The Hedging Basis Trap: Unhedgeable Offshore Rho

A quantitative extension of the onshore–offshore basis framework — divergent hedging products and unhedged rho leakage. It reuses the same $T=1$ generative calibration ($S_0=40$, $r_f=4\%$, $F^{\text{off}}=56.20$, $F^{\text{on}}=60.88$, $\sigma_{\text{ATM}}^{\text{off}}=22\%$).

Abstract. A regulatory firewall does not merely sustain a static cross-currency basis $b$ — it bifurcates the stochastic properties of the onshore and offshore TRY forwards into distinct asset classes. The onshore forward (CBRT-anchored) carries near-zero rate volatility; the offshore forward is a hybrid FX–rates swaption that absorbs the duration-weighted variance of a free-floating basis. We show the “4-vol-point phantom vega” a local desk perceives is not rich FX vega but unhedgeable offshore rho, trace a $+1000$ bps squeeze that wipes out a delta-neutral book with zero spot movement, and derive the offshore-basis rho in closed form. The inflated offshore implied vol is the market-clearing XVA reserve for warehousing this rho leakage.

Keywords: FX options · multi-curve · NDF vs deliverable forward · stochastic cross-currency basis · rho / basis risk · XVA

Context and institutional microstructure

When a regulatory firewall (such as BDDK limits) segments domestic and foreign TRY liquidity, it does more than sustain a static pricing wedge (the cross-currency basis $b$). It fundamentally bifurcates the stochastic properties of the interest rates and, by extension, the derivative products themselves.

1. The onshore product (deliverable forward). Excess TRY liquidity is parked at the Central Bank (CBRT). The onshore rate $r_d^{\text{on}}$ is rigidly tethered to the policy corridor. Because local banks have elastic access to central bank liquidity, the onshore rate exhibits near-zero stochastic volatility ($\sigma_r^{\text{on}} \approx 0$).
2. The offshore product (NDF). Foreign counterparties seeking TRY to settle trades cannot access the CBRT and are restricted by legal quotas. The offshore rate $r_d^{\text{off}}$ acts as an unanchored, free-floating release valve for cross-border liquidity demand. It exhibits massive volatility ($\sigma_r^{\text{off}} \gg 0$).

Because standard FX options are derivatives on the forward exchange rate — not just the spot — this segmentation means the two forwards are entirely different asset classes. We formally map this divergence and quantify the hidden basis risk accumulated by an onshore bank attempting to cross-hedge an offshore-priced option with a zero-volatility onshore forward.

1. The variance wedge: mathematical decoupling of the forwards

Let the spot rate $S_t$ follow the heavily managed physical crawl of the generative model with minimal diffusion ($\sigma_{\text{real}}\approx 4\%$) and jump variance $V_J$. Let the USD rate $r_f$ be constant. By CIP operating strictly within each habitat, the observable forwards traded by the desks are $F_t^{\text{on}} = S_t \exp((r_t^{\text{on}} - r_f)\tau)$ and $F_t^{\text{off}} = S_t \exp((r_t^{\text{off}} - r_f)\tau)$, with $\tau = T-t$.

Applying Itô to the log-forwards (taking spot and rate diffusion orthogonal for clarity), we extract their instantaneous quadratic variations.

Theorem 1 (Habitat variance wedge). The total variance of the hedging instrument strictly depends on the habitat of the rate curve.

Onshore forward (Product B). Because the central bank pegs the domestic rate, $dr_t^{\text{on}}\approx 0$:

$d\langle \ln F^{\text{on}} \rangle_t = \underbrace{\sigma_{\text{real}}^2\, dt}_{\text{Spot Diff}} + \underbrace{V_J\, dt}_{\text{Jump Risk}} + \underbrace{0\cdot dt}_{\text{Rate Vol}}$

The onshore forward is a one-dimensional, pure-FX product. It moves strictly tick-for-tick with the managed spot.

Offshore forward / NDF (Product A). The offshore rate floats freely as a liquidity premium, $dr_t^{\text{off}} = \kappa(\cdots)dt + \sigma_r^{\text{off}}\, dW_t^r$:

$d\langle \ln F^{\text{off}} \rangle_t = \underbrace{\sigma_{\text{real}}^2\, dt}_{\text{Spot Diff}} + \underbrace{V_J\, dt}_{\text{Jump Risk}} + \underbrace{\tau^2 (\sigma_r^{\text{off}})^2\, dt}_{\text{Offshore Rate Vol}}$

The offshore forward is a two-dimensional hybrid FX–rates swaption. It absorbs an entirely separate source of orthogonal risk: the duration-weighted variance of the stochastic cross-currency basis.

2. Disentangling “phantom vega”

When an international dealer quotes a 1Y offshore USD/TRY option at $\sigma_{\text{ATM}}^{\text{off}}=22\%$, they price the total variance of the offshore forward. A local bank observes this $22\%$, compares it to the heavily suppressed spot diffusion ($\approx 4\%$), and falsely concludes the option is “structurally overpriced,” prompting it to aggressively sell USD calls.

This is a category error. Stripping the offshore rate variance from the total implied variance isolates the true underlying volatility of the onshore product. Reusing the $T=1$ generative calibration (total variance $0.0484$, offshore rate-variance contribution $0.0165$):

$\text{Native onshore variance} = 0.0484 - 0.0165 = 0.0319$

$\sigma_{\text{ATM}}^{\text{on}} = \sqrt{0.0319} = 17.86\%$

The $4.14$ volatility-point gap ($22.00\% - 17.86\%$) is not “rich FX vega.” It is offshore rho volatility. When a local bank sells an option at $22\%$ and hedges with an onshore forward, it believes it is capturing an FX arbitrage; in reality it is selling unhedged interest-rate insurance to London.

