Price formation on the EuroMTS platform

Abstract

This article examines the process of price discovery in the Mercato Telematico dei Titoli di Stato (MTS) system, which builds on the parallel quoting of euro-denominated government securities on a number of (relatively large) domestic markets and on a (relatively small) European marketplace (EuroMTS). Using 27 months of daily data for 107 pairs of bonds, we present unambiguous evidence that trades on EuroMTS have a sizeable informational content.

Notes

¹See Baillie et al. (2002) for a detailed discussion of the two price discovery measures.

²Following Upper and Werner (2002), in the case of missing observations (owing to lack of transactions) we use the last available transaction price (‘fill-in’ method).

³Complete results of this section are available upon request.

⁴In three models, the rank of Π turns out to be two, which is not consistent with the conclusions from the unit root tests but confirms that condition (2) holds in these cases too.

⁵Although asymptotic intervals are not very sensitive to the assumption of normality, QQ-plots and normality tests indicate clear departures from this assumption for γ _E and ζ _E in the two samples.

⁶The ‘fill-in’ method may influence the short-term information flow for the less frequent trading marketplace (EuroMTS, in the present case) even if the trades taking place on that market do contain information (Lehmann, 2002). Accordingly, our estimated values for γ _E and ζ _E can be considered as lower bounds.

⁷By comparing the mean value of γ _E (ζ _E ) in Panel A to its counterpart in Panel B, the null of equivalence is not rejected according to asymptotic and bootstrap-based tests. Furthermore, replicating these computations for weighted quantities by traded volumes or by average number of trades over the 27-month horizon considered gives quantitatively similar results (not reported) to those in Table 2.