AI Research FANGUSOmacro:brent

FANG daily beta to crude oil (USO close) by USO volume regime (top quartile high volume)

0.6485
Beta (low USO vol)

FANG's sensitivity to crude oil is often treated as a constant. It isn't. Over 753 trading days, the stock's daily beta to an oil proxy shifted from roughly 0.65 on ordinary days to about 0.52 when USO volume landed in the top quartile — a decline of roughly 0.13. The interaction term is statistically meaningful (p ≈ 0.02), suggesting this is unlikely to be random noise.

The theory framing the test: when crude-ETF trading spikes, speculative activity dominates, institutional sellers may step in and break the usual stock-oil linkage. Using USO returns as the crude proxy and separating days by volume regime, the evidence supports that hypothesis — with the caveat that Brent data was unavailable.

The full analysis below lays out the regression design, interaction model, R-squared, and the threshold that defines a high-volume day. The takeaway for risk modeling: a single fixed beta will misprice exposure on exactly the days when oil trading heats up.

The research question

For FANG over the past ~3 years, does its daily return's beta to Brent crude decrease on days when USO's daily trading volume is in its top quartile? I expect beta to shrink because elevated crude-ETF activity signals speculative noise, prompting institutional selling in leveraged oil names and breaking the link between stock and commodity.

How this was measured

Daily FANG and crude-oil returns were computed from resampled minute bars and, where available, the brent_daily_df macro frame. USO daily trading volume was aggregated to a calendar-day sum. Days were classified into 'high' volume (≥ 75th percentile of USO volume) and 'low' volume (< 75th percentile). Simple OLS regressed FANG returns on crude returns separately for each regime, and a full interaction model (FANG_ret ~ crude_ret + high_vol + crude_ret*high_vol) was estimated via statsmodels.OLS. The interaction coefficient and its p-value test whether the beta differs between regimes. Brent crude data (brent_daily_df) was unavailable; USO closing price returns serve as a proxy for oil returns, which may slightly alter the measured exposure.

The key numbers

Beta (low USO vol)
0.6485
OLS slope of FANG on crude proxy, N=564 days
Beta (high USO vol)
0.5162
OLS slope of FANG on crude proxy, N=189 days
Beta difference (high - low)
-0.1324
Positive = beta higher in high-volume regime
Interaction coefficient (crude_ret * high_vol)
-0.1324
from regression: FANG_ret ~ crude_ret + high_vol + crude_ret*high_vol
Interaction p-value
0.0206
p=0.0206 < 0.05 → significant interaction
R-squared (interaction model)
0.3724
USO vol top quartile threshold
6,543,379
Days analyzed
753
Of which 189 high-USO-vol days, 564 other days

Reading the numbers

On low-USO-volume days FANG's beta to crude was 0.65; on high-USO-volume days it fell to 0.52. That 0.13 drop is statistically meaningful (p=0.02), so the data support the prediction that elevated crude-ETF activity loosens the stock-oil link.

The charts

FANG beta to crude oil (USO close) by USO volume regime
What this chart says

The tall left bar shows FANG's beta to crude on low-USO-volume days at 0.65, while the shorter right bar shows beta dropping to 0.52 on high-volume days. That gap of about 0.13 is exactly the effect you predicted: when crude-ETF trading spikes, FANG's daily return tracks oil less tightly. The size of the drop is meaningful enough statistically to support the idea that speculative ETF activity weakens the stock-commodity link.

Distribution of USO daily volume (vertical threshold = top quartile)
What this chart says

USO daily volume is heavily bunched at lower levels, with a long right tail reaching up to roughly 144 million shares, and the mean of about 7.5 million is pulled up by those extreme days. The vertical line marks the top-quartile cutoff at about 6.54 million shares, so the 189 high-volume days are the tail observations to the right. This chart shows the high-volume regime is not a typical state of the market but a set of unusual spikes, which fits the story that speculative noise is what breaks FANG's link to crude.

Regression output summary

Regime / coefficientEstimatep-valueN
Low vol beta0.6485p=0.0000N=564
High vol beta0.5162p=0.0000N=189
Interaction coeff-0.1324p=0.0206

The takeaway

Yes — the data supports your hypothesis. On high-USO-volume days (top quartile), FANG's daily beta to the oil proxy drops to about 0.52, versus about 0.65 on lower-volume days, a difference of roughly -0.13. That gap is statistically meaningful: there's only about a 2-in-100 chance you'd see this by luck if the true betas were equal, based on 753 trading days (189 of them high-volume). So this isn't a coin flip, but it's also not a massive effect — high oil-ETF activity is associated with a modest but real weakening of FANG's oil sensitivity. The practical takeaway is that FANG's oil beta isn't a single stable number; it shrinks when crude-ETF trading spikes, so risk models and hedging decisions built on a fixed beta will be off during exactly those high-activity periods. One important caveat: the analysis used USO returns as the crude proxy, not Brent, so the exact betas apply to WTI-style exposure rather than Brent directly.

The fine print