Barrier shifting can create a conservative managed claim for a short barrier-option liability. Payoff dominance, reserve margins and executable hedge protection are distinct.
Barrier shifting and overhedging
Barrier shifting uses a displaced barrier in a comparison claim to manage the risk of a barrier option. Overhedging, in this desk context, uses a conservatively valued managed claim with reduced digital risk. The extra provision and the hedge derived from that claim are separate choices from simply holding more stock. The client’s contract keeps its original barrier.
Koster and Oeltz, §1, describe this managed-product approach: for a short position, seek a higher valuation together with smaller maximum absolute deltas. Near a knock-out, abrupt changes in eligibility can make unwinding a hedge difficult when trading is delayed or prices jump.
Move the comparison barrier, keep the contract
Consider a fictional European up-and-out call on one underlying unit, with prices and cash payments in USD:
Term
Client contract
Comparison claim
Strike
100
100
Upper knock-out barrier
110
112
Monitoring
Continuous over years 0 to 1
Identical
Trigger
Touching the barrier knocks out
Identical
Rebate
Zero
Zero
Settlement
Eligible call payoff at year 1
Identical
The supplied continuous path joins the values 100,108,111,105 by straight lines at times 0,1/3,2/3,1 year. These segments define the whole fictional path, rather than merely connecting isolated observations of an unknown path.
Its maximum is 111. The contract touches 110 and permanently knocks out, paying 0. The comparison never touches 112, so it pays max(105−100,0)=5 at expiry. The client receives zero; the comparison’s five-dollar payoff does not become a contractual entitlement.
Fictional continuous path 100 to 108 to 111 to 105 over one year. It hits the contractual upper barrier 110 but stays below comparison barrier 112. With strike 100 and zero rebate, contract payoff is USD 0 and comparison payoff is USD 5 per underlying unit.
Original fictional piecewise-linear continuous path. Prices and expiry cash payoffs are USD per underlying unit; strike 100; touching counts as a hit; zero rebate. This is a payoff comparison, not an option valuation or hedge backtest.
Why the outward shift is conservative here
Write MT=max0≤t≤TSt. For a zero-rebate upper knock-out, the payoff at barrier H is
XH=(ST−K)+1{MT<H}.
Every path surviving 110 also survives 112. The larger barrier adds survival paths without removing any, and the call payoff is nonnegative. Consequently,
X112−X110=(ST−K)+1{110≤MT<112}≥0.
The strict upper inequality preserves the touch rule: a path reaching 112 loses both claims. A path staying below 110 pays the same under both. A path ending below the strike can pay zero under both even when only the comparison survives.
Under the same pricing model and other terms, an expectation with a positive discount factor preserves this ordering. If the payoffs are integrable, V112≥V110. For a dealer short the client claim, V112−V110 is a nonnegative comparison margin. The path above does not determine that margin: valuation needs a model for possible future paths.
For a lower knock-out, the analogous outward move lowers the barrier. A knock-in instead pays on the hit set, so the same displacement changes eligibility in the opposite direction. Rebates require a fresh cash-flow comparison: a hit can itself pay money, potentially destroying the ordering proved above. The holder’s side also matters when deciding what valuation is conservative.
A larger reserve does not mean a larger delta
Price ordering does not order price derivatives. From V≥V, it does not follow that ∂V/∂S≥∂V/∂S. A managed claim can have a larger value and a smaller absolute delta near the contractual barrier. Its sensitivity must be calculated; the payoff ordering alone proves no risk reduction.
Koster and Oeltz also describe modifying PDE boundary conditions to enforce a delta limit while preserving a conservative value. Their objective is a trade-off between provision and digital-risk reduction. A simple barrier shift need not achieve the desired limit across the whole price surface. See §§1–2.
For comparison, suppose a trader adds five shares, funded by borrowing USD 500 at spot 100. Over one interval with zero interest, no dividends and no transaction costs, the extra position contributes
5(S1−100).
It gains USD 5 if spot reaches 101 and loses USD 5 if spot reaches 99. Buying extra shares therefore adds directional exposure; neither the purchase price nor the share count measures a guaranteed protection benefit. Multiplying a model delta by 105% has no general connection to the conservative managed-claim construction.
Even a dominating comparison payoff is not an executable trading strategy. Guaranteeing coverage through jumps, limited liquidity and discrete rebalancing requires a separate argument about available instruments, funding and trading constraints. A reserve and a changed model hedge alone do not establish dynamic superreplication.
Another barrier shift answers another question
A barrier continuity correction approximates a contract’s discrete monitoring using continuous-barrier pricing. Broadie, Glasserman and Kou, pp. 326–327, derive such a monitoring adjustment under Black–Scholes assumptions. Its purpose is to match an observation convention. The 110-to-112 example instead holds continuous monitoring fixed and constructs a different comparison payoff; no monitoring frequency or universal shift size is inferred.
For the underlying eligibility rules, the Barrier Options lesson develops the pathwise in–out partition and distinguishes contractual monitoring from simulation dates. Its public preview is available; the full lesson requires membership.
Sources
Frank Koster and Daniel Oeltz, A Generic Approach to Manage Gap Risk, §§1–2: managed products, conservative valuation, barrier displacement and delta-constrained PDE management.
Mark Broadie, Paul Glasserman and Steven Kou, A Continuity Correction for Discrete Barrier Options, Mathematical Finance 7(4), 1997, pp. 326–327, especially Theorem 1.1: shifts used to approximate a different monitoring convention.
