1

Hidden Risks and Optionalities in American

Options
Noura El Hassan∗ , Bacel Maddah† , Nassim Nicholas Taleb†‡
∗ American University of Beirut–Mediterraneo

Paphos, Cyprus
† American University of Beirut, Beirut, Lebanon
‡ Universa Investments

nnt@[Link]
arXiv:2602.14350v1 [[Link]] 15 Feb 2026

Abstract—We develop a practical framework for identifying ∆a, the mean deviation of a, and estimating the convexity
and quantifying the hidden layers of risks and optionality bias π∆a
embedded in American options by introducing stochasticity
into one or more of their underlying determinants. The
heuristic approach remedies the problems of conventional Z K
1
pricing systems, which treat some key inputs deterministi- π∆a ≈ f (x | ā + ∆a)
cally, hence systematically underestimate the flexibility and −∞ 2
convexity inherent in early-exercise features.  Z K
+ f (x | ā − ∆a) dx − f (x | ā) dx. (3)
−∞

Furthermore, particularly in the case of options, even if a

I. U NACCOUNTED O PTIONALITY pricing approximation is used, the result may not illuminate
A. A note on model nonlinearity as fragility us on option valuation but will give us a degree of model risk.
Fragility to model error has been mapped in terms of Under the principle in [1], a bad ruler might not give us the
convexity [1], and its heuristic testing presented below applied precise height of a growing child, but will inform us whether
among others by the IMF to gauge portfolio risks of banks, the child is growing. As we are looking for fragilities, this
see [2]. allows us some approximations that work well with otherwise
The logic is as follows. Having the right model, but being computationally onerous American options.
subjected to parameter uncertainty will invariably lead to an
expected increase in model error in the presence of convexity B. Application to American Options
and nonlinearities, particularly when the second order effect American options differ from their European counterparts
is not negligible. Assume f (.) an estimated function: in allowing early exercise. This single feature introduces a
f (x | a), (1) set of nonlinearities and latent exposures that remain largely
invisible to conventional risk systems.
where a is fixed, assumed to be the average or expected A standard option O(.) is a function
parameter, taking u as the distribution of a over its domain
A: Z O(S, K, σ, T, r1 , r2 ), (4)
ā = a u(a) da. (6) where S is the underlying security price at time 0, K the strike
A price, σ the volatility, T the time to nominal expiration, r1 the
The mere fact that a is uncertain (since it is estimated) might funding rate, and r2 the “carry” of the underlying (which could
lead to a bias if we perturbate from the outside (of the integral), be the discrete dividend or continuous foreign rate).
i.e. stochasticize the parameter deemed fixed. Accordingly, Under conventional pricing models (starting with [3] ),
fragility to model error πA is easily measured as the difference only S is stochastic. In further refinements and adaptations,
between (a) f integrated across values of potential a and (b) σ is treated as stochastic, with a rich literature [4], [5],
f estimated for a single value of a deemed to be its average. [6], see [8] for a review. We note that stochasticity of an
The convexity bias π becomes additional variable entails additional parameters, particularly
the centering and scale of the stochastic variable.
Z Z European Options do not heed the stochasticity of r2 , or,
πA = f (x | a) u(a) da dx − rather the differential between r1 −r2 as it is entirely inherited
X A
Z  Z  in the volatility of the forward F = Ser1 −r2 t, where t is
f x a u(a) da dx. (2) the time period, which can be captured by expert operators
X A who typically use the volatility of the latter (at the nominal
This can be approximated by an interpolated estimate ob- maturity) instead of that of the spot. However American
tained with two values of a separated from a mid-point by options are specially –and seriously– affected by both r1 and
 2

