본 논문은 한국거래소(KRX)에 상장된 상장지수펀드(ETF)를 대상으로 가격효율성을 검증하기 위해 추적오차와 괴리율을 분석한다. 2002년 12월 23일부터 2009년 12월 30일까지 35개 ETF 의 일별 자료를 사용해 분석한 본 논문의 주요 결과는 다음과 같다. 첫째, NAV와 벤치마크 지수의 수익률 차이로 정의된 추적오차는 분배금 지급, 벤치마크 지수의 변동성 및 구성종목 변경 빈도에 따라 커지지만 예상과는 달리 운용사의 운용능력에 의해 영향을 받지 않는다. 둘 째, ETF 가격과 NAV 차이의 절대값으로 정의된 괴리율은 보수율, ETF 가격의 일중 변동성이 클수록 높게 나타난다. 또한 개인투자자의 거래비중이 높을수록 그리고 유동성이 작을수록 괴리율이 크게 나타난다. 이 결과는 본질가치 대비 가격의 괴리 현상에 대한 노이즈 투자자 및 유동성에 기반한 가설을 모두 설명한다. 셋째, 괴리율을 이용한 차익거래 기회는 일중에서 완전히 제거되지 않고 그 다음날에 가서야 해소된다. 요컨대 본 논문의 연구결과는 분배금, 보수율과 같은 제도적 요인과 ETF 시장참여자의 특성에 따라 추적오차나 괴리율이 커질 수 있음을 시사한다. 따라서 ETF의 시장효율성을 제고하기 위해 유동성을 높이고 기관투자자의 참여를 유도하는 등 제도적 정비가 요구된다.

There have been long debates about whether stock prices reflect their intrinsic value or instead the psychological outgrowth of irrational investors. The former is based on Friedman`s (1953) argument that prices converge to intrinsic value because arbitragers` activities eliminate price deviation from intrinsic value produced by irrational investors; the latter is based on Keynes`s (1936) argument that prices are moved by ignorant individuals` mass psychology. Since DeLong, Shleifer, Summers, and Waldmann 1990), however, the view that price deviations from intrinsic value may persist for a long time and the deviations may come from the noise-trader risk has become more common. As an example, if noise-trader risk is systematic, this will be priced into the equation and make arbitrage trading troublesome. This argument by DeLong et al. is strongly supported by Lee, Shleifer, and Thaler (1991) and Pontiff (1996) who study closed-end fund discount puzzle. Separately, there has been widespread discussion about the effect of liquidity on an asset`s equilibrium price (e.g. Amihud, Mendelson, and Pederson, 2005). Applied to the fund, it implies that an illiquid fund is underpriced and therefore provides high expected returns. Cherkes, Sagi, and Stanton (2009) show that the closed-end fund discount is not from irrational investors but illiquidity using their liquidity-based theoretical model. To sum up the research on the fund, we are able to reach the following hypotheses : Both noise-trader risk and illiquidity are systematic risk factors and therefore they make prices deviate from intrinsic value. Hence, if these risk factors are incorporated into an asset`s market price and in turn it becomes below intrinsic value, the asset (e.g. fund) will be illiquid and individual investors` trading activities on the asset will be high. Meanwhile, if proportion of trading volume is dominated by individual investors while trading volume is high, then the level of price deviations will not be predictable. This paper tests the aforementioned hypotheses in the exchange- traded fund (ETF) market. That is, we analyze whether noise traders, liquidity, or both drive the deviation between an ETF `s market price and intrinsic value. For the analysis, we use "tracking error," which is the difference between net asset value (NAV) and benchmark index returns, and "differentials," which are the discount or premium between NAV and ETF market price; we investigate the cross-sectional and time-series differences of these two measures among ETFs. Using the daily data of 30 domestic ETFs and 5 foreign ETFs from December 23, 2002 to December 30, 2009, we find the following results. First, tracking errors increase as the dividend payment, the volatility of benchmark index, or the frequency of change of constituent stocks increases. Second, foreign ETFs` tracking errors are higher than domestic ETFs.` However, the management abilities of asset management companies and the difference in domestic ETFs` benchmark types do not affect tracking errors statistically significantly. Third, the differentials increase as market capitalization and liquidity decrease and the intraday volatility of ETF price and the proportion of individual investors` trading increase. Differentials are the smallest where liquidity is high and the proportion of individual investors is low, and it is the largest where liquidity is low and the proportion of individual investors is high. This finding is consistent with both the liquidity-based explanation and noise-trader-risk-based explanation for price deviation from intrinsic value. In addition, expense ratio and ex-dividends are the factors that widen differentials. Fourth, we find that when the ETF price closes higher than its NAV, the ETF price declines on the next day; when the ETF price closes lower than its NAV, the ETF price increases on the next day. This implies that arbitrage opportunities from differentials are not eliminated intra-daily but rather inter-daily. This paper shows the tracking errors and differentials for ETF prices are affected by trader characteristics and liquidity as well as policy factors such as dividends and expense ratios. Our results imply that it is necessary to augment liquidity and institutional investors` participation to facilitate arbitrage trading and thereby enhance market efficiency in the ETF market. To elevate ETF liquidity, it is important to streamline the requirement standards and evaluation of liquidity provider (LP) and to give incentives to pension funds to participate in the ETF market. Finally, considering that ETF trading tends to gravitate into certain ETFs, strengthening retail- investor education and promotion activities is necessary to expand market attention to a variety of ETFs.