Warp speed price moves: Jumps after earnings announcements

This repository contains Matlab code and, partly artificial (as described below), data for the paper "Warp speed price moves: Jumps after earnings announcements" to appear in Journal of Financial Economics.

The research findings in the paper are based on the following data sources:

* Bloomberg
Supplied the weights of the S&P 500 index members
* CRSP
Employed for close-to-open return-based jump test in the supplemental Appendix D
* Factiva
Provided timing of earnings announcement (to nearest minute)
* Refinitiv
Provided I/B/E/S analyst earnings information
* TAQ
Source of high-frequency transaction and quotation data
* Wall Street Horizon
Provided timing of earnings announcement (to nearest minute)
* Zacks Investment Research
Supplied the placement of earnings announcement (“Before Market Open”, “During Regular Trading”, or “After Market Close”).

The data are protected by copyright and cannot be shared in original form. Thus, to be able to run the supplied code, we generated some synthetic data for a small subset of the stock universe analyzed in the paper, i.e. Facebook (FB) and Apple (AAPL) for the period 2008 - 2012.

The following modifications of the original data were made:

The S&P 500 index weights were randomly reshuffled.

In the original tick-by-tick transaction data files, the traded prices were replaced by a Geometric Brownian motion (GBM) with an annuliazed volatility of 20%. Each day, we start the GBM in the first available transaction price, as extracted from the TAQ database, and round the GBM to two decimals. Trade sizes are drawn randomly from the set [1,2,10]. Transaction times are randomly shifted by a number of seconds drawn uniformly from the interval [-5,5] and resorted.

In the original quotation data files, the bid is simulated by a GBM, as above, and the ask is constructed from the simulated bid after a random permutation of the original quoted spread. Bid and ask sizes are drawn randomly from the set [1,2,10].

We should note that realized volatility measures are provided for ALL stocks over the WHOLE sample, which allows for replication of several results presented in the paper. However, computation of these volatility estimators requires access to NYSE TAQ data.

In the file "crsp_ea_info.mat", the "placement" indicator provided by Zack Investment Research has been randomly reshuffled.

In the file "crsp_price_info.mat", the variables "cfacpr" (correction factor), "volume", and "ntrade" were permuted, while the remaining variables can be obtained from public sources and are unaltered.