기본변수모형의 이익예측력 비교: 구조적 접근과 경험적 접근 = Fundamental Variables Predictability of Future Earnings: Structure-based Approach and Experience-based Approach

다국어 초록 (Multilingual Abstract)

Research on fundamental analysis has reported the usefulness of accounting information on financial statement to predict future earnings which are deemed to be the most important determinant of firm value. Theses studies assume that information in general purpose financial statement can help investors to forecast earnings and firm`s intrinsic value which, then, can be compared to the observed market prices. In this way, fundamental analysis can help investors separate ex-post winners from losers based on the financial statement information. Following the lead of Ou and Penman(1989), many studies used accounting variables to predict future earnings and future returns, assuming that if the market is not efficient in terms of timely reflecting the value-relevant information into stock prices, then better forecasts of earnings will predict future returns. For example, Lev and Thiagarajan(1993) selected 12 accounting variables that are used by analysts in firm value determination process. They reported that these variables are associated with contemporaneous returns, after controlling for current earnings, firm size, and macro-economic conditions. With regard to the market` efficiency in reflecting value-relevant information, Abarbanell and Bushee(1997, 1998) examined the ability of Lev and Thiagarajan(1993)`s variables to predict future revisions in analysts` earnings forecasts and future returns. They reported that these variables can explain future revisions of analysts` earnings forecasts and that the trading strategy based on these variables generate significant abnormal returns. Piotroski(2000) applied fundamental analysis to high book-to-market ratio(or value firms), assuming that these firms tend to be neglected by the market participants and do not timely reflect financial statement information. He reported that financial statement information can be used to separate winners from losers in terms of excess returns. Mohanram(2005) applied fundamental analysis to low book-to-market ratio(or glamour firms) and reported that measures tailored to growth firms help to separate ex-post winners from losers in terms of excess returns. He also argued that growth oriented fundamental variables are effective for low book-to-market ratio firms. In this way, many studies have focused on the usefulness of fundamental analysis in predicting future earnings and future returns. However, due to the lack of a theory relating fundamental variables to firm value, different studies employed different fundamental variables in the analysis and many correlated variables were included in the predictive regression for future earnings and future returns. Recently, Penman and Zhang(2006) took a structured approach in selecting the fundamental variables based on residual income valuation model. Specifically, they examined the predictive ability of the components of return on net operating assets(RNOA) for the one-year-ahead changes in RNOA. In comparison to previous studies, their variables selection was theoretically guided by incorporating the accounting structure involved in earnings measurement. Subsequent studies reported that the structure based variables are useful in predicting future earnings changes. For example, Wieland(2011) employed the Penman and Zhang`s variables to identify analysts` earnings forecasts that correctly and incorrectly predict an earnings increase. Also Wahlen and Wieland(2011) reported that Penman and Zhang`s variables can be used to identify companies with a greater likelihood of future earnings increases. However, whether the theoretically sound fundamental variables are superior to traditional experience-based fundamental variables remains an empirical question. The purpose of this study is to compare the predictive ability of two fundamental variable models: the experience based model by Lev and Thiagarajan(1993) and the structure based model by Penman and Zhang(2006). This comparison is motivated by the importance of the accurate predictive ability of fundamental variables in investment decisions. We design our empirical tests with the goal of identifying a fundamental variable model with more accurate predictive ability for future earnings. This identification is obtained by comparing the predictive ability of individual and aggregate fundamental variables between the two models. The comparison was performed using 6 variables in both approaches. Each year between 2000 and 2013, we identify non-financial December year-end firms with sufficient earnings and fundamental variables data on KOSPI market. Earnings and fundamental variables data are obtained from New-KISVALUE database. In order to compare the two fundamental variables model, we need sample firm-years that have both fundamental variables. This selection procedure yields the final sample of 4,189 firm-years. Empirical results from both ordinary least square regression and logit regression indicate that both fundamental variables have information contents for future earnings. However, the structure based model outperforms experience based model in terms of accurately predicting future earnings changes. The adjusted  in the structured approach is higher than the one in the experience-based approach. The Z statistics of Vuong test and M statistics of Clarke test indicate that the difference is significant, respectively. While all variables of the structured approach except for profit margin(PMC) indicated statistical significance in predicting future earnings changes, only inventory(INVC) variable showed statistical significance in the experience-based approach. We also employed a summary measures of individual variables in order to estimate the overall informativeness of the fundamental variables. The summary measures were estimated by adding all individual probability of earnings increase from LOGIT regression, each year using a rolling 5 year period. A low(high) summary measure represents a firm with very few (mostly) good news. The results remains unchanged. The results remains unchanged after we controlled the heteroskedasticity and time-series autocorrelation problems. In short, in comparison to the experience-based fundamental variables, the structure-based fundamental variables with a solid theoretical basis appear to have a better forecasting ability for future earnings.

