Contents lists available at ScienceDirect International Review of Economics and Finance journal homepage: www.elsevier.com/locate/iref Overnight returns of stock indexes: Evidence from ETFs and futures ☆ Qingfu Liu a , Yiuman Tse b, ⁎ a Fudan University, China b University of Missouri – St. Louis, United States ABSTRACT During the period 1999–2014, overnight returns of US exchange-traded index funds and most international index futures are significantly positive, while returns during trading hours are negative. The overnight volatility is lower than the trading volatility. Estimating the value at risk and expected shortfall by incorporating the daytime and overnight returns into a joint distribution with a copula method, we find that the risk contribution of trading hours is substantially higher than that of nontrading hours. The results are not consistent with the usual risk-return tradeoff. For US ETF and futures markets, we also show that overnight returns can forecast both the in-sample and out-of-sample returns during the first half-hour (with a negative relation) and last half-hour (with a positive relation) of trading hours. 1. Introduction Previous studies (Berkman, Koch, Tuttle, & Zhang, 2012; Branch & Ma, 2012; Cliff, Cooper, & Gulen, 2008) have showed that US stock returns overnight are significantly positive, while returns during trading hours are insignificant or negative. However, it is widely reported that the variance of overnight returns is lower than that of daytime returns (French & Roll, 1986; Lockwood & Linn, 1990). We confirm these results with US exchange-traded funds (ETFs) of the S & P 500 and nine sectors and 12 international index futures during the period 1999–2014. We do not use the underlying cash indexes, which are not tradable, and both the open and close prices may be biased by the nonsynchronous trading problem. Futures markets with low transaction costs, a narrow bid-ask spread, and having no short-selling constraints also facilitate our examination of return distributions. If the positive overnight returns are driven by positive information, how could the overnight variance be lower than the daytime variance? As Ross (1989) points out, variances are directly related to information flow. Branch and Ma (2012) attribute the positive overnight stock returns and subsequent negative daytime returns to the trading strategy of market makers. However, all the futures markets use electronic trading with no designated market markets, and most of the ETF trading on the New York Stock Exchange (NYSE) is handled electronically. The theory of attention-triggered retail buying at the start of the trading day proposed by Berkman et al. (2012) cannot explain the current results of overnight and daytime returns. Berkman et al. examine US stocks and find that attention measures (proxied by the previous day's close-to-close return volatility and other variables) are correlated positively with overnight returns and negatively http://dx.doi.org/10.1016/j.iref.2017.01.005 Received 21 September 2016; Received in revised form 1 November 2016; Accepted 3 January 2017 ☆ We thank three anonymous reviewers for valuable comments. Liu acknowleges financial support from the National Nature Science Funds of China (71473042). An earlier version of the paper was completed when Tse visited Fudan University in 2015. ⁎ Correspondence to: College of Business Administration, University of Missouri- St. Louis, St. Louis, MO 63121, United States. E-mail address: tseyi@umsl.edu (Y. Tse). International Review of Economics and Finance 48 (2017) 440–451 Available online 03 January 2017 1059-0560/ © 2017 Elsevier Inc. All rights reserved. MARK