Recently, reliability of accounting information has been one of the most controversial issues among the practices, regulatory bodies and accounting academic. This study investigates whether a firm’s governance practices has an effect on the quality of the publicly released financial information. In particular, I examined the relationship between governance scores and the extent of corporate earings quality as measured by the level of current accruals' prediction errors suggested by Dechow and Dichev(2002) model. The model estimates earnings quality as the extent to which working capital accruals map into operation cash flow realizations. This model is predicated based on the idea that earings quality is affected by the measurement error in the process of working capital accruals' realization into operating cash flow. Greater portion of accruals unrealized into cash flows indicates the lower quality of earings. Data sources are the governance practice scores evaluated by the Korean Corporate Governance Service(CGS), primarily based on the 2003 annual reports of the firms listed in the Korea Exchange. The CGS has been established to help the companies listed in the Korea Exchange to improve the governance level after the 1997 Korea financial crisis. The CGS annually evaluates all the listed firms’ governance index scores and presents an award to the outstanding firms. The index composes of five subindices including several governance elements: (1) Shareholder Rights, (2) Board Structures, (3) Disclosure, (4) Internal Audit Function, (5) Distribution of Earnings. I excluded the scores concerning Internal Audit Function because they are mainly dealing with only large firms with audit committee. In this paper, a research hypothesis is formated: there is a positive relationship between the corporate governance scores and earnings quality. The research hypothesis has been tested by the two stage least square regressions(2SLS) considering the endogeniety of corporate governance. In the first stage, a probit model is formulated to identify the factors affecting the governance practice level. The independent variables of the probit model are firm size, firm risk, capital intensity, investment in R&D and advertising, ROA, leverage, controlling shareholder’s equity ratio, firm size dummy indicating whether total asset amount exceeds 20 trillion won, and industry factors. They are selected based upon Black et al.(2005a) model. The probit model predicts governance probability level for each firms. In the second stage, the governance probabilities estimated in the first stage is used as the level of experiment variable. Several control variables are included in the second stage regression model: firm size, variability of cash flows from operation, variability of sales, length of operation cycle, and frequency of losses based on Dechow and Dichev(2002) study. The second stage regression models are composed of both reduced and full ones. Reduced models’ experiment variables are each of the subindices governance scores: score 1 (Stockholder Rights), score 2(Board Structures), score 3(Disclosure Level), and score 4(Distribution of Earnings), conscore(composite scores including all the subindices scores). This reduce model determines whether each of subindices score, if any, affects the earings quality. The full model’s experimental variables are the four subindices scores to determine their incremental effect on the level of earings quality. The results of this study report that firms with higher level of disclosure and more earnings distribution are likely to have the high quality of earnings, supporting the research hypothesis. This results indicate that firms disclosing detailed information, hence having more management transparency and well protecting investors’ dividends right report higher quality of earnings. This evidence implies that corporate governance contributes to enhancing the reliability level of accounting information, theoretically underpinning the empirical evidence that disclosure level lowers the cost of equity capital.