More Friction, Less Noise? Relative Efficiency in a Two-Tier Stock Exchange

Abstract

If trading becomes easier and cheaper, does this attract more noise trades or more informed trades? We investigate this question in the Brussels market, where about 100 stocks were simultaneously traded in two parallel tiers, whereof one, the forward'' segment, was organizationally far superior than the tier. In an exploratory test based on Roll (1984), we write spot and forward prices as equal to a common true value, plus a market-specific settlement effect (a correction for time value until payment), plus an unpredictable noise term. We expect to find that in the forward markets the noise has lower variance, that is, the pricing is more efficient. Yet we find no such thing; in fact, by some measures the forward market may actually have been the more noisy one.Going on to the issue of price discovery, which is a more formal way of testing whether the forward market is less noisy, we extend the Margrabe-Silber price-discovery model to take into account the asynchronism in the opening forward and spot prices. Although the presence of the latent true morning return in our extended model precludes us from estimating explicitly the price adjustment coefficients, we can still identify the sign of the estimated difference of these coefficients and the lower and upper bounds of its t-statistic. This information enables us to conclude on the significance of the test. Also, our results reject the potential price discoverer status of the forward market: spot prices seem more informative than forward prices. Even more puzzlingly, the phenomenon is most pronounced for active stocks. For robustness, our Hasbrouck (1995) information-share analysis also supports this finding. This result raises the issue of how far the financial markets perform their central function of price discovery and how far the conventional wisdom can be trusted (e.g. the higher the trading volume or the lower the friction, the less noise the observed price contains).