David Bernstein
September 2026
While watching CNBC’s coverage of a recent Oracle earnings announcement, I noticed an intriguing pattern: losses before the announcement appeared to be largely offset by gains after the results were released. Because Oracle reported after the market closed, the two movements occurred in distinct trading windows. That raised a broader question: Do stocks typically continue their pre-announcement movement after earnings—or reverse it?
This study examines four quarterly announcements for each of seven large technology companies—Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia and Tesla—and seven established dividend-paying companies—Visa, Costco, Amgen, Chubb, Aflac, McKesson and Clorox. The sample is limited to earnings announcements released after the regular market close.
Returns are measured relative to RSP, the Invesco S&P 500 Equal Weight ETF. Unlike the conventional capitalization-weighted S&P 500, RSP gives approximately equal weight to every company in the index and is therefore less dominated by the largest technology stocks. Subtracting RSP’s return from each company’s return helps separate company-specific movement from changes in the overall market.
The pre-announcement window runs from the previous trading-day close to the announcement-day close. Because the earnings release occurs after the announcement-day close, this window ends before investors receive the report. The post-announcement window runs from the announcement-day close to the following trading-day close and therefore captures the first trading session following the release.
For both windows, the stock’s return is reduced by RSP’s return over exactly the same period. All results described below are therefore market-adjusted rather than raw stock returns.
This is what I found.
Key Findings:
Post-announcement movements were substantially larger than pre-announcement movements in both groups.
Technology stocks experienced larger post-announcement movements than dividend-paying stocks.
Dividend-paying stocks had higher average post-announcement returns in this sample, but the difference was not statistically decisive.
Reversals were frequent in both groups, but pre-announcement returns did not reliably predict subsequent movements.
Analysis:
Pre-announcement movements were nearly identical across the groups. The average absolute market-adjusted return was 1.35 percent for the technology stocks and 1.29 percent for the dividend payers—a difference of only 0.06 percentage points. Absolute returns measure the size of a movement without regard to whether it was positive or negative.
Post-announcement movements were substantially larger, particularly for technology stocks. The average absolute market-adjusted return increased to 6.38 percent for the technology group and 4.03 percent for the dividend payers. The 2.35-percentage-point difference had a Welch-test p value of 0.035.
Dividend-paying stocks had higher average post-announcement returns. Their average market-adjusted return was positive 1.28 percent, compared with negative 0.83 percent for the technology group—a difference of 2.11 percentage points. However, the difference was not statistically decisive: the Welch test produced a p value of 0.243.
The median results also favored dividend-paying stocks. Their median post-announcement market-adjusted return was positive 1.52 percent, compared with negative 2.07 percent for the technology stocks—a descriptive difference of 3.59 percentage points.
Technology stocks displayed greater post-announcement dispersion. Their standard deviation was 7.89 percent, compared with 5.23 percent for the dividend payers. The 2.66-percentage-point difference produced a Brown-Forsythe p value of 0.052—borderline evidence rather than statistical significance at the conventional 5 percent level.
Reversals were frequent in both groups. A reversal occurs when the pre- and post-announcement market-adjusted returns have opposite signs. Sixteen of the 28 technology events and 18 of the 28 dividend-paying events met that definition. Fisher’s exact test produced a p value of 0.785, providing no evidence that reversal frequency differed between the groups.
The four possible movement patterns produced no simple rule. The technology group recorded five gains followed by gains, 10 gains followed by losses, six losses followed by gains and seven losses followed by losses. The dividend payers recorded nine gains followed by gains, 11 gains followed by losses, seven losses followed by gains and one loss followed by a loss.
The correlations did not establish a reliable relationship between pre- and post-announcement returns. The technology group’s Pearson and Spearman correlations were positive 0.164 and positive 0.083, respectively; neither was statistically significant. For the dividend payers, the Pearson correlation was negative 0.368, with a p value of 0.054, but the Spearman correlation was only negative 0.147, with a p value of 0.456. The conflicting dividend results suggest that a few relatively large observations influenced the Pearson estimate. A larger study incorporating more companies, more quarters and information about earnings, revenue and guidance surprises is needed.
These results should be considered exploratory. Each group contains 28 events but only seven companies, with four repeated observations for each company. The conventional tests do not fully account for that dependence, so the observations should not be treated as if they came from 28 independent companies in each group.
Source note: Nasdaq historical-price and earnings-surprise data retrieved September 2026. Calculations by the author. The workbook containing all 56 observations, formulas, statistical results and variable definitions is available to paid subscribers. The blog is inexpensive with the 50 percent off coupon available through the subscribe now button.
The spreadsheet is below paywall.
