본고는 2000년대 중반이후 외국인 주식 및 채권 투자행태가 어떻게 변화했는가를 실증 분석한다. 2006년 이후 주별 시계열을 사용하여 수익률(국내외 장기 및 단기 금리차) 및 위험도(선진국 및 신흥국 위험지표)를 설명변수로, 외국인 증권(주식+채권)순매수를 종속변수로 하는 단일선형모형을 이용했다. 글로벌 금융위기 (GFC)를 유일한 투자행태 변환점으로 가정하는 기존 연구와 달리 GFC 이후 기간에서 경제일지 및 통계적으로 유의한 구조적 변환점을 찾아 네 개의 하부 표본을 적시하였다. 각 표본별로 외국인 주식과 채권순매수 모형을 OLS와 GMM으로 추정했다. 위험도 변수들이 더 유의한 패턴을 보였으나 표본에 따라 금리차도 유의성을 보였다. 주식과 채권순매수가 동조화 되는 표본에서 설명변수들이 더 유의한 결과를 보였다. 2015년 11월 이후 2.5개월 기간을 이용하여 표본 외 예측 (out-of-sample)을 하여 모형의 예측력을 검토하였다.
This paper empirically examines how the investment patterns of foreign investors in Korean stocks and bonds have changed since the mid 2000s. Linear equations of foreign net purchases of stocks and bonds were estimated using weekly data since 2006 of the following variables; market risk measures of emerging and advanced economies, and interest rate spreads of both long- and short-term interest rates of Korea and the US, and the won-dollar exchange rate. The Bai-Perron method was applied to endogenously identify three structural breaks in the linear relationship over the whole sample period, unlike most empirical analyses that assume an exogenous break in the investment relationships around 2008. An identical model was estimated from each sub-samples employing both OLS and GMM. Risk measures remain significant more consistently across different samples than other explanatory variables. Explanatory variables were uniformly more significant in the sample in which foreigners’ net purchases of Korean securities showed a tighter co-movement. The end of sample period since November 2015 was used for an out-of-sample dynamic forecast exercise to evaluate the predictive power of select models.
This paper empirically examines how the investment patterns of foreign investors in Korean stocks and bonds have changed since the mid 2000s. Linear equations of foreign net purchases of stocks and bonds were estimated using weekly data since 2006 of the following variables; market risk measures of emerging and advanced economies, and interest rate spreads of both long- and short-term interest rates of Korea and the US, and the won-dollar exchange rate. The Bai-Perron method was applied to endogenously identify three structural breaks in the linear relationship over the whole sample period, unlike most empirical analyses that assume an exogenous break in the investment relationships around 2008. An identical model was estimated from each sub-samples employing both OLS and GMM. Risk measures remain significant more consistently across different samples than other explanatory variables. Explanatory variables were uniformly more significant in the sample in which foreigners’ net purchases of Korean securities showed a tighter co-movement. The end of sample period since November 2015 was used for an out-of-sample dynamic forecast exercise to evaluate the predictive power of select models.