Time scales involved in market emergence

TL;DR

This paper investigates the time scales of market formation by analyzing high-frequency data, revealing how correlations develop over different time horizons and noting that significant correlations emerge faster in recent markets.

Contribution

It provides a comparative analysis of correlation eigenvalues across multiple time scales and markets, highlighting the dynamics of market emergence and the influence of the Epps effect.

Findings

Largest eigenvalue increases with time scale

Significant correlations emerge faster in modern markets

Correlation development is influenced by multiple factors

Abstract

In addressing the question of the time scales characteristic for the market formation, we analyze high frequency tick-by-tick data from the NYSE and from the German market. By using returns on various time scales ranging from seconds or minutes up to two days, we compare magnitude of the largest eigenvalue of the correlation matrix for the same set of securities but for different time scales. For various sets of stocks of different capitalization (and the average trading frequency), we observe a significant elevation of the largest eigenvalue with increasing time scale. Our results from the correlation matrix study go in parallel with the so-called Epps effect. There is no unique explanation of this effect and it seems that many different factors play a role here. One of such factors is randomness in transaction moments for different stocks. Another interesting conclusion to be drawn from our results is that in the contemporary markets the emergence of significant correlations occurs on time scales much smaller than in the more distant history.