Basis Risk in Proxy Hedging: Measuring the Gap Between Your Stock and Its Synthetic
Earnings week. Your proxy basket — 60% small-cap industrials ETF, 40% two liquid single-names — closes the week up 2%. Your micro-cap closes down 8% on a revenue miss and a pulled guidance number.
Nothing broke. The options priced correctly. The basket tracked its own index fine. The problem is the 10-point gap between what you own and what you hedged with. That gap is basis risk, and in proxy hedging for micro-caps it is the entire game.
If you run a proxy hedge without measuring that gap weekly, you do not have a hedge. You have a second position.
What Basis Risk Actually Is
Basis risk in proxy hedging is the return difference between the micro-cap and the synthetic basket built to stand in for it.
Define it as a time series, not a concept:
basis_t = R_microcap,t − R_proxy,t
Cumulate that series and you get the drift between the two legs. Take its standard deviation and annualize it and you get tracking error — the single number that tells you whether the hedge is doing its job.
A proxy basket with a factor R² of 0.78 still leaves 22% of variance idiosyncratic. That residual lives entirely in the basis. Earnings, restatements, customer loss, a CEO walking out — all of it hits the micro-cap leg and nowhere else. No reweighting of the liquid side fixes that. The only question is how large the basis gets and whether you have a rule for when it invalidates the hedge.
How to Measure It: Three Numbers, No Shortcuts
You need three metrics, updated on the same schedule. Monthly minimum, weekly if the position is large enough to matter.
1. Tracking error (annualized). Standard deviation of daily basis_t over a trailing 60-day window, multiplied by √252. This is your headline gauge. A proxy hedge on a micro-cap running 8–12% annualized tracking error is normal. Above 15%, you are paying for protection that may not arrive.
2. Rolling 60-day correlation. Pearson correlation of daily returns, micro-cap vs. proxy basket. Healthy range for a working proxy hedge is 0.70–0.85. Below 0.60 for two consecutive months, the factor loadings you built the basket on have decayed.
3. Beta drift. Re-run the original factor regression monthly and compare realized basket beta to target. If realized beta has moved more than 0.15 from the sizing beta, your notional is wrong even if correlation still looks acceptable. Resize or rebalance.
Track all three together. Tracking error tells you the cost of the gap. Correlation tells you whether the relationship still exists. Beta drift tells you whether the sizing is still correct. Any one of them alone will mislead you.
12-Month Walkthrough: What the Chart Actually Looks Like
Take the $2M micro-cap position from the proxy hedging setup: β 1.35 to the Russell 2000, R² 0.78, $1.7M proxy basket. Here is a full year of trailing 60-day tracking error and correlation, month by month, with what drove each move:
| Month | Ann. Tracking Error | 60-Day Correlation | Realized Beta | What Happened |
|---|---|---|---|---|
| Jan | 9.1% | 0.81 | 1.33 | Clean start. Basket tracks. Basis noise only. |
| Feb | 8.7% | 0.83 | 1.36 | Best month. Broad small-cap rally lifts both legs together. |
| Mar | 10.4% | 0.77 | 1.31 | Sector rotation. Proxy ETF drifts from micro-cap by 3% intra-month, then reconverges. |
| Apr | 11.2% | 0.74 | 1.29 | Micro-cap misses revenue by 4%. Proxy flat. First idiosyncratic print. |
| May | 9.8% | 0.78 | 1.34 | Mean reversion. Basis from April partially reverses. |
| Jun | 12.6% | 0.69 | 1.22 | Low-volume summer tape. Micro-cap trades 40% below average volume; bid-ask widens to 5%. Proxy stays liquid. |
| Jul | 14.1% | 0.64 | 1.18 | Contract delay announced. Micro-cap −11% in two sessions. Proxy +1%. Largest single basis print of the year. |
| Aug | 13.5% | 0.66 | 1.24 | Aftershock. Correlation stabilizes but tracking error stays elevated on the July print still in window. |
| Sep | 10.9% | 0.73 | 1.30 | Rebalance month. Basket weights shifted toward the stronger single-name leg. Beta back near target. |
| Oct | 9.4% | 0.79 | 1.35 | Normal regime returns. Tracking error bleeds down as July rolls out of the window. |
| Nov | 8.9% | 0.82 | 1.37 | Strongest co-movement since February. Hedge efficiency highest here. |
| Dec | 16.3% | 0.58 | 1.12 | Tax-loss selling hits the micro-cap (−9% in December on no news). Proxy basket up 3% into year-end rally. Textbook divergence. |
Read the year as three regimes, not twelve independent months. January–May is the working state: tracking error under 11%, correlation above 0.74. June–August is stress: one idiosyncratic event plus thin liquidity pushes tracking error past 12% and correlation under 0.70. October–November recovers after a rebalance. December is the regime break.
