Financial econometrics at the high frequency

Abstract

This project focuses mainly on three lines of research. The first follows up on _Price discovery in high-dimensional arbitrage portfolios_ (FAPESP 2013/22930-0), which extended the standard price discovery analysis to a number of different settings. Fernandes & Scherrer (2018) show how to tackle the situation in which the underlying asset trades at multiple trading platforms, not necessarily using the same currency. Dias, Fernandes & Scherrer (2018) discuss how to conduct price discovery analyses in continuous time. In particular, we show that the component share measure of price discovery is invariant to the sampling interval regardless of whether the market-specific variances and covariances change over time. This is extremely convenient because one may carry out statistical inference about the continuous-time component share using lower-frequency data less prone to market microstructure noise. Dealing with a continuous-time setting is a very timely topic. Hasbrouck (2018) indeed argues for running price discovery analyses in higher resolutions now that the TAQ database time-stamps prices and quotes in nanoseconds.The second strand relates to my research on realized measures of variation and covariation. Corradi, Distaso & Fernandes (2012) show how to test for volatility transmission between markets in a nonparametric manner. Corradi, Distaso & Fernandes (2018) contemplate a similar nonparametric framework to identify jump spillovers. In both papers, the hardship lies on the fact we do not observe the realizations of the volatility and jump processes. This means that we must explicitly account for the estimation error in the derivation of the limiting distribution of the tests. The asymptotic theory is even more difficult for jump analyses because prices might not even jump at all. That is why we are very careful to design a test for jump spillovers that neither suffers from pre-test/misclassificaton bias nor depends on the presence of jumps. The idea now is to derive a nonparametric framework to spot mispricings in financial markets that accounts for the fact we have only nonparametric estimates of the risk factor loadings.The third line of research deals with high-frequency trading (HFT). Such traders run automated trading strategies that stand out by their dazzling speed. As they usually avoid holding nonzero positions for more than a few seconds, it is standard to view them as intermediaries. The literature on HFT has been gaining momentum very fast in the recent times, tackling topics such as speed dispersion in limit-order markets, preying behavior, market fragmentation, endogenous quote flickering, and determinants of order cancellations, among others. Regardless of the specific theme, the underlying question is always the same, namely, how HFT affects market liquidity and quality? I am starting to work on this literature using a unique data set from B3 in order to better understand how HFT affects not only market quality and liquidity, but also the price discovery mechanism. (AU)