Lecture Notes Liu, Whited, and Zhang (2009, J. of Political Economy): Investment-based Expected Stock Returns Lu Zhang1 1 The Ohio State University and NBER BUSFIN 920: Theory of Finance The Ohio State University Autumn 2012 Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Introduction What We derive and test q-theory implications for the cross-section of expected stock returns Introduction Why: Many characteristics-return relations in asset pricing Realized returns z}|{ rjt+1 Abnormal returns = Et [rjt+1 ] | {z } + z}|{ jt+1 Expected returns Use the q-theory of investment to link expected returns to firm characteristics Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Key Results Ten SUE portfolios, the CAPM Average predicted returns 0.3 0.25 0.2 Low High 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten SUE portfolios, the Fama-French model Average predicted returns 0.3 0.25 0.2 Low High 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten SUE portfolios, the consumption-CAPM Average predicted returns 0.3 0.25 Low High 0.2 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten SUE portfolios, the q-theory model 0.3 Average predicted returns High 0.25 0.2 0.15 0.1 Low 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten B/M portfolios, the CAPM Average predicted returns 0.3 0.25 0.2 Low High 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten B/M portfolios, the Fama-French model Average predicted returns 0.3 0.25 0.2 0.15 Low High 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten B/M portfolios, the consumption-CAPM Average predicted returns 0.3 0.25 0.2 High Low 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten B/M portfolios, the q-theory model Average predicted returns 0.3 0.25 High 0.2 Low 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten CI portfolios, the CAPM Average predicted returns 0.3 0.25 0.2 0.15 0.1 High 0.05 0.05 0.1 Low 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten CI portfolios, the Fama-French model Average predicted returns 0.3 0.25 0.2 0.15 0.1 0.05 0.05 High 0.1 Low 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten CI portfolios, the consumption-CAPM Average predicted returns 0.3 0.25 High Low 0.2 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Ten CI portfolios, the q-theory model Average predicted returns 0.3 0.25 High 0.2 Low 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Key Results Predicted vs. realized stock return volatilities, joint estimation of mean and variance, the q-theory model 0.4 Predicted volatilities 0.35 Low 0.3 0.25 0.2 0.15 0.1 High 0.05 0 0 0.1 0.2 0.3 Realized volatilities 0.4 Key Results Average predicted vs. realized returns, joint estimation of mean and variance, the q-theory model Average predicted returns 0.3 0.25 Low High 0.2 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Model The neoclassical q-theory of investment. Firms use capital and costlessly adjustable inputs such as labor Operating profits, Π(Kit , Xit ), with Yit ∂Π(Kit , Xit ) =α ∂Kit Kit with Yit = Sales Capital evolves as: Kit+1 = Iit + (1 − δit )Kit Convex adjustment costs: a Φ(Iit , Kit ) = 2 Iit Kit 2 Kit Model Equity-value maximization B One-period debt, Bit+1 , with corporate bond return rit+1 Payout, Dit , defined as: (1−τt )[Π(Kit , Xit )−Φ(Iit , Kit )]−Iit +Bit+1 −ritB Bit +τt δit Kit +τt (ritB −1)Bit The cum-dividend market value of the equity: "∞ # X Vit ≡ max Et Mt+s Dit+s ∞ {Iit+s ,Kit+s+1 ,Bit+s+1 }s=0 s=0 in which Mt+1 is the pricing kernel, correlated with Xit+1 Model The investment return I I Et [Mt+1 rit+1 ] = 1, in which rit+1 is the investment return: Marginal benefit of investment at time t+1 z " }| 2 # Yit+1 a Iit+1 (1 − τt+1 ) α + K 2 Kit+1 it+1 | {z } Marginal product plus economy of scale (net of taxes) Iit+1 +τt+1 δit+1 + (1 − δit+1 ) 1 + (1 − τt+1 )a Kit+1 | {z } I ≡ rit+1 Expected continuation value 1 + (1 − τt ) | {z Iit Kit } Marginal cost of investment at time t { Model The WACC Proposition Ba = (1 − τ B Ba Define rit+1 t+1 )rit+1 + τt+1 , then Et Mt+1 rit+1 = 1 Define Pit ≡ Vit − Dit and the stock return S rit+1 ≡ (Pit+1 + Dit+1 )/Pit Under constant returns to scale, the investment return is the weighted average of stock and after-tax bond returns: I Ba S S Iw rit+1 = wit rit+1 + (1 − wit )rit+1 ⇒ rit+1 = rit+1 ≡ I Ba rit+1 − wit rit+1 1 − wit in which wit is market leverage, wit ≡ Bit+1 /(Pit + Bit+1 ) Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Econometric Methods GMM Do expected stock returns equal expected levered investment returns? h i S Iw E rit+1 − rit+1 =0 Do stock return variances equal levered investment return variances? h i2 h i2 S S Iw Iw − E rit+1 − E rit+1 E rit+1 − rit+1 =0 Econometric Methods Testing portfolios Three sets of testing portfolios Ten Standardized Unexpected Earnings (SUE) portfolios of Chan, Jegadeesh, and Lakonishok (1996) Ten book-to-market portfolios as in Fama and French (1993) Ten “abnormal” investment portfolios of Titman, Wei, and Xie (2004) Why portfolios? Larger and more reliable expected return spreads across portfolios than across individual stocks Smoothing lumpy investment as in Thomas (2002) middle of year t to the middle of year t+1. The bottomline is that the investment return timing matches naturally with the stock return timing from the Fama-French portfolio approach. Econometric Methods Timing Figure 1: Timing Alignment between Annual Stock and Investment Returns I rit+1 - (from July of year t to June of t+1) - τ t , Iit (from January of year t to December of t) τ t+1 , δ it+1 Yit+1 , Iit+1 - (from January of year t+1 to December of t+1) December/January December/January December/January t June/July t+1 June/July t+2 Kit Kit+1 S rit+1 B , r Ba rit+1 it+1 (from July of year t to June of t+1) Kit+2 - Econometric Methods Measurement Kit : gross property, plant, and equipment Iit : capital expenditure minus sales of property, plant, and equipment Yit : sales Bit : total long-term debt Pit : market value of common equity δit : the amount of depreciation divided by capital B : impute bond ratings, assign corporate bond returns of a rit+1 given rating to all firms with the same rating τt : statutory tax rate of corporate income Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Expected Stock Returns Point estimates and tests of overidentification a [ste] α [ste] χ2 d.f. p m.p.e. SUE B/M CI 7.68 [1.72] 0.32 [0.03] 22.34 [25.47] 0.50 [0.31] 0.97 [0.29] 0.21 [0.02] 4.37 8 0.82 0.74 5.99 8 0.65 2.32 6.52 8 0.59 1.51 Expected Stock Returns Euler equation errors, ten SUE portfolios Low 5 High H−L [tH−L ] Panel A: Ten SUE portfolios ei eiFF eiC eiq −1.69 −4.59 −8.07 0.26 6.56 1.96 −0.04 1.66 10.86 9.47 5.31 −0.15 12.55 14.06 13.38 −0.40 [5.53] [5.31] [1.35] [−0.41] Expected Stock Returns Euler equation errors, ten B/M portfolios Low 5 High H−L [tH−L ] Panel B: Ten B/M portfolios ei eiFF eiC eiq −4.91 −0.54 −5.43 −3.94 5.19 1.80 0.27 2.35 13.65 6.76 6.88 −2.73 18.56 7.30 12.31 1.21 [2.51] [3.25] [0.26] [0.79] Expected Stock Returns Euler equation errors, ten CI portfolios Low 5 High H−L [tH−L ] Panel C: Ten CI portfolios ei eiFF eiC eiq 8.21 6.45 4.03 −0.97 5.89 1.54 0.46 2.72 1.91 0.11 −4.35 −1.45 −6.30 −6.34 −8.38 −0.49 [−3.88] [−3.99] [−1.35] [−0.41] Expected Stock Returns Economic determinants of expected stock returns (1 − τt+1 ) I rit+1 ≡ Iw rit+1 ≡ Yit+1 αK it+1 a 2 Iit+1 Kit+1 2 + + τt+1 δit+1 i h +(1 − δit+1 ) 1 + (1 − τt+1 )a KIit+1 it+1 Iit 1 + (1 − τt )a Kit I Ba rit+1 − wit rit+1 1 − wit B Yit+1 /Kit+1 , Iit+1 /Iit , δit+1 , and Iit /Kit ; wit and rit+1 Expected Stock Returns Characteristics, ten