Scaling the Overlay: Applying Proxy Carry Across a Portfolio of Illiquid Positions

Three illiquid positions. $2.0M in an industrial micro-cap, $1.5M in a regional bank stub, $3.0M in a specialty chemicals name. Total book: $6.5M. No listed options on any of them. No borrow.

Run separately, that is three factor regressions, three proxy baskets, three options overlays — roughly $5.5M in combined proxy notional and three sets of options tickets to manage. Run as a portfolio, the factor exposures net against each other and collapse to a single $4.2M proxy basket with one consolidated overlay. Same hedgeable risk covered. Less notional, fewer tickers, lower cost.

That netting effect is why the overlay gets better as the book gets bigger.

Single-Name vs Portfolio Overlay

A single-name overlay is a one-to-one mapping. One micro-cap, one regression, one proxy basket, one put spread plus one short call stream. It works, but it carries the full overhead per dollar hedged: separate basket, separate hedge cost, full idiosyncratic residual.

A portfolio overlay aggregates the book first, then hedges the aggregate. The regression runs on the position-weighted portfolio return series against the same factor set: broad market, sector, size, value, momentum. The loadings describe net systematic exposure, not any single name.

The practical difference: offsetting exposures cancel before they ever reach the proxy. A long-value industrial and a growth-tilted chemicals name partially net on HML. Two names with opposite rate sensitivity partially net on the sector leg. You only buy proxy for what survives netting.

Netting Factor Exposures

Factor netting is the core mechanic of proxy hedging a portfolio. Each position contributes dollar-beta to each retained factor. Sum them, and longs in one name offset shorts-equivalent tilts in another.

Take the three-position book:

• Position A ($2.0M industrials): market β 1.35, sector +0.48, value +0.22.
• Position B ($1.5M regional bank): market β 1.10, sector (financials) +0.61, value −0.15 — a growth tilt that subtracts from A’s value loading.
• Position C ($3.0M specialty chemicals): market β 0.95, sector (materials) +0.40, momentum +0.28.

Gross market-equivalent exposure summed name-by-name: roughly $2.70M + $1.65M + $2.85M = $7.20M. Netted at the portfolio level, value and momentum partially offset, sector legs consolidate into two ETF allocations, and blended portfolio beta lands near 1.08 — about $7.0M market-equivalent, hedged with a $4.2M proxy basket at basket β ~1.65.

The arithmetic matters less than the direction: every additional name with a non-identical factor profile reduces proxy notional per dollar of book. A five- or ten-name book nets further, provided the names are not all the same factor bet in different tickers.

The constraint: netting only helps when factor profiles differ. Five identical small-cap industrials with β 1.3+ do not net. They stack. Diversify the factor mix or accept that the proxy scales linearly.

One Consolidated Basket, Not Three

Three separate overlays mean six to nine tickers, three ETF legs, three sets of chains to monitor and roll. A consolidated overlay needs one basket: one broad small-cap ETF leg plus one or two sector legs, sized to netted loadings.

Fewer tickers buys two things.

First, position size per ticker goes up, which fixes options liquidity. A $600K single-name proxy leg struggles to find 500+ contracts of open interest near the money. A $2.5M ETF leg does not. IWM, XLI, XLF, XLB carry OI in the thousands with penny-wide spreads. Consolidation moves notional from thin single-name chains into deep ETF chains.

Second, operations collapse. One put spread to price, one call stream to roll, one delta to monitor. Rebalance monthly on the portfolio regression, with an intra-month trigger if realized beta drifts more than 0.15. Same rule as the single-name case: rolling 252-day OLS, retention t > 2.0.

The rule for instrument selection does not change — dollar volume above 20× notional, OI above 500 near the money — but it gets easier to satisfy because each allocation is larger and sits in a more liquid instrument.

Diversification on the Premium Side

The cost side — the put spread — scales with proxy notional, so netting cuts it directly. A $4.2M basket costs less to protect than $5.5M across three. At representative pricing (5–8% OTM spread, 60 DTE), that gap returns roughly $8,000–$10,000/month to net carry.

The revenue side improves differently. Short-call premium on a consolidated ETF basket is smoother than on three single-name proxies. Single-name IV spikes around earnings and headlines the proxy is not supposed to track. ETF IV mean-reverts more reliably, which steadies the 25–30Δ monthly call roll. Premium per dollar is slightly lower on ETFs, but the hit rate is higher and forced rolls around idiosyncratic gaps disappear.

The portfolio’s idiosyncratic residual diversifies across names. One name’s 22% residual is concentrated risk. Three names’ residuals, imperfectly correlated, partially diversify. The overlay still does not hedge idiosyncratic risk — but a single restatement dents the book instead of defining it.

Worked Example: Three Positions, One Overlay

Book value $6.5M. Portfolio regression R² 0.74. Retained factors and netted dollar exposures:

Proxy basket: $4.2M total. 55% small-cap broad ETF ($2.3M), 25% industrials ETF ($1.05M), 20% split across financials/materials ETF legs ($0.85M). Blended basket β ~1.67, matching the $7.02M market-equivalent exposure.

Overlay, same architecture as the single-name case — 5–8% OTM put spread at 60 DTE for the hedge layer, 25–30Δ short call at 30 DTE rolled monthly for the premium layer:

• Hedge cost: ~$28,000/month amortized on $4.2M notional.
• Premium collected: ~$148,000/month at blended ETF IV ~28%.
• Net carry: ~$120,000/month, or roughly 22% APY on the $6.5M book.

Run separately, the three overlays would collect more gross premium but spend ~$37,000/month on hedge cost across $5.5M in fragmented notional, with triple the tickets and worse fills. Consolidation does not manufacture yield. It removes cost and variance from the yield already there.

What Changes at Five or Ten Names

Scale helps until factor concentration stops it. Each additional name with a distinct sector or style tilt adds netting benefit and smooths the residual further. Each additional name that duplicates the existing book — another high-beta industrial value name — adds gross exposure with no offset.

The practical ceiling is operational. Beyond eight to ten names, the portfolio regression stays stable but attribution gets harder: isolating which name’s beta drifted requires name-level regressions alongside the portfolio one. Run both. Portfolio regression sizes the overlay; name-level regressions diagnose drift.

The entry condition is unchanged: portfolio R² ≥ 0.70 and liquid proxies for every retained factor. Below that, the book is too idiosyncratic and the overlay is guesswork. Above it, every new name makes the structure cheaper per dollar hedged.