r2 . In what follows, after discussing some empirical episodes, The resolution was conceptually simple yet operationally
we will focus on injecting the dynamics of r1 and r2 . decisive: exercise the calls up to the amount of the short
(by the hedge ratio). By doing so, the trader obtained the
**
shares and neutralized the squeeze. Had the options been
The remaining part of this article is organized as follows. European, early exercise would have been impossible, and
We present the real world problems as encountered by option the losses potentially catastrophic. The episode demonstrated
operators in section II, discuss their typology in section III, that the American call possesses not only market optionality
briefly link model error to fragility in section I-A, present but also “model error optionality”–the ability to adapt to
the master equation and the various possible dynamics and unexpected discontinuities in the underlying or in the market
probability distributions for pricing in section in section IV. microstructure. We note that such optionality can be modeled
We perform a broad set of calculations under different models with a jump in the financing rate.
showing the robustness of our findings in the final sections. Practitioner Episode 3: The Equity Index Squeeze: A related
mispricing witnessed by the corresponding author occurred in
II. I LLUSTRATIVE P RACTITIONER E PISODES the period covering 1998-1999 (in the wake of the failure of
Practitioner Episode 1: The Currency Interest Rate Flip: the hedge fund Long Term Management). It concerned long-
Not only the sign of (r1 − r2 ) can vary, but it can flip from dated, over-the-counter European call options on an equity
positive to negative –and the tradition for option pricing and index. These instruments traded at prices corresponding to
hedging systems is to use a flag to price either the put or the volatility levels far below any plausible historical measure.
call as if they were European since the early exercise feature Traders were long the calls and short the index futures, contin-
can be ignored. uously rebalancing as the market rose slowly but substantially.
During the 1980s, German interest rates were generally The problem is that the rebalancing led to an increase of short
lower than those in the United States. In such a configuration– futures. They lost on the futures (which for these contracts
where the foreign rate is below the domestic rate–standard were to be settled daily with an outflow of cash), but were
pricing systems value the American put on a currency pair unable to monetize gains on the options, which remained
higher than its European counterpart, while assigning identical heavily discounted.
values to the corresponding calls. At one point, the options were offered below their intrinsic
When interest rates later converged, and subsequently re- value relative to the forward (at a standard funding rate)–an
versed following the post-reunification rise in German yields, apparent market inefficiency. Yet, capital constraints prevented
many believed they were executing an arbitrage-free trade by arbitrage, as carrying the long position required margin and
selling the American option and buying the European one. funding, not available to risky positions during that period.
Initially, their mark-to-model valuations appeared profitable, Earlier, during the crash of 1987, similar distortions were
as the systems treated both options as equivalent. However, observed when the cash-futures discount widened to nearly
when interest rate differentials inverted, the mark-to-market 10% –an arbitrage that failed to attract operators owing to the
values diverged dramatically. The models, which had ignored stress on the financial system.
the early-exercise possibility of such options, failed to capture With European options, such dislocations can become ter-
the exposure. Several trading desks incurred significant losses minal to a trading desk, that is, they threaten extinction.
before realizing that the American call carried embedded By contrast, the American contract provides a lower bound
optionality on the path of the rate differential. to adverse mark-to-market movements (and an option on
Similar opportunities reappeared during subsequent cur- funding rates): its early-exercise right effectively caps the
rency crises and devaluations, whenever interest rates became degree of mispricing to which the holder can be exposed. This
unstable. The pattern was recurrent: volatility in the rate differ- feature embodies an additional, often unrecognized, layer of
ential would amplify the hidden optionality of the American convexity.
instrument, while the European remained constrained by its
terminal payoff structure.
III. D IFFERENTIAL VALUATION C ASES
Option operators were unaware of the risks since both
the academic literature and option software designers (an Case 1: Convexity to Changes in the Carry: Consider an
overlapping community) did not count for it –even stochastic underlying forward and spot both initially at 100, and a
volatility wasn’t even implemented then prior to the late 1990s one-year at-the-money European and American call. Under
[9]. conventional pricing systems, both instruments will be marked
Practitioner Episode 2: The Stock Squeeze: In the early identically.
2000s, the corresponding author was confronted a problematic If the underlying rallies to 140, both options converge to
position: his desk was long listed American calls on an parity, each worth $40. However, assume that interest rates rise
Argentinian stock and short the corresponding delta amount to 10%. The European option’s value becomes the discounted
(hedge ratio) in the underlying shares. The stock, an obscure intrinsic value–approximately $36.36–while the American op-
ADR, was delisted unexpectedly, forcing an urgent buy-in. tion, which can be exercised immediately, retains a value of
No liquidity was available, and attempts to borrow the stock– $40.
ironically through the firm Bear Stearns at the time–proved Thus, a change in the carry–here, the discounting
futile. environment–benefits the American option disproportionately.
 3

