AMZN monthly returns vs US retail-sales acceleration, with QQQ beta control (last ~36 months)
The short answer: over the trailing ~36 months Amazon did not reliably rally when U.S. retail sales accelerated. The contemporaneous correlation between AMZN monthly returns and retail-sales acceleration is negative (corr = −0.234). I aligned calendar-month returns with month‑over‑month retail-acceleration and compared outcomes directly, then added a market factor.
What dominates the story is market exposure, not a consumer-spend signal. AMZN carries a large QQQ beta (≈1.125) that explains roughly half of its monthly variation; adding retail-acceleration barely moves the fit and the retail coefficient is small and statistically indistinguishable from zero. The full dataset, charts, t‑tests and regression diagnostics are below.
For AMZN over the past ~3 years, is Amazon actually the 'bet on the US consumer' it's sold as — do months when US retail sales accelerate deliver stronger AMZN returns, and does that link survive once you account for plain Nasdaq beta (QQQ)? Thesis: AMZN's monthly returns show little-to-no positive tie to accelerating retail sales and are dominated by tech-market beta, so the consumer-spending narrative is marketing while the tape prices Amazon as a big-cap growth name.
How this was measured
From AMZN_df and QQQ_df minute bars, built daily closes and compounded to calendar-month returns via log-sum/exp. Retail-sales acceleration was defined on retail_sales_df as Δ(MoM growth): mom=value.pct_change(); acceleration=mom.diff(). Aligned by calendar month (PeriodIndex) and analyzed over the overlapping trailing ~36 months. Tested (i) univariate contrast of AMZN returns when acceleration>0 vs ≤0 (Welch t-test), and (ii) OLS with HC3-robust SEs: AMZN ~ alpha + β·QQQ + γ·retail_accel, comparing R² to a beta-only model.
The key numbers
Reading the numbers
Across 36 months, AMZN monthly returns are slightly negatively correlated with retail-sales acceleration (corr = -0.234). QQQ beta (β = 1.125) explains about 49.8% of returns; adding retail acceleration raises R² by only 0.0037 and the retail coeff is small and insignificant (p=0.618).
The charts
This scatter plots each month's AMZN return against the month-over-month retail-sales acceleration. The cloud of points shows no upward slope; the measured correlation is modestly negative (corr = -0.234), with retail accel ranging from -0.3076 to 0.1948 and AMZN returns spanning -0.1255 to 0.2604. In plain terms, months when retail growth accelerates do not coincide with stronger AMZN returns and the visual spread reinforces that any link is weak or absent.
This scatter shows AMZN returns versus QQQ returns and displays a clear positive relationship: the estimated beta is 1.125 (HC3 p = 0.000) and the model explains about 49.8% of monthly variance (R² = 0.498). Look at the alignment along a rising line rather than a random cloud — AMZN tends to move more than QQQ, and that single factor captures roughly half the variation in AMZN monthly returns. That dominance of market/tech beta undercuts the idea that retail-sales swings are the primary driver of AMZN's monthly performance.
This bar chart compares mean AMZN monthly returns when retail acceleration is positive versus nonpositive: mean return is 0.013657 when accel > 0 and 0.026837 when accel ≤ 0. Contrary to the 'bet on the consumer' story, the mean is actually higher when retail acceleration is absent, and the Welch two-sided test has p = 0.621, indicating the difference is not statistically significant. The takeaway is that bucketed means offer no support for a meaningful positive consumer-driven boost to AMZN returns.
Regression (full model: AMZN ~ const + QQQ + retail_accel, HC3 SEs)
| variable | coef | se | t | p |
|---|---|---|---|---|
| const | -0.00312 | 0.009913 | -0.315 | 0.753 |
| qqq_mret | 1.100146 | 0.260568 | 4.222 | 0 |
| retail_accel | -0.03678 | 0.073764 | -0.499 | 0.6181 |
Regression (beta-only: AMZN ~ const + QQQ, HC3 SEs)
| variable | coef | se | t | p |
|---|---|---|---|---|
| const | -0.003651 | 0.009901 | -0.369 | 0.7123 |
| qqq_mret | 1.125006 | 0.261219 | 4.307 | 0 |
The takeaway
No — months when U.S. retail sales accelerate do not produce stronger AMZN monthly returns; if anything AMZN was slightly weaker in those months. The contemporaneous correlation is negative (−0.234), and mean AMZN return was about +1.37% in accel months versus +2.68% when retail decelerated (Welch two-sided p = 0.621, i.e. ~62-in-100 chance the difference is noise). The tape is dominated by a strong QQQ exposure: beta ≈ 1.125 with R² ≈ 0.498 (the beta is highly significant), and adding retail acceleration only nudges R² to 0.502 (ΔR² ≈ 0.0037) while the retail coefficient is small and non-significant (−0.0368, p≈0.62). Bottom line: monthly returns in this ~36-month window look like a tech/market-beta story, not a robust signal that Amazon reliably rallies when U.S. retail spending accelerates.
The fine print
- Months are aligned contemporaneously; the Advance Retail Sales report is released mid-month for the prior month, so this is not a point-in-time predictive test.
- Only 36 months of data — thin sample for monthly regressions, so small signals may be lost.
- Retail acceleration is Δ(MoM growth) on nominal Advance Retail Sales; revisions, seasonals, or real-sales variants could change results.
- QQQ contains AMZN and is a tech-heavy proxy; a fuller multi-factor attribution could shift how much exposure is labeled "beta."