quasirational investors and information asymmetry between different classes of investors. – Properties of market equilibrium models in this complex setting are not well-known. • Debatable whether the ex post return is an appropriate proxy for a firm’s cost of capital. – May reflect the shocks to a firm’s growth opportunities, expected growth rates or investors’ risk aversion. – Fama and French (1997) conclude that expected returns estimated by ex post returns are imprecise because of the uncertainty of factor premiums and factor loading estimates. • Firms may have a very active financing policy. Control Variables • • • • • • • • • Beta Size Book-to-Market Leverage Price Momentum Forecast Errors Forecast of Long-term Growth Lagged Industry Risk Premium Year fixed effects. Main Specification Ri*,t Rf ,t MTCovi,[1,t ] AQxMTCovi,[1,t ] AQi,t Z ei,t • AQ: a measure of accounting quality similar to Francis et al. (2005). Data and Sample Selection • Start from Compustat/CRSP merged file. • Eliminate utilities and financial firms. • Require firms to be listed for at least 3 years. • Require observations to have all four cost of equity estimates and control variables. • Final sample contains 26,286 firm-year observations from 1981 to 2007. Descriptive Statistics Variables R*-Rf N 26,286 Mean 5.351 Stdev Q1 3.134 MTCov(1, t) 26,286 0.007 AQ 26,286 Beta LogMV LogBM Leverage Price Momentum (MMT) Forecast error (Ferr) Long-term earnings growth forecast (Fltg) Industry risk premium (IndRp) 26,286 26,286 26,286 26,286 26,286 26,286 26,286 26,286 3.362 Median 4.787 Q3 6.698 0.114 -0.009 0.003 0.022 -0.049 0.034 -0.063 -0.040 -0.026 1.161 6.632 -0.766 0.130 0.102 -0.017 0.163 5.182 0.628 1.659 0.704 0.131 0.404 0.053 0.072 1.448 0.754 5.446 -1.178 0.016 -0.113 -0.019 0.117 4.303 1.084 6.531 -0.724 0.096 0.105 -0.003 0.150 5.112 1.467 7.674 -0.299 0.201 0.323 0.003 0.200 6.030 Is Market Timing a Firm Characteristic? • For each year from 1970 to 2002 (or 1997), we estimate the following cross-sectional regression MTCovi,[t ,t N ] 1MTCovi,[0,t 1] ei,[t ,t N ] • We try N = 5 and 10. • β1 is positive and significant at 5% level or below – in 27 (32) of the 33 years at 5% (10%) when N=5 – in 26 (28) of the 28 years at 1% ( 5%) when N=10. Table 3 Dependent variable: R* - Rf Predicted Signs OLS Regression MTCov(1,t) ? Beta + 0.150*** (2.34) LogMV - -0.315*** (-7.13) LogBM + 0.657*** (6.59) Leverage + 3.581*** (11.79) Price Momentum (MMT) - -1.873*** (-11.98) Forecast error (Ferr) - -11.493*** (-10.60) Long-term earnings growth forecast (Fltg) ? 8.009*** (12.46) Industry risk premium (IndRp) + 0.421*** (8.46) Year Fixed Effects Adjusted R2 N -0.293*** (-3.25) Yes 0.412 26,286 Table 4 Predicted Sign MTCov - AQ×MTCov - AQ - Beta + LogMV - LogBM + Leverage + Price Momentum (MMT) - Forecast error (Ferr) - Long-term earnings growth forecast (Fltg) + Industry risk premium (IndRp) + Year Fixed Effects Adjusted R2 N Model 1 -0.617*** (-4.97) -6.040*** (-3.33) -5.071*** (-4.23) 0.098*** (2.71) -0.267*** (-13.44) 0.716*** (14.12) 3.784*** (15.82) -1.878*** (-35.43) -10.980*** (-17.73) 7.610*** (15.15) 0.411*** (18.94) Yes 0.419 26,286 Table 5 (I) (II) (III) Predicted Sign MTCov_Sent(1,t) - MTCov_Resid(1,t) - MTCov_Pred(1,t) 0 Beta + 0.175*** (2.88) 0.145** (2.22) LogMV - -0.308*** (-7.13) -0.302*** (-6.75) -0.303*** (-15.23) LogBM + 0.652*** (6.72) 0.671*** (6.60) 0.680*** (13.10) Leverage + 3.768*** (11.91) 3.617*** (11.91) 3.667*** (14.95) Price Momentum (MMT) - -1.893*** (-12.04) -1.886*** (-12.00) -1.882*** (-35.04) Forecast error (Ferr) - -11.477*** (-10.59) -11.425*** (-10.66) -11.439*** (-18.21) Long-term earnings growth forecast (Fltg) ? 8.117*** (12.27) 7.942*** (11.53) 7.859*** (15.58) Industry risk premium (IndRp) + 0.414*** (8.50) 0.422*** (8.59) 0.419*** (19.22) Year Fixed Effects Adjusted R2 N -0.442*** (-3.63) -0.223*** (-3.22) -0.099 (-1.35) 0.146*** (3.86) Yes Yes Yes 0.413 26,286 0.412 25,935 0.411 25,935 Robustness Tests • Estimation of cost of equity • Estimation of transparency • Estimation of market timing activity Table 6 – Panel C Alternative measure of transparency Coefficient estimates of TRAN*MTCOV (t-statistic) TRAN = Innate component of AQ -7.219*** (-2.21) TRAN = Discretionary component of AQ -4.844*** (-2.07) TRAN = negative absolute value of discretionary revenue -3.985** (-1.93) TRAN = natural logarithm of number of analyst following. -0.308*** (-3.26) Table 7 Coefficient Estimate (t-statistic) * Dependent variable: R - Rf Predicted sign MTCov - PIN ? MTCov×PIN + DedOwn ? MTCov*DedOwn - -1.098*** (-4.77) -2.557*** (-3.14) 4.722*** (3.94) + LogMV LogBM + Leverage + Price Momentum (MMT) Forecast error (Ferr) Long-term earnings growth forecast (Fltg) ? Industry risk premium (IndRp) + Adjusted R2 N -0.166 (-1.54) -0.290 (-0.55) -1.752** (-2.19) Beta Year Fixed Effects Coefficient Estimate (t-statistic) 0.171** (2.54) -0.357*** (-7.41) 0.789*** (9.14) 2.496*** (9.79) -1.866*** (-10.59) -12.353*** (-10.15) 7.787*** (11.02) 0.387*** (7.25) 0.151** (2.33) -0.314*** -(7.26) 0.880*** (8.25) 2.433*** (10.68) -1.869*** -(12.00) -11.747*** -(10.89) 7.915*** (12.27) 0.424*** (8.52) Yes Yes 0.42 21,092 0.41 26,286 Conclusions • Our results suggest that managers can reduce the cost of equity by timing the market. – Effects are both statistically and economically significant. – Robust to multiple specifications • The effects are stronger for transparent firms than for opaque firms. Thank You