Model Specifications

Three different general specifications are used for the ordered logistic re- gressions, and two different models are used for each general specification. The first general specification is the baseline specification for the indepen- dent variables which is used to generate the primary results reported in this study. Within the baseline specification for the independent variables there are two models which are used to perform the ordered logistic re- gressions. The first model includes scaled advertising (AD), scaled R&D (R&D/A), scaled net fixed assets (PPENT), scaled intangible assets (IN- TAN), scaled cash flow (CF), scaled accruals (ACC), investor sentiment (Sent), several interaction terms between firm characteristics and investor sentiment (AD*Sent, R&D*Sent, PPENT*Sent, INTAN*Sent, CF*Sent, and ACC*Sent), Altman’s Z-Score (AltZ1), firm age (Age), firm size (Size), the standard deviation of returns (StdRet), stock turnover (Turnover), the standard deviation of return on assets (StdROA), the stock’s β (Beta), return on assets (ROA), and leverage (Leverage). The first baseline speci- fication model omits delta and vega because the limited availability of the delta and vega data significantly reduces the sample period. The model is

presented below in Equation 16:

Decilei,t =f(α+γ3ADi,t+γ4R&D/Ai,t +γ5P P EN Ti,t +γ6IN T ANi,t+γ7CFi,t+γ8ACCi,t+γ9Senti,t +γ10(ADi,t∗Senti,t) +γ11(R&Di,t ∗Senti,t)

+γ12(P P EN Ti,t∗Senti,t) +γ13(IN T ANi,t∗Senti,t) +γ14(CFi,t∗Senti,t) +γ15(ACCi,t∗Senti,t)

+γ16AltZ1i,t+γ17Agei,t+γ18Sizei,t

+γ19StdReti,t +γ20T urnoveri,t+γ21StdROAi,t

+γ22Betai,t+γ23ROAi,t+γ24Leveragei,t +εi,t) (16) The second model adds delta (Delta) and vega (Vega) to the model pre- sented in Equation 16. The model which results is presented below in Equation 17:

Decilei,t =f(α+γ1V egai,t+γ2Deltai,t+γ3ADi,t

+γ4R&D/Ai,t+γ5P P EN Ti,t +γ6IN T ANi,t +γ7CFi,t +γ8ACCi,t+γ9Senti,t+γ10(ADi,t∗Senti,t)

+γ11(R&D/Ai,t∗Senti,t) +γ12(P P EN Ti,t∗Senti,t) +γ13(IN T ANi,t∗Senti,t) +γ14(CFi,t∗Senti,t) +γ15(ACCi,t∗Senti,t) +γ16AltZ1i,t

+γ17Agei,t+γ18Sizei,t +γ19StdReti,t

+γ20T urnoveri,t+γ21StdROAi,t+γ22Betai,t

+γ23ROAi,t +γ24Leveragei,t+εi,t) (17)

The second general specification adds single digit SIC industry fixed effects to the baseline specification models. Industry fixed effects are in- cluded in this specification because it is possible that industry effects may drive some of the overvaluation which is observed.2 _{Aside from the addi-}

tion of the single digit SIC industry fixed effects, the same combinations of variables are used for the two industry fixed effects specification models as are employed for the two baseline specification models. In the model equations for the industry fixed effects specification, Equations 18 and 19, the industry fixed effects are represented by the addition of the vector of industry dummy variables to the baseline models in Equations 16 and 17. 2_{An ordered logistic regression with firm level fixed effects has been attempted, but}

the model failed to converge.

The models used in the industry fixed effects specification are as follows:

Decilei,t =f(α+γ3ADi,t+γ4R&D/Ai,t +γ5P P EN Ti,t +γ6IN T ANi,t+γ7CFi,t+γ8ACCi,t+γ9Senti,t +γ10(ADi,t∗Senti,t) +γ11(R&D/Ai,t∗Senti,t) +γ12(P P EN Ti,t∗Senti,t) +γ13(IN T ANi,t∗Senti,t) +γ14(CFi,t∗Senti,t) +γ15(ACCi,t∗Senti,t)

+γ16AltZ1i,t+γ17Agei,t+γ18Sizei,t

+γ19StdReti,t +γ20T urnoveri,t+γ21StdROAi,t +γ22Betai,t+γ23ROAi,t+γ24Leveragei,t

+θIndDU M +εi,t) (18)

