What Pre-Open News Tells About the Same Day’s Stock Price, and What Is Left for a Trader at the Open: Evidence from Korean Equities, 2007–2026

Junyoung Jung, Yyoungseo Cho and Jingi Kim

EconStor Research Reports from ZBW - Leibniz Information Centre for Economics

Abstract: We code every pre-open news article about a listed Korean firm (BigKinds archive, 2007-2023 for the information study) into an event type and a good/bad/neutral polarity, and ask how much the code tells about the same day's abnormal return. The outcome is a 10-cell variable: the size quintile of the standardised abnormal return, with quintile cuts taken each day from that day's no-news firms, crossed with its sign. Our measure is the cross-fitted log-score gain of a code-conditional predictor over a no-news baseline distribution, in bits per news firm-day. With rules frozen before a never-opened era (2007-2016) was scored once, three findings hold in all three eras. (i) The gain is positive, 0.033-0.044 bit per news firm-day for the type×polarity code, and a same-firm no-news placebo gives about zero. Most of it comes from the presence of news alone (0.024-0.038 bit). (ii) Most of the gain is about the size of the move: direction accounts for 8-19%. (iii) Measured on the open-gap outcome, more than half of the gain is direction. Measured on the open-to-close outcome, the direction gain is at most 0.002 bit. Several exploratory patterns did not replicate, and we report them. We then trade the signal, with rules fixed before any trade return was computed. Buying good-news stocks at the open and selling at the close loses about 45-57 bp per day after costs in every era, and is not positive before costs either. A long good-news/short bad-news spread earns 8-28 bp per day before costs (t = 3-7), mainly because bad-news stocks keep falling during the day. Under the main cost settings, the transaction tax and costs remove it. For good-news stocks bought at the close and held for 1-20 days we detect no positive net return in any era, and the intervals are wide. The only positive net result with nominal t > 2 is the long-short spread in 2024-26 at zero slippage (t = 2.09); its short leg is idealised, and the t is not corrected for the sensitivity settings searched, so it is not evidence of an executable profit. All trading results are simulations on daily bars under our labels and cost assumptions

Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://www.econstor.eu/bitstream/10419/344566/1/E ... -Paper-2007-2026.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:zbw:esrepo:344566

Access Statistics for this paper

More papers in EconStor Research Reports from ZBW - Leibniz Information Centre for Economics Contact information at EDIRC.
Bibliographic data for series maintained by ZBW - Leibniz Information Centre for Economics ().