SUE portfolios Iit /Kit (Iit+1 /Kit+1 )/(Iit /Kit ) Yit+1 /Kit+1 δit+1 wit B rit+1 Low 5 High H−L [tH−L ] 0.12 0.89 1.52 0.08 0.30 9.44 0.11 1.00 1.50 0.08 0.28 9.76 0.12 1.06 1.83 0.08 0.21 9.38 0.00 0.17 0.31 0.00 −0.10 −0.06 [0.70] [4.06] [5.16] [0.63] [−5.83] [−0.27] Expected Stock Returns Expected returns accounting, ten SUE portfolios Iit /Kit qit+1 /qit Yit+1 /Kit+1 w it Low 5 High H−L −2.48 −5.23 −0.78 0.13 4.45 1.76 0.39 1.89 −4.26 3.62 3.53 −1.46 −1.78 8.85 4.31 −1.58 Expected Stock Returns Characteristics, ten B/M portfolios Iit /Kit (Iit+1 /Kit+1 )/(Iit /Kit ) Yit+1 /Kit+1 δit+1 wit B rit+1 Low 5 High H−L [tH−L ] 0.18 0.98 1.95 0.10 0.08 8.17 0.11 1.00 1.45 0.07 0.27 8.09 0.08 1.02 1.38 0.07 0.53 8.52 −0.10 0.04 −0.57 −0.03 0.44 0.35 [−7.95] [0.68] [−6.77] [−5.01] [12.44] [1.05] Expected Stock Returns Expected returns accounting, ten B/M portfolios Iit /Kit qit+1 /qit Yit+1 /Kit+1 w it Low 5 High H−L −42.06 −1.92 0.16 −6.00 4.69 2.11 0.92 2.19 48.17 −4.06 −6.33 5.58 90.23 −2.14 −6.49 11.58 Expected Stock Returns Characteristics, ten CI portfolios Iit /Kit (Iit+1 /Kit+1 )/(Iit /Kit ) Yit+1 /Kit+1 δit+1 wit B rit+1 Low 5 High H−L [tH−L ] 0.09 1.25 1.84 0.08 0.35 8.47 0.11 1.04 1.58 0.07 0.25 8.27 0.16 0.81 1.89 0.08 0.28 8.44 0.07 −0.44 0.05 0.00 −0.07 −0.03 [11.06] [−7.23] [0.38] [−0.46] [−2.59] [−0.15] Expected Stock Returns Expected returns accounting, ten CI portfolios Iit /Kit qit+1 /qit Yit+1 /Kit+1 w it Low 5 High H−L 2.86 0.73 0.57 1.80 3.50 2.97 −0.44 2.61 −5.67 −3.87 0.09 −0.91 −8.53 −4.60 −0.48 −2.71 Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Joint Estimation Point estimates and tests of overidentification SUE B/M CI 28.88 [16.25] 0.61 [0.27] 11.48 [4.75] 0.35 [0.07] 16.23 [5.53] 0.36 [0.08] χ2(2) d.f.(2) p(2) m.p.e.(2) 5.14 8 0.74 0.03 6.18 8 0.63 0.04 6.05 8 0.64 0.02 χ2(1) d.f.(1) p(1) m.p.e.(1) 5.22 8 0.73 3.45 4.38 8 0.82 2.58 4.81 8 0.78 2.22 χ2 d.f. p 5.45 18 1.00 6.17 18 1.00 6.62 18 0.99 a [ste] α [ste] Joint Estimation Predicted vs. realized stock return volatilities, ten B/M portfolios, the q-theory model 0.4 Predicted volatilities 0.35 0.3 High 0.25 0.2 0.15 0.1 Low 0.05 0 0 0.1 0.2 0.3 Realized volatilities 0.4 Joint Estimation Average predicted vs. realized returns, ten B/M portfolios, the q-theory model Average predicted returns 0.3 0.25 0.2 High Low 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Joint Estimation Predicted vs. realized stock return volatilities, ten CI portfolios, the q-theory model 0.4 High Predicted volatilities 0.35 0.3 0.25 Low 0.2 0.15 0.1 0.05 0 0 0.1 0.2 0.3 Realized volatilities 0.4 Joint Estimation Average predicted vs. realized returns, ten CI portfolios, the q-theory model Average predicted returns 0.3 High 0.25 0.2 Low 0.15 0.1 0.05 0.05 0.1 0.15 0.2 0.25 Average realized returns 0.3 Outline 1 What/Why 2 Key Results 3 Model 4 Econometric Methods 5 Matching Expected Stock Returns 6 Matching Expected Returns and Variances 7 Summary and Future Work Conclusion Summary and interpretation Derive and test the q-theory model for cross-sectional returns Portfolios of firms do a good job in aligning their investment policies with costs of equity capital, and this alignment drives many characteristics-return relations Conclusion Future Work More realistic ingredients, such as decreasing returns, investment lags, financing constraints, labor, organizational capital: Balancing realism and analytical tractability More puzzles in cross-sectional returns (momentum, asset growth, accruals, distress, M&As, net equity issues, governance, etc) Methodology applicable in dynamic corporate finance An investment-based theory of corporate bond returns Belo, Xue, and Zhang (2011): Fit cross-sectional equity valuation ratios; firm-level estimation, allow industry specific parameters