AI Research METAMETA_earnings

META earnings: EPS surprise magnitude vs earnings-day move (last ~3 years)

Surprising result up front: META’s quarterly EPS surprise magnitude barely moves the needle on the stock’s earnings-day reaction. I matched each of the last 11 quarterly EPS surprises to the first regular-session close-to-close return and ran Pearson/Spearman correlations and simple OLS fits; the signed Pearson r is essentially zero (r ≈ -0.05), indicating no reliable linear link between beat/miss size and the direction or size of the next-session return.

Big one-day moves do happen — the mean absolute earnings-day return is about 8.44% — but they don’t track headline EPS beats in any consistent way. The size-on-size relationship even leans negative but is statistically inconclusive; the detailed charts, regressions and event-level data below show the full evidence and how guidance, commentary and positioning appear to matter more than the raw EPS surprise.

The research question

For META over the past ~3 years, does the size of the quarterly EPS surprise actually drive the size of the earnings-day move — is a bigger beat rewarded with a proportionally bigger pop — or is the reaction basically disconnected from the headline number? Thesis: the correlation between surprise magnitude and the absolute earnings-day return is weak, because the market trades forward guidance and positioning rather than the reported EPS beat, so how far the stock jumps barely tracks how big the surprise was.

How this was measured

For each META quarterly earnings with a known release date in the last ~3 years, the earnings-day reaction is defined as the close-to-close return of the first trading day on or after the reported_date (this captures both after-hours and pre-market releases as the next regular session's move). We pair that return with the reported EPS surprise_percentage (data-source units are percentage points) and compute Pearson/Spearman correlations for the signed relation (surprise vs signed return) and for the size relation (|surprise| vs |return|). A simple OLS fit reports slope and R²; slope is interpreted as return per 1 percentage-point of surprise.

The key numbers

Earnings events analyzed
11
2023-07-26 to 2026-01-28
Mean earnings-day return (signed)
1.7442%
Median earnings-day return (signed)
5.3346%
Mean |earnings-day| return
8.4441%
Pearson r (surprise vs signed return)
-0.051
Linear correlation (signed)
Pearson p-value (signed)
0.8820
Spearman rho (signed)
-0.327
Rank correlation (signed)
Spearman p-value (signed)
0.3259
OLS slope (per 1pp surprise → return)
-0.0686%
Return change per +1 percentage-point surprise
R² (signed OLS)
0.003
Slope per +10pp surprise (return)
-0.6865%
Pearson r (|surprise| vs |return|)
-0.403
Size-on-size correlation
Pearson p-value (abs)
0.2196
R² (abs OLS)
0.162

Reading the numbers

Across 11 earnings events, the average absolute earnings-day move was 0.0844407 while the correlation between surprise and signed return is essentially zero (Pearson r = -0.0508, p = 0.882), so surprise size alone shows no clear linear link to the direction of the move.

The charts

META: EPS surprise (%) vs earnings-day return
What this chart says

This scatter plots each quarter's EPS surprise (2.4055–23.1801) against the actual close-to-close return (ranging -0.1691 to 0.1564). The points are all over the place rather than forming an upward line you’d expect if bigger beats produced bigger pops; Pearson r = -0.0508 with p = 0.882 indicates no detectable linear relationship. The OLS slope is effectively zero (slope per +10pp surprise = -0.0068648, R² = 0.002585), so a larger reported surprise did not meaningfully predict a bigger signed move in this sample.

META: |EPS surprise| vs |earnings-day return|
What this chart says

This scatter uses absolute values: |surprise| (same 2.4055–23.1801 range) versus |return| (0.0207–0.1691, mean 0.0844407). Rather than a positive connection, the Pearson r = -0.4026 points to a modest negative relationship — bigger surprises were not followed by larger absolute moves and in this small sample even trended slightly the other way. Visually the cloud of points lacks a clear monotonic rise, so surprise magnitude does not reliably predict how far the stock jumps on earnings.

META earnings events — surprise vs earnings-day move

reported_trading_daysurprise_pctearnings_day_returnabs_surprise_pctabs_earnings_day_return
2023-07-262.410.06572.410.0657
2023-10-2520.94-0.067320.940.0673
2024-02-017.460.15647.460.1564
2024-04-249.03-0.16919.030.1691
2024-07-317.950.10917.950.1091
2024-10-3013.99-0.055113.990.0551
2025-01-2920.060.020720.060.0207
2025-04-3023.180.053323.180.0533
2025-07-3021.840.097721.840.0977
2025-10-298.05-0.07698.050.0769
2026-01-288.560.05748.560.0574

The takeaway

Short answer: no — across the last 11 META quarters there’s essentially no reliable link from the EPS surprise size to the earnings-day move. The signed Pearson correlation is basically zero (r = -0.0508, p = 0.882) and the OLS for signed returns explains almost nothing (R² = 0.0026). The size-on-size relation leans negative (|surprise| vs |return| Pearson r = -0.4026) but that pattern is not statistically convincing (p = 0.220 and abs OLS R² = 0.162), and a +10pp surprise maps to only about a -0.69% change in the session move by the reported slope. Put bluntly: big earnings-day moves (mean absolute move ~8.44%) happen often, but they don’t track the headline EPS beat or miss in a consistent, predictable way — the signed relationship is a clear null and the size relationship is at best a weak, inconclusive lean. Practical takeaway: don’t expect a proportionally larger pop from a larger EPS beat; guidance, commentary and positioning are likely the drivers you should watch instead.

The fine print