KO vs SPY over ~3 years: downside vs upside beta and extreme-day cushioning
When the market plunged, Coca‑Cola barely budged: on SPY’s worst decile (mean SPY return −1.63%) KO averaged essentially 0.00% (-0.0004%). That gap is the clearest single signal in this analysis — over ~3 years and 749 trading days KO behaved like a buffer on extreme down days rather than a mirror of broad-market selloffs.
I tested that formally by estimating conditional betas on SPY up‑ and down‑days and by comparing decile tails. Downside beta (0.1393) is meaningfully lower than upside beta (0.2105), and the upside slope is highly significant (p=0.0001) while the downside slope is only marginal (p=0.0901). Full regressions, capture ratios, and day‑by‑day charts are shown below for the detailed evidence and robustness checks.
For KO over the past ~3 years, does it actually live up to its defensive reputation — on SPY's worst days does it fall meaningfully less than the market while still keeping pace on the best days? Thesis: KO's downside beta to SPY runs well below 1 and below its upside beta, confirming it cushions drawdowns far more than it gives up on rallies.
How this was measured
Resampled KO and SPY minute bars to daily closes, computed close-to-close returns, and limited to the last ~3 years of overlapping data. Estimated conditional betas via linear regression of KO returns on SPY returns separately for SPY down-days (downside beta) and SPY up-days (upside beta). Identified SPY's bottom and top deciles by daily return to summarize KO's average performance on the market's worst and best days and computed capture ratios (KO mean ÷ SPY mean) for those tails. All computations are close-to-close; no intraday information is used.
The key numbers
Reading the numbers
Across 749 trading days, KO's downside beta to SPY is 0.1393 while its upside beta is 0.2105, so KO moves much less than SPY on down-days and still lags on up-days, with an up-minus-down gap of 0.0713.
The charts
This scatter plots 326 SPY down-days: SPY returns run from -0.0575 to 0.0 with mean -0.0065, while KO's returns on those same days cluster around zero (mean 0.0004) and range -0.0482 to 0.0328. Look at how many KO points sit at or above zero even when SPY is negative — that visual cluster is the cushion people talk about. That pattern is consistent with the reported downside beta of 0.1393: KO tends to move far less than SPY on bad days.
This scatter covers 422 SPY up-days: SPY runs from 0.0 to 0.1125 with mean 0.0066, while KO's returns are tighter (mean 0.0007, range -0.0268 to 0.0402). Notice KO often registers smaller gains and even some losses on days SPY is positive — the cloud is much flatter than the x-axis spread. That matches the upside beta of 0.2105: KO participates in rallies but to a much smaller degree than SPY.
The two bars compare means in SPY worst and best deciles: on the worst-decile days SPY mean is -0.0163 while KO's mean is essentially -0.0, producing a tiny downside capture of 0.000231; on the best-decile days SPY mean is 0.0166 and KO mean is 0.0027, giving an upside capture of about 0.1594. The takeaway is stark: KO is nearly flat on the market's worst days and captures only a small fraction of the biggest rallies, which is exactly why downside beta is much lower than upside beta.
Conditional beta summary (KO_ret ~ SPY_ret)
| condition | N | beta_slope | r_squared | p_value |
|---|---|---|---|---|
| Downside (SPY<0) | 326 | 0.1393 | 0.0088 | 0.0901 |
| Upside (SPY>0) | 422 | 0.2105 | 0.0344 | 0.0001 |
| All days | 749 | 0.1096 | 0.0132 | 0.0017 |
Extreme-day (decile) summary
| bucket | N | SPY_mean_ret | KO_mean_ret | KO/SPY_mean_capture |
|---|---|---|---|---|
| SPY worst decile | 75 | -0.0163 | 0 | 0 |
| SPY best decile | 75 | 0.0166 | 0.0026 | 0.159 |
The takeaway
Yes — over the past ~3 years KO acted like a defensive stock: it fell far less than SPY on the market's worst days while giving up most upside on the big rally days. The conditional betas show this directly: downside beta = 0.1393 vs upside beta = 0.2105 (gap 0.0713), so KO's sensitivity to SPY declines is materially lower than its sensitivity to rallies. Looking at extremes, on SPY's worst decile (mean SPY return -1.63%) KO averaged essentially 0.00% (KO mean = -0.0003766) — a mean capture of 0.00023× (N=75 days), i.e., KO barely moved when SPY plunged. On SPY's best decile (mean SPY return 1.66%) KO averaged 0.265% (capture ~0.159×), so it participated only modestly in big up days. Statistically, the upside slope is highly significant (p = 0.0001, N=422) while the downside slope is only marginal (p = 0.0901, N=326), so the pattern is strong for upside behavior and suggestive — not ironclad — for the downside slope itself. Practically: KO cushions the market much more often than it keeps pace — it outperformed SPY on 72.4% of down-days but only met a 75%-of-SPY threshold on 29.6% of up-days — so the defensive reputation is supported in this window.
The fine print
- Window limited to the last ~3 years of overlapping KO and SPY data; results may differ in other regimes or longer horizons.
- Downside beta estimate is marginal (p = 0.0901) and overall R²s are small, so the regression slopes explain little of daily return variation.
- Extreme-day summaries use in-sample deciles (N=75 per decile); different tail definitions (e.g., 5% tails) will change the averages.
- These are close-to-close returns (overnight gaps included); intraday cushioning behavior is not captured.