Whole Period Model

3. Regression Results

Panel regression is used to consider cross-sectional and time-series effects.

3.1. Whole Period Model

Table 3 and Table 4 report the panel regression results of the whole sample model for EPS and EVA. Several noticeable empirical results are significant, based on Tables 3 and 4.

First, by comparing Table 3 and Table 4, we find that the explanatory power of EVA is

dominant over those of EPS in every model, (Average adj-R²：0.261 > 0.211，0.414 > 0.399，

0.583 > 0.560); therefore, it confirms H1: EVA is more informative due to lower information asymmetry.

Furthermore, although EPS in Table 3 is positively related to total returns in

omitted-collinearity-adjusted models, EPS is not significant and negatively related with firm value, especially in some models whose dependent variable is stock return (Ri). This

19 CAPM derived ICOE (ri) = Rf +βi × [E(Rm) － Rf] (2)；Industry betas (market value weightedβi within a single industry ) for 19 industries (classified by SIC division) are once conducted; however, it does not seem to improve the empirical results. Therefore, it is omitted in the final table.

finding corresponds with prior studies claiming that EPS may be a misleading indicator of returns, likely because EPS can affect investment decisions due to earning management problems (Stern, 1974; Stewart & Jones, 2001)²⁰

Second, we can observe the increasing explanatory power as we gradually change the dependent variable from stock return to excess return 2 (Average adj-R²：0.211 < 0.399 <

0.560；0.261 < 0.414 < 0.583). This finding illustrates the improvements made by adding Fama and French risk factors on valuation in the market perspective.

. In addition, this may illustrate some occasions in which valuations from operating and market perspective move in

counter-directions such as dot-com companies during Internet bubbles, which might encounter long-term loss on earnings but still receive high valuations and external funds from the market. In addition, as Table 4 shows, EVA is nearly negatively related to stock returns (Ri) in every case, suggesting that EVA may illustrate more behavioral finance effect.

Table 3 Panel Regression — Whole Sample Model for EPS

This table reports the regression coefficients but omits the associated t-statistics from the panel regression model with fixed effect and no intercept settings for whole sample period 1994 to 2009. The number of total sample firm-years (N) is 1760 for every model. F-statistics of validity tests of panel regression models are also provided.

Three dependent variables, including stock returns, excess stock returns 1, and excess stock returns 2, are all used in each model, and earning-based valuation indicator EPS is used as the major independent variable. The table summarizes all model combinations composed of controlled variables: size, tangibility 1, tangibility 2, profitability, investment, Tobin’s Q, current ratio times interest earned, and debt ratio. (The detailed definition is listed in Table 1.) For example, Model 1 indicates three regression results conducted by three dependent variables explained by EPS as major independent variable plus first combinations of controlled variables in the model. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively.

20 Some also suggest that the negative coefficients may due to long-run reversal condition of returns, because individual beta is related to previous returns.

Panel A Model 1 Model 2 Model 3

Table 4 Panel Regression — Whole Sample Model for EVA

This table reports the regression coefficients but omits the associated t-statistics from the panel regression model with fixed effect and no intercept settings for the whole sample period from 1994 to 2009. The number of total sample firm years (N) is 1760 for every model. F-statistics of validity tests of panel regression models are also provided. Three dependent variables, including stock returns, excess stock returns 1, and excess stock returns 2, are all used in each model, and earning-based valuation indicator EVA is used as the major independent variable.

The table summarizes all model combinations composed of controlled variables: size, tangibility 1, tangibility 2, profitability, investment, Tobin’s Q, current ratio times interest earned, and debt ratio (the detailed definition is listed in Table 1.) For example, Model 1 indicates three regression results, which are conducted by three dependent variables explained by EVA as major independent variable plus the first combinations of controlled variables in the model.

*, **, and *** indicate significance at 10%, 5%, and 1%, respectively.

Panel A Model 1 Model 2 Model 3

Panel A (continued) Model 7 Model 8 Dependent Variable (1) Ri (2) ER1 (3) ER2 (1) Ri (2) ER1 (3) ER2

EVA -0.021 *** 0.063 *** 0.08 *** -0.021 *** 0.062 *** 0.079 ***

ln(TA) -0.09 *** 0.776 *** 0.542 *** -0.087 *** 0.715 *** 0.482 ***

Tang1 -0.057 0.321 0.421

Tang2 -0.008 -0.946 *** -0.804 ***

Investments

Q 0.111 *** 0.097 *** 0.1 *** 0.111 *** 0.095 *** 0.099 ***

Profitability -0.002 -0.065 *** -0.05 *** -0.002 -0.059 *** -0.045 ***

Current Ratio -0.013 0.068 * 0.024 -0.013 0.096 ** 0.049

TIE

Debt Ratio 0.000 -0.002 * -0.001 0.000 -0.002 * -0.001

Dum09 0.296 *** 3.623 *** 0.471 * 0.295 *** 3.625 *** 0.415

F-stat. 1.43 *** 2.54 *** 2.23 *** 1.7 *** 3.28 *** 2.82 *** Avg. Avg. Avg.

Adj. R-square(%) 0.318 ^{0.406 0.579 0.318 0.409 0.581 0.261 0.414 0.583}

Third, we compare from the figures Table 3 to Table 4 and the expectation of variables in Section 2.2.

Coefficients of financial constraints are partly consistent with the expectation (Opler &

Titman, 1994; Maury, 2006; Lee & Chuang, 2009); thus, significance is achieved. Among three measures for financial constraints (i.e., inverse measures), tangibility is positively related to stock returns, but negatively related to excess returns, whereas size is negatively related to stock returns and positively related to excess stock returns. These results may hint that small firms facing lower financial constraints tend to have higher stock returns, and large firms facing greater financial constraints tend to have more excess stock returns and vice versa, but we cannot conclude that strongly. Hahn & Lee, 2009 state that size effect cannot fully explain the difference created by financial constraints, even after considering FF factors. Even, in their study, they find that especially when under constraint-group (by lots of measures, such size, tangibility, bond rating…,etc), bigger tangibility is still more related to bigger excess returns significantly, and our results is similar to theirs.

Coefficients of growth of investments and profitability are not fully consistent with the expectation (Hahn & Lee, 2009; Fama & French, 1992). Significance is achieved in most cases, except in some models of Ri. Some models are negatively related to stock returns,

and even more negatively related to excess returns. This situation does not alter models avoiding collinearity. Possibly, for manufacturing firms in S&P 500 index members, firms making less investment tend to have higher stock returns. In addition, after considering corresponding opportunity cost of equity funds, firms making less investment tend to earn even higher excess stock returns.

On the one hand, coefficients of current ratio and debt ratio are mostly consistent with the expectation, although significance is only achieved with some models. On the other hand, this significantly improves models avoiding collinearity. Current ratio is positively related with stock returns (Menon, 1987; Richards, 1980; Donaldson, 2000), whereas debt ratio is negatively related with stock returns (Opler & Titman, 1994; Majumdar & Chhibber, 1999; Weill, 2008); however, times interest earned is not significant in all cases.

In sum, our empirical results illustrate that, for manufacturing firms in S&P 500 index members, small firms with less financial constraints, less investment, less profitability, higher current ratio, and less debt ratio tend to have significantly higher stock returns, whereas large firms facing greater financial constraints, less investments, less profitability, higher current ratio, and less debt ratio tend to have higher excess stock returns significantly.