The European price is anchored to a fixed maturity, while the First, using the earlier notation in eq. 4 we write down the
American’s exercise flexibility preserves nominal value under price of an American option at time 0 and underlying price S
higher rates.
Case 1B: Asymmetric Rate Shifts: Assume now that only
the domestic rate increases to 10%, with the spot unchanged at
140. The forward declines to roughly 126. The European call, h i
valued off this forward, drops to approximately the present OA (S, K, σ, t, r1 , r2 ), = sup EQ e−r1 τ g(Sτ ) (5)
τ ∈T0,T
value of 26, or $23.64. The American call, which may be
exercised immediately, remains worth $26.
In both scenarios, the American option systematically out-
performs the European because it benefits from convexity to
the interest rate differential. Any model that prices the two where g(S) is the payoff function (intrinsic value) at exercise:
identically under changing carry assumptions is misspecified. g(S) = max(Φ(S − K), 0) where Φ = +1 for a call
From this, a general principle follows: if option A is worth option and Φ = −1 for a put option, Tt,T is the set of
at least as much as option B in all scenarios, and strictly more all stopping times τ such that t ≤ τ ≤ T almost surely,
in some, it is suboptimal to sell option A and buy option B EQ [ ·] is the conditional expectation under the risk-neutral
at equal prices. Yet this qualitative inequality still leaves open probability measure Q, given the information available at time
the quantitative question–how much more should one pay for t and, finally, Q is the risk-neutral (equivalent martingale)
the flexibility? measure. Now eq. 5, the "master" equation does not specify
Case 2: Sensitivity to Changes in the Foreign or Dividend methodologies.
Rate: Let S = F = 140 with both domestic and foreign
rates initially at zero. Again, the European and American Owing to the path dependence of American options, their
options start at the same model price. Suppose the foreign rate pricing has always been fraught with difficulties, even in the
rises sharply to 20%. The forward now appreciates to roughly very standard situation when only S is stochastic.
Se(rd −rf )T ≈ 1.16S.
Note that conventional Monte Carlo methods are ill-suited to
The European call, lacking early exercise, is now worth only
capturing this additional stochasticity, as the stopping time is
about $16 (its discounted intrinsic value). The American call,
path-dependent and endogenous. More sophisticated numerical
however, retains the full intrinsic value of $40. The rationale
approaches–such as least-squares Monte Carlo or hybrid ana-
is straightforward: the American option dynamically selects
lytical methods–are required to quantify the magnitude of this
the more favorable exercise basis–cash or forward–depending
latent premium. In practice, however, even ordinal (directional)
on which maximizes its immediate payoff. It "chooses" the
comparisons can reveal substantial model risk when early-
superior underlying, adapting endogenously to the change in
exercise rights are ignored.
rate environment.
Case 3: Sensitivity to the Yield Curve Slope: Consider now Some complexity arises from the uncertainty of the hedge
a non-flat term structure, such as those frequently observed horizon for the underlying. The effective forward hedge of
around year-end or policy rollovers. When the yield curve an American option is unknown, since the exercise time is
contains inflection points, the conventional valuation using stochastic. The situation resembles that of a barrier option
only the terminal forward rate FT becomes unreliable. with an uncertain trigger: termination depends on multiple
Intermediate fluctuations in the carry can significantly affect stochastic variables, including volatility, the base rate, and the
the American option’s value, as the optimal exercise point rate differential.
may occur precisely at one of those kinks. A pricing or risk-
management system that collapses the full term structure into A hidden risk arises from the following. Intuitively, the
a single terminal forward will therefore misprice the American "smart" American option positions itself, in principle, at the
option–often marking it equal to the European, when in fact point on the forward curve that maximizes its discounted
it should be higher. value. A risk-management system that allocates all forward
The intuition is clear: an American option allows the holder delta exposure to the terminal maturity–treating the forward
to "lock in" the forward at any intermediate date, capturing as if exercise can occur only at T –commits a structural
transient peaks in synthetic carry. The European option, con- error. Such systems underestimate the embedded additional
strained to final maturity, lacks such adaptability. optionality and misstate both value and hedge sensitivities.