The 12-month average tracking error is 11.2%. That number is useless. Nobody experiences the average. You experience July and December, when the hedge is weakest at exactly the moment you check it most.
Why the Gap Widens: Three Regimes
Basis risk in micro-cap proxy hedging does not widen randomly. It widens in three repeatable regimes:
1. Idiosyncratic events. Earnings, guidance cuts, contract wins or losses, management changes. The micro-cap moves 8–15% while the proxy moves with the sector. July in the table is the template. Expect 2–4 of these per year in a single micro-cap name. Nothing in the proxy construction prevents them — the 22% residual from the factor regression is precisely this bucket.
2. Liquidity droughts. Summer months, holiday weeks, any stretch where micro-cap ADV drops below $250K. The proxy keeps trading tight; the micro-cap gaps on small prints and the bid-ask alone contributes 1–2% of spurious basis per round trip. June shows this: no news, but tracking error still climbs to 12.6% on microstructure alone.
3. Flow-driven divergence. Tax-loss selling in December, index reconstitutions that include the proxy ETF but not the micro-cap, sector ETF inflows that lift liquid names and skip the illiquid one. December is the template: fundamentals unchanged, shareholder base changed. These are the most dangerous episodes because correlation breaks without a headline, and managers keep trusting a hedge that has quietly stopped working.
The Kill Threshold: When the Hedge Stops Being Useful
A proxy hedge has a defined operating envelope. Outside it, keeping the hedge on costs premium and delivers no offset. Use hard rules:
- Tracking error above 15% annualized for 40 consecutive trading days. At 15%+ tracking error, the one-month 1σ basis band is ±4.3%. A typical put-spread hedge caps basket losses at ~14% but the micro-cap can sit 4–8% outside the basket with normal probability.
- Rolling 60-day correlation below 0.60 for two consecutive month-ends. One bad month can be an event. Two is a broken relationship. Re-estimate the factor model from scratch. If the new R² comes back under 0.70, the name is no longer hedgeable by proxy. Stand the hedge down.
- Realized beta drift above 0.25 from target with no reconvergence after one rebalance. The micro-cap’s systematic loading itself has changed. Resizing will not fix it because the target moved. Pause and rebuild the decomposition.
- December rule. If tax-loss pressure or year-end illiquidity pushes the micro-cap more than 6% below the proxy in a single month with no fundamental news, cut hedge notional in half until January volume returns. Paying full put-spread cost against a leg that is trading on flows, not factors, is negative expected value.
When any single threshold trips, halve the hedge. When two trip together, remove it and reassess. A proxy hedge that fails these checks is not conservative to keep — it is an uncompensated short-volatility position on a relationship that no longer exists.
What to Do Monday Morning
Compute trailing 60-day tracking error, rolling correlation, and realized beta for your position. Log all three weekly. If all three sit inside their bands, the hedge is working. If tracking error climbs while correlation holds, you are in an event window — sit tight but do not add notional. If correlation breaks below 0.60 and stays there, the relationship that justified the structure is gone.
Basis risk is not a footnote to the hedge. It is the hedge.