Decilei,t =f(α+γ1V egai,t+γ2Deltai,t+γ3ADi,t

+γ4R&D/Ai,t+γ5P P EN Ti,t +γ6IN T ANi,t +γ7CFi,t +γ8ACCi,t+γ9Senti,t+γ10(ADi,t∗Senti,t)

+γ11(R&D/Ai,t∗Senti,t) +γ12(P P EN Ti,t∗Senti,t) +γ13(IN T ANi,t∗Senti,t) +γ14(CFi,t∗Senti,t) +γ15(ACCi,t∗Senti,t) +γ16AltZ1i,t

+γ17Agei,t+γ18Sizei,t +γ19StdReti,t

+γ20T urnoveri,t+γ21StdROAi,t+γ22Betai,t

+γ23ROAi,t +γ24Leveragei,t+θIndDU M +εi,t) (19) The third general specification incorporates several adjustments to ad- dress potential endogeneity concerns, producing the endogeneity specifica- tion models. Specifically, three issues arise which are related to the baseline specification approach described so far. First, it is possible that vega and delta could be affected by managers seeking particular compensation pack- ages when their firms are overvalued. Second, one possible concern with the scaled intangible assets in the ordered logistic regressions is that over- valued firms may engage in more acquisitions, thereby potentially creating an endogeneity problem for the scaled intangible assets. To address both potential endogeneity problems lagged values of vega, delta, and scaled intangible assets are used as the independent variables in the endogeneity specification models. Third, scaling R&D by assets may be cause for con- cern due to the fact that higher R&D spending reduces earnings, thereby reducing assets in the long run and affecting both the earnings and book value components of the residual income model. To address that concern R&D spending is scaled by sales for the endogeneity specification models. Aside from the adjustments made to delta, vega, scaled intangible assets, and scaled R&D spending, the same combinations of independent variables are used for the two endogeneity specification models as are employed for

the baseline specification models. The resulting endogeneity specification models are presented below, with the endogeneity specification models in Equations 20 and 21 corresponding to the baseline specification models in Equations 16 and 17 after the endogeneity adjustments described:

Decilei,t =f(α+γ3ADi,t+γ4R&D/Si,t+γ5P P EN Ti,t +γ6IN T ANi,t−1+γ7CFi,t+γ8ACCi,t+γ9Senti,t +γ10(ADi,t∗Senti,t) +γ11(R&D/Si,t∗Senti,t)

+γ12(P P EN Ti,t∗Senti,t) +γ13(IN T ANi,t−1∗Senti,t) +γ14(CFi,t∗Senti,t) +γ15(ACCi,t∗Senti,t)

+γ16AltZ1i,t+γ17Agei,t+γ18Sizei,t

+γ19StdReti,t+γ20T urnoveri,t+γ21StdROAi,t

+γ22Betai,t+γ23ROAi,t+γ24Leveragei,t +εi,t) (20)

Decilei,t =f(α+γ1V egai,t−1+γ2Deltai,t−1+γ3ADi,t

+γ4R&D/Si,t+γ5P P EN Ti,t+γ6IN T ANi,t−1+γ7CFi,t +γ8ACCi,t+γ9Senti,t +γ10(ADi,t ∗Senti,t)

+γ11(R&D/Si,t∗Senti,t) +γ12(P P EN Ti,t∗Senti,t) +γ13(IN T ANi,t−1∗Senti,t) +γ14(CFi,t∗Senti,t) +γ15(ACCi,t∗Senti,t) +γ16AltZ1i,t

+γ17Agei,t+γ18Sizei,t+γ19StdReti,t

+γ20T urnoveri,t+γ21StdROAi,t+γ22Betai,t

+γ23ROAi,t+γ24Leveragei,t +εi,t) (21)

The baseline, industry fixed effects, and endogeneity specifications are used to generate ordered logistic regression analyses of the determinants of mispricing. In order to generate the ordered logistic regression analyses of the determinants of mispricing the valuation deciles, CEO compensation characteristics, firm characteristics, sentiment variables, and control vari- ables discussed previously are used to implement the models depicted in Equations 16 through 21. Although the main focus in this study is upon the baseline specification encompassing the models in Equations 16 and 17, the alternative models in Equations 18 through 21 make it possible to illuminate the robustness of the findings to adjustments applied to the inde- pendent variables and models used in the analyses. In the following section the methodology discussed in this section is applied to the preparation of the independent and dependent variables and the subsequent results from the ordered logistic regression analyses are summarized and discussed.