Asiasn review of Financial research

Past Issues

HOMEPast Issues Past Issues

Asian Review of Financial Research Vol.26 No.3 pp.311-351
A Construction of the Korean Stock Database and its Applications
Jangkoo Kang Professor, Graduate School of Finance & Accounting, College of Business, KAIST
Deokhyeon Lee Ph.D. Candidate, Department of Management Engineering, College of Business, KAIST
Changjun Lee Assistant Professor, College of Business Administration, Hankuk University of Foreign Studies
Jejoon Choi* Researcher, Financial Engineering Research Center, KAIST
Key Words : CAPM,Security Prices,Holding Period Return,Dividend Adjusted Return,CAPM,Fund Performance Evaluation


Up to now, major Korean data vendors have provided financial researchers with partially adjusted stock price data. As a result, holding period returns and the number of shares outstanding are sometimes inaccurate. This is a serious problem since stock price data are the most important ingredient for financial studies. To remedy this problem, benchmarking the methodologies used in the Center for Research in Security Prices (CRSP), we construct the Korean stock price data (1999. 12~2011. 09) which fully reflect corporate actions. We then investigate how security prices affect empirical studies in financial economics, with a particular focus on the effect of cash dividends. To begin with, we compute holding period returns and the number of shares outstanding by taking into account the effect of any corporate action. Though our approach is based on the methodology of the CRSP, it is modified to reflect the Korean stock market practices. By crosschecking several corporate event data sources, we build fully adjusted stock price data. The construction procedures are as follows. First, we categorize corporate events into nine large groups: Cash dividends, liquidations and transfers of exchange, mergers and splits, paid-in capital increases, capital increases without consideration and stock dividends, capital decreases and retirements of shares, stock splits and reverse stock splits, stock conversions and stock options, and other outstanding share changes. The nine large groups are subdivided into the smaller groups. Second, we examine all corporate events during our sample period. Since all corporate events except cash dividends, and liquidations and transfers of exchange entail the change of outstanding shares, we can inversely trace the corresponding corporate event when changes in outstanding shares are discovered. Although there are a few corporate actions in which outstanding shares are unchanged, those actions can be checked by the Korea exchange (KRX) corporate action data. Examples include stock split after reduction of capital stock without any refund. Third, we investigate the corresponding effective-dates of all the corporate events. Generally, there are two-or three-week gaps between the dates in which outstanding shares actually changed and the corresponding effective-dates, and this time disparity stems from related administration processes. Yet, these time gaps cause some erratic results because investors recognize that corporate events take place at the effective date but the changes of outstanding shares happen a few weeks later. Due to this time gap, firm sizes can be miscalculated. Accordingly, we revise the stock price data so that the changes of outstanding shares occur at the corresponding effective-dates and thus the gaps are removed. Basically, we check the effective-dates in Data Analysis of the Retrieval and Transfer System (DART) of Financial Supervisory Service (FSS). Also, the excess of daily limit of price variation is helpful in detecting the effective-dates because the excess indicates that there is a corporate event such as stock split and capital reduction. Fourth, we calculate corresponding holding period returns at the effective-dates. For each event, the cash dividend amount and price adjust factor are figured out. Using the values, we calculate the accurate holding period returns at each effective-date through a simple equation. Then, we finally obtain the fully adjusted stock price data. With the accurately constructed stock price data, we study how our construction of security prices affects empirical studies in financial economics. To this end, we need a proxy for the risk-free rate. In this study, we use one month spot rate of a monetary stabilization bond as a proxy for the risk-free rate. Certificate of Depository (CD) rate, the prevailing alternative of risk-free rate in Korean financial market research, bears the default risk of banks as they are the issuers of CDs. Thus, CD rate has a serious flaw to be an alternative of risk-free rate. In contrast, monetary stabilization bonds issued by the Bank of Korea do not have default risk. Therefore, we adopt one month spot rate of the monetary stabilization bond from Korea Asset Pricing (KAP) as an alternative of risk-free rate. We then concentrate on the effect of cash dividends because researchers often overlook the effect they bring on the overall outcome when they conduct research. Instead, they usually use value-weighted index of the KOSPI and KOSDAQ as market portfolio. However, since these indices do not contain cash dividend returns, considerable biases can result from relying only them. This paper, therefore, points out that such biases indeed exist and how significant those biases are. The central empirical findings are as follows: First, while the average market excess return which reflects all corporate actions is 0.97% per month, the average market excess return which does not include cash dividends is 0.82% per month. Second, the annual CAPM alphas with security prices reflecting all corporate actions are 1.70%~1.85% lower than the ones that exclude cash dividends during our sample period. Finally, the estimated annual abnormal returns of equity funds are 1.32%~1.80% lower than the abnormal returns estimated with factors that do not include cash dividends. In sum, when the security prices that exclude the cash dividends are used in empirical studies, we find that holding period returns are underestimated, and, consequently, abnormal returns are overestimated.
Export citation