In summary, American options possess multiple layers of

IV. P RICING I MPLEMENTATIONS unaccounted convexity beyond their explicit early-exercise
The preceding examples illustrate that the value differential feature. These include sensitivity to stochastic rates, curvature
between American and European options grows with the in the term structure, model error, and liquidity constraints.
volatility of interest rates and the curvature of the term Properly accounting for these requires stochasticizing the
structure. The greater the uncertainty in the path of the carry, underlying rate processes and evaluating expected value under
the larger the unpriced optionality embedded in the American the distribution of exercise times–a problem intimately linked
contract. to the concept of the fugit.
 4

Comment 1 an American option". Taleb terms this result "Omega". The

formula is
The difference between various methods should be minor
compared to parameter uncertainty. We are looking for ρ2A
Ω=t
, (6)
the first order effect of the stochasticity in rates, largely ρ2E
to gauge the magnitude of the hidden risk ignored so where t is the nominal time to expiration, ρ2A and ρ2E are
far. "Rhos", the sensitivities of the American and European options
Risk management is about scenario analyses across a to changes in the underlying nominal carry yield.
parameter set, not precise pricing; our approach allows 2) The Stochastic Fugit: Distribution of Exercise Times: A
parametrization. deterministic-rate American pricer (binomial, finite-difference,
or least-squares Monte Carlo) naturally yields:
• a discrete set of candidate exercise times t1 , . . . , tK ,
V. I NTEGRATING AN A MERICAN O PTION ACROSS • the corresponding exercise probabilities pk = P(τ = tk ).
S TOCHASTIC R ATES
This defines the stochastic fugit:
In short, in what follows, we try the following simplified
heuristics to grasp the hidden exposure. All are based on a T̃ ∈ {t1 , . . . , tK }, P(T̃ = tk ) = pk .
separation of OA using a separation of the sort used in eq. 2, The classical deterministic fugit is merely the expectation
that is integrating OA across r1 or r2 . K
X
• Method 1- One single integration OA across τ ∗ = E[T̃ ] = pk tk . (7)
k=1
stochastic rates at a distribution of optimal stopping
times, the " fugit based heuristic". By retaining the full distribution (tk , pk ), we preserve the
• Method 2- Multiple integrations of OA across time convexity inherent in the early-exercise feature.
stochastic rates at a given optimal stopping time τ , The fugit provides a principled estimate of the expected
the "fugit". exercise horizon for use in the rate distribution. It adjusts
automatically to the option’s moneyness. This heuristic cap-
Let us use the shortcut OA (r1 , r2 , t) to denote the price tures first-order effects of rate uncertainty without solving a
of an American option computed under a deterministic carry full two-factor PDE. It can be extended by integrating over
and foreign or dividend rates r2 using any standard numerical a discrete distribution of fugit times tk from a Bermudan
method (binomial, lattice, or PDE). We wish to approximate exercise histogram,
the price of the same option when either r1 or r2 is stochastic, X Z
OA ≈ pk OA (r, t) fr(tk ) (r) dr, (8)
by integrating over the distribution of the stochastic rate(s) at
k
the effective exercise time. We note that perturbations for r1
∗
can cover squeezes of financing (in section III), while r2 can where pk = P(τ = tk ).
cover changes in the security yield, which includes dividends. 3) Fugit-weighted integration heuristic: We define the
fugit-weighted American price under rate stochasticity as
Z
A. Various stopping times methodologies OA,τ ∗ = OA (r, t) fr(τ ∗ ) (r) dr, (9)
We first proceed by assuming that one of the rates is Dr
stochastic, then expand for both assuming either independence where fr(τ ∗ ) is the density of the stochastic rate evaluated at
or some correlation between the rates. the expected stopping time τ ∗ . This represents a weighted av-
1) Single expected Stopping Time: The "Fugit"-based erage of deterministic-rate American prices, with the weights
Heuristic, see [9], is as follows. Let τ ∗ be the expected given by the probability distribution of the relevant rate at the
discounted stopping time of the American option, measured fugit time.
in risk-neutral time units. If τ denotes the random stopping
time (optimal exercise), then the fugit is defined as B. Extension to Two Stochastic Rates
Z τ  When both rates r1 (t) and r2 (t) are stochastic, possibly
∗ Q −r(u−t0 )
τ (S0 , t0 ) = E e du , correlated, the extension is immediate:
t0
(stoch-fugit)
which can be interpreted as the "effective maturity" or the OA =
time-to-exercise that discounts equivalently to the American L
X ZZ
payoff. For European options, τ ∗ = T − t0 ; for deep-in-the- pl OA (r1 , r2 , tl ) f(r1 ,r2 )(tl ) (r1 , r2 ) dr1 dr2 . (10)
R2
money Americans, τ ∗ is substantially shorter. l=1
This quantity can be estimated directly from a binomial If independence is assumed,
or finite-difference grid as the expectation of discounted time
f(r1 ,r2 )(tk ) (r1 , r2 ) = fr1 (tk ) (r1 ) fr2 (tk ) (r2 ).
spent before exercise.
A trick is proposed by [9] as a "shortcut method... to
find the right duration (i.e., expected time to termination) for
 5

