Proxy Basket Construction: Instrument Selection, Liquidity Constraints, and Beta Matching

A $2.0 million position in a micro-cap industrials name. Market cap under $150M. Average daily dollar volume under $400K. No listed options. No borrow. Selling $200K in a day pushes the print 2-4%.

You cannot hedge it directly. So you hedge something else — a liquid proxy basket that moves like it. Get the basket wrong and everything downstream is mispriced. Get it right and 78% of the variance becomes manageable.

This is how to build that basket.

Start With What You Cannot Use

The micro-cap itself fails every liquidity test. That is the point. The proxy must pass tests the underlying never could.

Two hard filters. No exceptions, no judgment calls:

1. Average daily dollar volume > 20x allocated notional.

If you allocate $680K to one leg, that leg needs ADV above $13.6M. The 20x ratio means a full unwind costs roughly 5% of a day’s volume — small enough to exit in hours without signaling. Anything below 10x and you are building a second illiquid position to hedge the first.

2. Listed options with open interest > 500 contracts at strikes within 15% of spot.

Volume lies. Open interest does not. You need 500+ contracts outstanding near the money so the put spread and the 30-delta call you plan to write later actually have a bid. Check the chain before you allocate a dollar. A name with $50M ADV but 40 contracts of OI is unusable — you can buy the stock but you cannot run the overlay.

Run every candidate through both filters. Most single names fail the second one. That is why the basket is a mix, not a stock pick.

ETF vs. Single Names: The Actual Tradeoff

A pure ETF basket is tempting. One ticker, deep options, trivial to rebalance. For a micro-cap industrials position, that might be a small-cap industrials ETF with $85M ADV and 2,000+ contracts of OI at every strike. It clears both filters by a mile.

The problem: it tracks the sector, not your stock. In the worked example, the Russell 2000 loading is 1.35 (t-stat 6.1) and the sector loading is 0.48 (t-stat 3.4). A broad ETF captures the 1.35 but washes out the 0.48 — the sector tilt, the size tilt, the value loading of 0.22 that make this specific name move the way it does. Pure ETF replication typically lands at R² around 0.60-0.65 for single micro-caps. You leave 10+ points of explainable variance on the table.

A pure single-name basket has the opposite problem. Pick three liquid industrials peers, match the factor profile closely, push R² to 0.75+. But single names gap on earnings. You swap one idiosyncratic exposure for three smaller ones. And options liquidity is thinner — 600 contracts of OI instead of 2,000, wider spreads, worse fills on the roll.

Neither extreme is optimal. The fix is a split.

Why 60/40 Tracks Better

Allocate 60% to the ETF, 40% to two liquid single names. The logic is mechanical:

• The ETF leg anchors the market beta. It is the stable, high-R² core that does not gap on earnings.
• The single-name legs add the sector and style tilt the ETF misses. They close the tracking gap without dominating the basket.

In practice this 60/40 structure tests at R² ~0.75-0.78 against the micro-cap, versus ~0.62 for ETF-only. Residual idiosyncratic variance drops from ~38% to ~22%. That 16-point improvement is the difference between a hedge that roughly works and one that offsets drawdowns.

The 40% is split across two names so one earnings print cannot break tracking. Never let a single name exceed 25% of the basket.

Here is the example basket for the $2.0M position:

Every leg clears both filters. The tightest ratio is peer A at 75x ADV-to-notional — nearly 4x the 20x minimum. The thinnest options chain is peer B at 650 contracts — above the 500 floor with room to trade spreads.

Dollar-Beta Sizing: The $2M x 1.35 Math

Basket notional is not $2.0M. The micro-cap moves 1.35x the market, so $2.0M of micro-cap behaves like $2.7M of market exposure on the systematic component. Size the basket to match dollar beta, not dollars.

The formula:

Basket notional = (Micro-cap notional × Micro-cap beta) / Basket blended beta

Plug in the example:

($2,000,000 × 1.35) / 1.59 = $1,698,000 ≈ $1.70M

The basket’s blended beta of ~1.59 comes from its components: the ETF at ~1.15, peer A at ~1.9, peer B at ~2.1, weighted 60/25/15. Verify each beta from the same 252-day regression window used for the factor decomposition — do not mix lookbacks.

Two errors to avoid:

• Sizing to notional ($2.0M). This underhedges by ~15%. A 10% market drop hits the micro-cap for ~$270K (1.35 beta) but the $2.0M basket for less, leaving a gap the put spread was not sized for.
• Sizing to full $2.7M market-equivalent. This overhedges because the basket itself is high-beta (1.59). $2.7M of basket would behave like $4.3M of market — nearly 60% too much short exposure once the overlay goes on.

The $1.70M figure is the one where a 1% market move produces the same dollar P&L on both sides of the systematic component.

Rebalance Rule: Monthly, Plus a Tripwire

Factor loadings drift. The 1.35 beta today is 1.20 or 1.50 in six months. Two rules:

Scheduled: re-estimate monthly. Re-run the 252-day regression, recalculate blended basket beta, resize legs. Frequent enough to catch drift, infrequent enough to avoid churning on noise.

Tripwire: rebalance immediately if realized beta drifts more than 0.15 from target. Track a 30-day realized beta of micro-cap vs. basket. If it prints 1.74 against a 1.59 target (or 1.44), do not wait for month-end. Resize or swap a leg.

Also re-screen both liquidity filters on every rebalance. ADV decays. Chains thin out after corporate actions. A leg at 800 contracts of OI in January can print 300 in March. Below 500, replace it — find another peer or push the weight back to the ETF.

What Good Looks Like

A finished basket has four properties:

1. Every leg passes both filters with margin — 50x+ on ADV, 600+ on OI minimum in practice.
2. Blended beta is known to two decimals and sizing math ties to it explicitly.
3. No single name exceeds 25% of notional.
4. Next rebalance date is on the calendar, and the 0.15 tripwire is monitored, not assumed.

Build that and the downstream work — put spreads, short calls — has a surface to stand on. Skip it and the overlay is priced against a basket that does not track: the hedge pays when it should not and misses when it should pay.

The micro-cap stays untouched. The basket does the work.