참고문헌 (Reference)

• 1 나종길, "기본변수의 정보성과 이익발표 후 잔류현상" 한국회계학회 38 (38): 355-390, 2013
• 2 나종길, "구조적 접근에 근거한 기본변수들의 정보성" 한국회계정보학회 31 (31): 33-60, 2013
• 3 Piotroski, J., "Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers" 38 (38): 1-41, 2000
• 4 Holthausen, R., "The Prediction of Stock Returns using Financial Statement Information" 15 : 373-411, 1992
• 5 Lipe, R, "The Information Contained in the Components of Earnings" 24 : 37-64, 1986
• 6 Durbin, J., "Testing for Serial Correlation in Least Squares Regression, II" 38 : 159-179, 1951
• 7 Durbin, J., "Testing for Serial Correlation in Least Squares Regression, I" 37 : 409-428, 1950
• 8 Stober, T. L, "Summary Financial Statement Measures and Analysts' Forecasts of Earnings" 15 : 347-372, 1992
• 9 Breusch, T. S., "Simple Test for Heteroskedasticity and Random Coefficient Variation" 47 (47): 1287-1294, 1979
• 10 Thompson, S. B, "Simple Formulas for Standard Errors that Cluster by both Firm and Time" 99 : 1-10, 2011
• 11 Fama, E. F., "Risk, Return., Equilibrium : Empirical Tests" 81 (81): 607-636, 1973
• 12 Penman, S, "Return to Fundamentals. Journal of Accounting" 7 : 465-483, 1992
• 13 Nissim, D., "Ratio Analysis and Equity Valuation : From Research to Practice" 6 (6): 109-154, 2001
• 14 Penman, S., "Modeling Sustainable Earnings and P/E Ratios with Financial Statement Analysis" Columbia University 2006
• 15 Vuong, Q, "Likelihood Ratio Tests for Model Selection and Non-nested Hypotheses" 57 : 307-333, 1989
• 16 Wieland, M. M., "Identifying Consensus Analysts' Earnings Forecasts that correctly and incorrectly predict an Earnings Increase" 38 (38): 574-600, 2011
• 17 Lev, B., "Fundamental Information Analysis" 31 (31): 190-214, 1993
• 18 Abarbanell, J. S., "Fundamental Analysis, Future Earnings., Stock Prices" 35 (35): 1-24, 1997
• 19 Ou, J., "Financial Statement Analysis and the Prediction of Stock Returns" 11 : 295-330, 1989
• 20 Greene, W. H., "Econometric Analysis" Prentice Hall 2003
• 21 Sloan, R, "Do Stock Prices fully reflect Information in Accruals and Cash Flows about Future Earnings?" 71 (71): 289-315, 1996
• 22 Sloan. R. G, "Discussion of : "Contextual Fundamental Analysis through the Prediction of Extreme Returns"" 6 : 191-195, 2001
• 23 Wooldridge, J. M., "Cengage Learning" South-Western 2012
• 24 Kothari, S. P, "Capital Market Research in Accounting" 31 : 105-231, 2001
• 25 Wahlen, J. M., "Can Financial Statement Analysis beat Consensus Analysts' Recommendations?" 16 (16): 89-115, 2011
• 26 Freeman, R., "Book Rate-of-Return and Prediction of Earnings Changes : An Empirical Investigation" 20 (20): 639-653, 1982
• 27 Anderson, M., "Are Selling, General" 41 : 47-63, 2003
• 28 Dechow, P., "An Empirical Assessment of the Residual Income Valuation Model" 26 : 1-34, 1999
• 29 Frankel, R., "Accounting Valuation, Market Expectation and Cross-Sectional Stock Returns" 25 : 283-319, 1998
• 30 Penman, S., "Accounting Conservatism, the Quality of Earnings., Stock Returns" 77 (77): 237-264, 2002
• 31 Fairfield, P., "Accounting Classification and the Predictive Content of Earnings" 71 (71): 337-355, 1996
• 32 Richardson, S., "Accounting Anomalies and Fundamental Analysis: A Review of Recent Research Advances" London Business School 2010
• 33 Abarbanell, J. S., "Abnormal Returns to a Fundamental Analysis Strategy" 73 (73): 19-45, 1998
• 34 Na, C. K, "A Stuty on the Accrual Levels and Information Contents of Fundamental Variables" 22 (22): 261-288, 2000
• 35 Newey, W. K., "A Simple, Positive Semi-definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix" 55 (55): 703-708, 1987
• 36 Clarke, K. A, "A Simple Distribution Free Test for Nonnested Hypotheses" 15 (15): 347-363, 2007
• 37 White, H, "A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity" 48 (48): 817-838, 1980