AI Research KMIKMI_fundamentalsKMI_earnings

KMI dividend-per-share increases vs forward 60-day returns (past ~3y, N=4 events)

4
DPS-increase events

A 6.6 percentage-point edge sounds like a signal. Over the past three years, KMI’s quarterly dividend hikes preceded an average forward 60-day return of +13.9%, versus a +7.3% baseline. But that gap is built on just four events—and one of them was flat. The statistics say what the sample size whispers: a p-value of 0.27 is nowhere near conclusive.

The question worth asking isn't whether KMI's payout increases move the stock. It's whether four data points can support any claim at all. The full analysis below walks through the event-by-event returns, the baseline comparison, and exactly why this edge fails the reliability test.

The research question

For KMI over the past ~3 years, does an increase in the quarterly dividend per share predict above-baseline forward 60-day returns? Dividend hikes signal management confidence in cash flow, attracting income investors and driving sustained outperformance.

How this was measured

Quarterly KMI fundamentals filtered to fiscal quarters with fiscal_date_ending ≤ today − 45 days. Dividend-per-share (DPS) computed as dividend_payout / shares_outstanding. A DPS increase is flagged when DPS exceeds the previous quarter's DPS. Event date = the reported_date from KMI_earnings when available, else fiscal_date_ending + 45 days (conservative reporting lag). Forward 60-trading-day return measured from the first trading day on or after the event date. The unconditional baseline is computed over every possible 60-day window in the same price history.

The key numbers

DPS-increase events
4
qualified quarters with both prior-quarter DPS and a hike
Event mean forward 60d return
13.8954%
N=4
Baseline mean forward 60d return
7.2997%
N=694 windows
Edge (event − baseline)
6.5957%
Event fraction positive
75.00%
Baseline fraction positive
75.65%
Welch t-statistic
1.358
positive = event returns exceed baseline
Welch p-value (two-sided)
0.2667
p=0.2667 ≥ 0.05 → no statistically-clear signal

Reading the numbers

In 4 dividend-hike quarters, KMI's average forward 60-day return was 13.9% vs a 7.3% baseline — a 6.6-point edge. With p=0.27, that gap isn't statistically clear, and the positive-window rate was basically the same (75% vs 75.6%).

The charts

KMI forward 60d returns after DPS increases
What this chart says

The histogram of the four post-hike windows shows mostly positive returns, but the sample is tiny: the worst outcome was essentially flat at -0.3%, the best reached +20.3%, and the average was 13.9%. Three of the four windows (75%) ended positive, which is almost identical to the baseline positive rate of 75.6%. The direction is encouraging, but one flat quarter and the small count make this a weak basis for claiming hikes reliably predict outperformance.

Mean forward 60d return: DPS-hike events vs baseline
What this chart says

The bar chart makes the headline contrast easy to see: dividend-hike events averaged a 13.9% forward 60-day return, nearly double the 7.3% average across all 60-day windows. That 6.6 percentage-point edge matches the dividend-signaling story, but with only four events and a p-value of 0.27, the gap is not statistically distinguishable from random variation. Treat it as a suggestive pattern, not a proven edge.

DPS-hike event‑level forward returns (chronological)

fiscal_date_endingevent_datedpsfwd_60d_return
2024-03-312024-04-170.28420.1582
2024-06-302024-07-170.28890.1975
2024-09-302024-10-160.28950.2034
2025-06-302025-07-160.2943-0.0032

The takeaway

The short answer is no: over the past three years, KMI's quarterly dividend hikes do not give a statistically reliable edge for forward 60-day returns — the evidence is suggestive but inconclusive. Across only four hikes, the average forward 60-day return was +13.9% versus a +7.3% baseline, an edge of about 6.6 percentage points. But three of those four events did the heavy lifting (+15.8%, +19.8%, +20.3%), while the latest hike was flat at -0.3%. The share of positive windows was essentially a wash — 75.0% after hikes versus 75.6% for the baseline — and the p-value of 0.27 means there's about a 1-in-4 chance this gap is just random noise. With only four events, that's too flimsy to call a real pattern or to rely on as a signal.

The fine print