C. Various distribution of rates

Let the funding rate r1 (t) or carry rate r2 (t) follow one of
the canonical short-rate dynamics:
a) Bachelier or normal world.:

dr = µr dt + σr dWt

⇒ rτ ∗ ∼ N (r0 + µr τ ∗ , σr2 τ ∗ ).

b) Vasicek / Hull–White world.:

dr = κr (θr − r) dt + σr dWt

∗

E[rτ ∗ ] = θr + (r0 − θr )e−κr τ ,
⇒ σ2 ∗
Var[rτ ∗ ] = r (1 − e−2κr τ ).
2κr
Fig. 2. Optionality versus standard deviation for a currency put option under
c) Lognormal world.: a lognormally distributed local interest rate with OA(r∗ ) = 15.1700 and
S/K = 1.00
dr2
= µr dt + σr dWt
r

√
 
1
⇒ rτ ∗ = r20 exp (µr − σr2 )τ ∗ + σr τ ∗ Z ,
2
Z ∼ N (0, 1).

**
The end result for us is testing , where e.(r) denotes
stochasticity over parameter r, O ^A (r) − OA (r), the extra
optionality, after clearing a few hurdles.
We will perform tests to establish whether the fugit shortcut
represents a good enough an approximation and whether
various rate dynamics (presented in the next section) make Fig. 3. Optionality versus standard deviation for a currency call option under
a difference for the convexity bias. a Hull-White distributed local interest rate with OA(r∗ ) = 12.7779 and
S/K = 1.00

VI. M AIN N UMERICAL I MPLEMENTATION

We work throughout these simulations with options with –to
normalize –a maturity of one year, hence no loss of generality.
We consider equity puts, currency puts, and currency calls with
stochastic local rate.
In our base example, we consider a generic at-the-money
American equity put option on a high volatility stock, with
the following common parameters, volatility (put on the upper
end of common values), σ = 40%, maturity, T = 12
months, initial underlying asset value, S0 = 100, strike price,
K = 100, and a stochastic interest rate with an initial value
r0 = 1%. (With respect to the notation in Section I of
this paper, this is an American option with r1 = r and
r2 = 0.) For the stochastic rate, we assume that it follows
Fig. 1. Optionality versus standard deviation for a an equity put option under
a normally distributed local interest rate with OA(r∗ ) = 14.3184, S/K = a Bachelier process (that is, normally distributed as defined in
1.00 Section V-C), with mean r̄ = 4.18% and standard deviation
σr = 1.28% at maturity, as described in Section V-C of this
 6

