SPY monthly returns vs inflation trend (cooling vs re-accelerating) — last ~3 years
Contrary to the disinflation-trade narrative, the recent three-year sample shows higher average monthly SPY returns in months when year-over-year inflation re-accelerated (≈2.76% per month) than when it cooled (≈1.00% per month). That raw split runs opposite the hypothesis that the bull market has been primarily driven by falling inflation.
This study aligns month-end SPY returns with CPI YoY momentum (ΔYoY) over the most recent ~3-year window, classifying months as Cooling, Re-accelerating, or Steady by small ΔYoY thresholds. The detailed analysis below presents the full sample, the Welch t-test and correlation results, and charts — the averages contradict the thesis, but the statistical evidence is thin and discussed in depth beneath.
For SPY over the past ~3 years, has the bull market really been a disinflation trade — are its average monthly returns meaningfully stronger in months when year-over-year inflation cooled than in months it re-accelerated? Thesis: SPY's monthly returns run clearly higher when inflation is decelerating and flat-to-negative when it ticks back up, so the market has been trading falling inflation more than growth.
How this was measured
Month-end SPY closes from minute bars produce monthly percent returns via month-end (ME) resampling. CPI-based year-over-year inflation is computed from cpi_df as CPI level pct_change over 12 months; the monthly change in that YoY rate (ΔYoY) defines regimes: 'Cooling' when ΔYoY ≤ −0.05 percentage points, 'Re-accelerating' when ΔYoY ≥ +0.05 pp, and 'Steady' within the band. Months are aligned on calendar month-end and restricted to the most recent overlapping ~36 months. We compare mean returns for Cooling vs Re-accelerating months via Welch's t-test and report the Pearson correlation between SPY monthly returns and ΔYoY.
The key numbers
Reading the numbers
Over the 33 months, average monthly SPY return was about 2.76% in inflation re-acceleration months versus about 1.00% in cooling months; the observed gap (~-1.76 percentage points) is not statistically significant (Welch p=0.219).
The charts
The bar chart groups months by inflation-change regime and shows the tallest bar is the Re-accelerating bucket (mean return 2.76%), with Cooling at about 1.00% and Steady near 0.39%. Look at the left bar first — it contradicts the simple thesis that falling inflation coincides with higher returns. Note the sample sizes behind those bars (Re-accel N=15, Cooling N=14, Steady N=4) and the statistical test: the mean gap is not statistically clear (Welch p=0.219), so the visual difference isn’t definitive.
The scatter shows monthly ΔYoY inflation on the x-axis (range −0.4586 to 0.8423, mean 0.0472) against SPY monthly returns on the y-axis (range −0.0599 to 0.1037, mean 0.0172); points are fairly spread out. The Pearson r is about 0.20, which in plain terms is a weak positive relationship, and its p-value (0.264) means that association isn’t statistically reliable. In short, the cloud of points offers at best a slight upward tilt (higher returns when inflation ticks up) but too noisy to support a clear disinflation-driven return story.
Per-bucket monthly return summary (last ~3 years)
| bucket | n_months | mean_return | median_return | std_return |
|---|---|---|---|---|
| Re-accelerating (YoY up) | 15 | 0.0276 | 0.0333 | 0.0374 |
| Steady (±≤0.05pp) | 4 | 0.0039 | 0.0059 | 0.0429 |
| Cooling (YoY down) | 14 | 0.01 | 0.0134 | 0.0381 |
The takeaway
No — this three-year view does not support the idea that the bull market was a ‘disinflation trade.’ Over the last 33 months, average monthly SPY returns were actually larger in re-accelerating inflation months (about 2.76% per month, N=15) than in cooling months (about 1.00% per month, N=14). The mean gap is −1.76 percentage points (cooling minus re-accelerating), but a Welch test gives p≈0.22 — roughly a 22-in-100 chance this difference is just noise, so it isn’t statistically-clear. The month-level correlation between ΔYoY inflation and returns is a weak positive (r≈0.20) with p≈0.26, again not significant and opposite the hypothesized negative relationship. Bottom line: the raw averages contradict the thesis, but the evidence is thin and inconclusive — you can’t reliably claim the market has been trading falling inflation from this sample alone. To be confident you’d need more data or alternate setups (different inflation measures, timing, or thresholds).
The fine print
- Small sample: 33 months total with only 14 cooling and 15 re-accelerating months — thin evidence.
- The observed mean gap isn’t statistically-clear (Welch p≈0.22), so the difference could easily be noise.
- Results hinge on choices: the ±0.05pp deadband for ΔYoY moves months between buckets and shifts means.
- Analysis aligns months to their CPI month (not release date) and uses headline CPI; core/PCE or release timing can relabel months.