3. The cross-hedging trap (worked example)

To show why cross-pollinating these products is catastrophic, we trace the mark-to-market P&L of a local bank caught in a liquidity squeeze.

Initial state ($t=0$). Anchors $T=1$, $S_0=40.00$, $r_f=4\%$:

• Offshore curve: $r_d^{\text{off}}=38\% \implies F^{\text{off}}=56.20$.
• Onshore curve: $r_d^{\text{on}}=46\% \implies F^{\text{on}}=60.88$.
• The trade: the bank sells a 1Y $K=70$ USD call (cash-settled offshore). At the correct $F^{\text{off}}$ and $\sigma=26.65\%$, the Prop-1 discounted upfront collected is $1.310$ TRY.
• The hedge: the risk system reports spot delta $\Delta=0.245$. Because BDDK limits prevent offshore swap hedges, the bank buys $0.245$ USD via the zero-volatility onshore forward at $60.88$. The book is now “delta-neutral.”

The shock. Overnight, foreign funds scramble to close short-TRY positions, triggering a severe offshore liquidity squeeze:

1. The offshore swap rate $r_d^{\text{off}}$ spikes $+1000$ bps (from $38\%$ to $48\%$).
2. The CBRT defends the local market; $r_d^{\text{on}}$ stays anchored at $46\%$.
3. Spot $S_t$ stays heavily managed, pegged at $40.00$.

Revaluation of the short-call liability. Because settlement is offshore, the option’s drift is governed by $r_d^{\text{off}}$, and the offshore forward surges:

$F_{\text{new}}^{\text{off}} = 40.00 \times \exp\!\big((0.48 - 0.04)\times 1.0\big) = 62.11$

Recomputing Black-76 at $F=62.11$ (implied vol held constant):

• $d_1 = -0.3160 \implies \Phi(d_1) = 0.3761$
• $d_2 = -0.5825 \implies \Phi(d_2) = 0.2802$
• undiscounted $c_{\text{fwd}} = 62.11(0.3761) - 70(0.2802) = 3.745$ TRY
• new discount factor $P_d^{\text{off}} = e^{-0.48} = 0.6188$
• new liability PV $= 3.745 \times 0.6188 = 2.317$ TRY
• loss on the option: $1.310 - 2.317 = \mathbf{-1.007}$ TRY

Revaluation of the onshore hedge. Because the CBRT held $r_d^{\text{on}}=46\%$ and spot at $40.00$, the onshore forward does not move a single pip: $F_{\text{new}}^{\text{on}}=60.88$, hedge MTM $= 0.000$ TRY.

Table 1 — Cross-hedge P&L under the squeeze ($+1000$ bps offshore, spot and onshore flat).

Leg	Before ($t=0$)	After (squeeze)	MTM (TRY)
Short call liability (PV)	1.310	2.317	−1.007
Onshore forward hedge	$F^{\text{on}}=60.88$	$F^{\text{on}}=60.88$	0.000
Net book			−1.007

The bank suffers an unhedged wipeout of −1.007 TRY per USD nominal on a “delta-neutral” book — while the underlying spot experienced absolute zero volatility.

4. Analytical proof of the rho trap

The bank blew up because it was entirely naked to the stochastic offshore basis. We formalize this by deriving the offshore basis rho $\rho_{\text{PV}}^{\text{off}} = \partial V_0^d/\partial r_d^{\text{off}}$ under the Prop-1 multi-curve framework.

For a base-currency-collateralized option discounted strictly by the offshore domestic curve ($V_0^d = e^{-r_d^{\text{off}} T} c_{\text{fwd}}$), the product rule with the chain rule $\partial c_{\text{fwd}}/\partial F^{\text{off}} = \Phi(d_1)$ (and $\partial F^{\text{off}}/\partial r_d^{\text{off}} = T F^{\text{off}}$) gives

$\rho_{\text{PV}}^{\text{off}} = T\, e^{-r_d^{\text{off}} T} \big[ F^{\text{off}} \Phi(d_1) - c_{\text{fwd}} \big].$

Substituting $c_{\text{fwd}} = F^{\text{off}}\Phi(d_1) - K\Phi(d_2)$, it collapses to the standard continuous-time domestic rho:

$\rho_{\text{PV}}^{\text{off}} = T\, e^{-r_d^{\text{off}} T}\, K\, \Phi(d_2).$

Intuition: a USD call is fundamentally an option to pay $K$ TRY in the future. When offshore TRY rates spike, the present value of that fixed TRY strike payment falls, making the call structurally more valuable.

Evaluating at the initial state ($T=1$, $r_d^{\text{off}}=0.38$, $K=70$, $\Phi(d_2)=0.1691$):

$\rho_{\text{PV}}^{\text{off}} = 1.0 \times 0.6839 \times 70 \times 0.1691 = 8.093\ \text{TRY per 100\% rate shock}.$

A $+1000$ bps ($+0.10$) linear shock predicts an immediate loss of −0.809 TRY. The true non-linear loss of −1.007 TRY reveals the severe negative convexity (basis vanna/gamma) the bank unknowingly warehoused.

Conclusion: the regulatory consequence

When interest-rate volatility is heavily anchored for one party (onshore) but structurally unanchored for another (offshore), treating the hedging instruments as fungible guarantees catastrophe. By delta-hedging an offshore option with an onshore forward, the local bank zeroes its spot delta but runs a naked short position on the cross-currency basis. The inflated $22\%$ offshore volatility is the exact, market-clearing XVA reserve required to warehouse this unhedgeable rho leakage.

---

Research/educational only; not investment advice. All figures are illustrative and reuse the onshore–offshore basis article’s calibration; day-count and compounding conventions shift third decimals but not the argument. Numerical values independently verified.