paper. These parameters align with recent values of the 1-year VII. F URTHER N UMERICAL R ESULTS AND T ECHNICAL
US treasury rate. Specifically, we consider the 1-year treasury D ETAILS
yield over the past three years [14], where the yield is reported In Section VII-A, we briefly discuss the binomial lattice
at the end of every month from January 2022 until March method ([16]) used to compute the integration and to obtain the
2025. expected fugit. In Section VII-B, we investigate the proposed
We start by computing the stopping time using expectation heuristics in (6) and then compare the results with the expected
of the classical deterministic fugit using (7). Accordingly, fugit estimated from the lattice in Section VII-A, which can
we compute the American option value, O eA (r), under a be seen as the “exact" baseline. In Section VII-C, we study
normal interest rate using (9). The parameters of the interest how changes in the dynamics of interest rate, under alterna-
rate process are determined based on matching the first two tive distributional assumptions, affect our results. Finally, In
moments of the rate at maturity, rT , with r̄ and σr . For Section VII-D, we briefly exhibit an alternative approach of
instance, under a normally distributed interest rate, the drift measure optionality by comparing the American option value
r̄−r0
√ obtained by moment matching as µ = T
and volatility are to a European counterpart.
and σ = σr / T . The binomial lattice is used to compute
the American option price for a fixed interest rate, taking A. Binomial Lattice and Fugit Distribution
into consideration the early exercise feature. Then, using the This section presents the numerical framework used
Gauss-Hermite quadrature we approximates the expectation throughout the paper. The starting point is the computation
of this price with respect to the stochastic interest rate by of τ ∗ = E[T̃ ] in (7), the expected optimal stopping time of an
evaluating the lattice-based price at a finite number of carefully American option using the classical deterministic fugit. Then,
chosen interest rate realizations and aggregating them using we employ the binomial lattice to compute the value of an
predetermined weights. More details about the approach can American option with stochastic local rate, OA(r) e , through
be found in section ??. numerical integration.
We compare the results with the corresponding deterministic Following the notation described in [15], we briefly recall
price of an American option, with the local rate being equal the elements of the lattice that are required for the computation
to the average interest rate at the expected fugit τ ∗ , OA (r∗ ). of the expected fugit and the subsequent numerical valuation.
We then estimate Let the maturity T be divided into n time steps of length
δt = T /ns . In accordance with [15], we set n = 2000, which
eA (r) − OA (r∗ ),
πA = O (11) provides stable and accurate numerical results. The stock price √
evolves on the lattice according to the multipliers u = eσ δt
(rf −δ)δt
as a measure of the gain from the hidden optionality of the and d = u1 , with a risk–neutral probability q = e u−d −d ,
American option. We compute πA and present the results in where rf denotes the continuously compounded risk–free rate
Table I. To investigate the impact of interest-rate uncertainty, and δ the dividend yield (or foreign rate in the currency case).
we vary the standard deviation parameter of the interest-rate Starting with S(i, j), the stock price at time step i and state j,
process σr . As expected, the difference πA increases with the and by P (i, j) the American option value can be obtained
variability of the interest rate. When the volatility is set to zero, through backward recursion. At maturity, i = n, the final
the stochastic model converges to the deterministic case and payoff is P (ns , j) = max(K − S(ns , j), 0) for a put option
the difference becomes exactly zero. This monotonic increase (with an analogous expression for a call). For i < n, the
in πA as standard deviation rises is clearly observed in Figure value of the option is the maximum between the exercise
1. We repeat the same numerical experiments under a local payoff and the continuation value, P (i, j) = max{K −
rate following Geometric Brownian motion and Hull-White S(i, j), Vc (i, j)},h where the continuation value is given i by
processes, as defined in Section V-C, and present the results −rf δt
Vc (i, j) = e q P (i + 1, j + 1) + (1 − q) P (i + 1, j) . The
in Figures 2 and 3, respectively. We observe a similar behavior,
as discussed in more details in section ??. backward recursion simultaneously determines the optimal
exercise region. Accordingly, we can define an early exercise
σr O
eA (r) πA indicator
(
0.0000 14.3184 0.0000 1, if K − S(i, j) ≥ Vc (i, j),
0.0028 14.3190 0.0006 I(i, j) =
0.0078 14.3231 0.0048 0, otherwise.
0.0128 14.3314 0.0130
0.0178 14.3445 0.0261 The optimal stopping time is therefore the first index
0.0228 14.3629 0.0446
0.0278 14.3878 0.0694 τ = min{i ≥ 0 : I(i, ji ) = 1},
0.0328 14.4162 0.0978
0.0378 14.4492 0.1308
0.0428 14.4892 0.1708 with corresponding point in time t∗ = τ δt.
0.0478 14.5319 0.2136 To compute the distribution of τ , the probability mass func-
TABLE I
O PTIONALITY AS A FUNCTION OF THE STANDARD DEVIATION OF A tion is propagated forward through the lattice while enforcing
NORMALLY DISTRIBUTED INTEREST RATE WITH OA(r ∗ ) = 14.3184, the optimal stopping rule. Let π(i, j) denote the probability of
S/K = 1.00 reaching node (i, j) without prior exercise, with initialization
π(0, 0) = 1. For each time step i = 0, . . . , ns − 1, the below
strategy is followed
 7

