30 Chapter III Research Methodology In this section, the research methodology framework and the regression models are discussed. The research methodology section elaborates on the type of suitable panel data method analysis used in the study, the possible violations of the regression assumptions, and how to account for these violations. The regression models section discusses all the regression models tested in this study, the appropriateness of these models to test the hypotheses, included all type of variables, and validate model with classic assumption test. III.1. Research Design The research design of this study is explained by figure below. First, begin with literature review and previous research about is any influence the Cost of Debt and ESG rating score and stock return of non-financial firms to find research gap. Furthermore, arrangement and develop research objective by determined and analysis collecting the data and examine the regression model. Finally, the discussion and conclusion revealed. Fig.III.1 Research Design 31 III.2. Research Methodology Framework Based on research design above, the research methodology framework is developed by figure below: Fig.III.2 Research Methodology Framework 32 Fig.III.3 Conceptual relationship among variable in cost of debt and stock return 33 After we make an analysis for the panel data regression method we have selected that we used for fitted model are the generalized random effect model least square and continue with classic assumption test aim to fulfil The Gauss Markov theorem says that, under certain conditions, the ordinary least squares (OLS) estimator of the coefficients of a linear regression model is the best linear unbiased estimator (BLUE), that is, the estimator that has the smallest variance among those that are unbiased and linear in the observed output variables. III.3. Hypotheses Testing Based on the theoretical above, our arrange the hypothesis development for this research with fourth hypothesis that are: H1.0 The overall ESG rating (measured through ESG score) has no a significant impact on cost of debt of a firm. H1.1 The overall ESG rating (measured through ESG score) has significant impact on cost of debt of a firm. H2.0. Environmental rating (measured through scores on environment component of ESG reporting) has no a significant impact on cost of debt of a firm. H2.1 Environmental rating (measured through scores on environment component of ESG reporting) has significant impact on cost of debt of a firm. H3.0 Social rating (measured through scores on social component of ESG reporting) has no a significant impact on cost of debt of a firm. H3.1 Social rating (measured through scores on social component of ESG reporting) has a significant impact on cost of debt of a firm. H4.0 Governance rating (measured through ESG score) has no a significant impact on cost of debt of a firm. H4.1 Governance rating (measured through ESG score) has significant impact on cost of debt of a firm. 34 H5.0 The overall ESG rating score has no a significant impact on raw stock return. Any of the three pilar has significant impact to the raw stock return. H5.1 The overall ESG rating score has a significant impact on raw stock return. Any of the three pilar has no significant impact to the raw stock return. H6.0 The overall ESG rating score has no a significant impact on abnormal stock return. Any of the three pilar has significant impact to the abnormal stock return. H6.1 The overall ESG rating score has significant impact on abnormal stock return. Any of the three pilar has no significant impact to the abnormal stock return. III.4. Regression Models Our research examines the following regression for 4 model to test all hypothesis: First hypothesis: CoDi,t = a + þ1 ESG Rating i,t-1 + þ2 Struc i,t + þ3 DER i,t +þ4 ICR i,t + þ5 RoA i,t + þ6 CashCov i,t + þ7 CR i,t + þ8Prob i,t + þ9 CRisk i,t + þ10 CovidYears + εi …(1) Second hypothesis: CoDi,t = a + þ1 Env Rating i,t-1 + þ2 Struc i,t + þ3 DER i,t +þ4 ICR i,t + þ5 RoA i,t + þ6 CashCov i,t + þ7 CR i,t + þ8Prob i,t + þ9 CRisk i,t + þ10 CovidYears + εi …(2) Third hypothesis: CoDi,t = a + þ1 Soc Rating i,t-1 + þ2 Struc i,t + þ3 DER i,t +þ4 ICR i,t + þ5 RoA i,t + þ6 CashCov i,t + þ7 CR i,t + þ8Prob i,t + þ9 CRisk i,t + þ10 CovidYears + εi …(3) Fourth hypothesis: CoDi,t = a + þ1 Gov Rating i,t-1 + þ2 Struc i,t + þ3 DER i,t +þ4 ICR i,t + þ5 RoA i,t + þ6 CashCov i,t + þ7 CR i,t + þ8Prob i,t + þ9 CRisk i,t + þ10 CovidYears + εi …(4) Fifth hypothesis: Raw Returni,t = a + þ1 ESG Rating i,t-1 + þ2 Prob i,t + þ3 RoE i,t + +εi …(5) Sixth hypothesis Abn. Returni,t = a + þ1 ESG Rating i,t-1 + þ2 Prob i,t + þ3 RoE i,t + +εi ..(6) 35 III.5. Dependent Variable For the cost of debt, we use an accounting proxy given by the ratio of interest expense on debt to total debt of a corporation; both measures are available through Data stream. In this regard, we follow the same approach research by Magnanelli et al. (2012) used interest expense on debt divided by total debt as a proxy for the cost of debt. We constructed this variable to increase the sample size available. While a built-in cost of debt is available through Data stream, only a few true corporate values are actually available, thereby drastically reducing the potential sample size. ���� �� ����∶ ���������� ������ ������������� ������� ���������� ���� Cumulative raw returns are based on the definition of a one-period simple return, the time-varying variable �� can represent the simple return for a holding asset from the time interval [�−1, �] (Tsay, 2005). �������� ������ �����= ������ ������−������ ������−1 ������ ������−1 Pt represent the closing price of the stock at the end of day. Pt−1 represent the closing price of the stock at the end of the previous day. For cumulative raw return based on equation: �� ������ =(1+� 1).(1+� 2)….(1+� ������) rt represent the return on the investment for a specific time period t. Crt represent the cumulative raw return at the end of time period t. n represent the total number of time periods. Cumulative abnormal returns are based on CAPM methods to calculate these returns. The test the impact of trading activities and events on stock prices based on the buy-and-hold abnormal return (BHAR) or the cumulative abnormal return (CAR) models (Barbers et. al, 1997; Ziobrowski, et. al, 2004). More efficient stock prices benefit shareholders by reducing information imbalance and improving liquidity. The aggregate of daily abnormal return based on equation. � �������������=� ������− ř ������,(������−1) 36 Specifically, a firm’s cumulative abnormal stock return is the difference between cumulative raw return and expected return calculated based on CAPM equation. We use the expected return calculation before period the sample observation (�ř ������(������−1)).