Time-varying correlation of the stock market returns across countries in the context of international investments has been well researched in the literature in last few years. It has also been recognized that there is "volatility effect in correlation", as the stock return correlations tend to rise on high-volatility days. Recent research has however, highlighted the pitfalls of using sample correlation for comparison, particularly when the conditional volatility across samples is not same. It has been shown that the sample correlation of two independent random variables is expected to rise when the conditional volatility of the variables is high and vice-versa, even if the unconditional correlation between them is constant. Empirically, it has been long well known that the overnight (closed-market) stock returns are less volatile than the open-market returns. Making use of this regularity, we test whether the stock returns correlation are higher during trading or non-trading hours. Using five years' daily returns of 30 constituent stocks of Sensitive Index (Sensex) of The Stock Exchange, Mumbai, we find that almost all the pair-wise closed market stock return correlations are higher than the open market correlations despite lower volatility of the closed market returns. These results are further reinforced after using the univariate and bivariate conditional volatility models. Higher closed market stock return correlations are consistent with the possibility that information, which arrives during non-trading, is more on common factors. Arrival of information on common factors would imply higher stock return correlations during non-trading hours. This has implication for managing risks associated with any stock portfolio, as the diversification benefits might be over-estimated for overnight positions and under-estimated during trading hours if daily close-to-close returns are used to estimate variance-covariance matrix.