• If I(i, j) = 1, the probability π(i, j) is recorded as set Pf (n) = Pf (n) + π(n, j).
stopping at time index i and is not propagated further.
• If I(i, j) = 0, the probability evolves according to the Step 7
Expected fugit (conditional on exercise):
risk–neutral dynamics, Set ti = i δt
Pnfor i = 0, . . . , n.
ti Pf (i)
π(i + 1, j + 1) = π(i + 1, j + 1) + q π(i, j), Set Et = Pi=0 n .
i=0 Pf (i)
π(i + 1, j) = π(i + 1, j) + (1 − q) π(i, j). (12) Output
Return the stopping distribution {Pf (i)}n
i=0 and
Denoting by Pf (i) the probability of optimal exercise at the expected fugit Et .
step i, the expected fugit conditional on exercise is Once the expected fugit has been obtained, the same bi-
Pns
(i δt) Pf (i) nomial lattice is used to evaluate the optimality measure πA .
τ ∗ = E[t∗ | exercise] = i=0
P ns . First, a deterministic benchmark is computed by evaluating
i=0 Pf (i) the lattice with an interest rate r∗ = E[r(t∗ )], . For instance,
A complete algorithm to compute the fugit pmf Pf (i) and under a normally distributed interest rate, the deterministic
expected value τ ∗ is presented next. While the algorithm is benchmark rate is simply E[r(t∗ )] = r0 +µτ ∗ . Having defined
straightforward, we could not identify any similar approaches the deterministic value OA(r∗ ), the optimality measure πA
in the literature. There are some online forums hinting to it is obtained by comparing OA(r∗ ) with the option value that
without enough details. The following complete algorithm can accounts for a stochastic interest rate, OA(r).
e Accordingly,
be seen as a side-contribution of this paper. OA(r) is obtained using the obtained expected fugit and the
e
integration in (9), which is evaluated numerically using the
Algorithm for the fugit distribution. Gauss-Hermite quadrature (e.g. [13]). This numerical integra-
Step 1
Input the values for S0 , K, T , n, R, δ, and tion method approximates the expectation by a weighted sum
σ. of lattice prices evaluated at optimally chosen rate nodes.
Set δt = T /ns . Construct the CRR parameters u,
d, and q (see [15].
B. Expected Sopping Time Ω
Step 2
Build the recombining stock-price lattice In this section, we compare the effective stopping time
{S(i, j)}0≤i≤n, 0≤j≤i . developed by Taleb [9] using (6) and using the expectation
of the classical deterministic fugit in (7), for an equity put,
Step 3
Initialize the terminal option values at currency put, and currency call.
maturity: [Link].1 Equity Put: We consider an equity put option with
For j = 0, . . . , n, set P (n, j) = Φ(S(n, j)), where the input parameters mentioned in Table I. We estimate
Φ(S) = K − S. OA (r) from the binomial lattice method, at the deterministic
benchmark, as explained in Section B.I. The European call
Step 4
Backward recursion (pricing + exercise OE (r) is valuated based on the Black-Scholes-Merton formula
indicator): ([17]). We first estimate the American and European option
For i = n − 1, . . . , 0: sensitivities with respect to the interest rate, ρ2A = ∂O∂rA (r)

For j = 0, . . . , i:  ∂OE (r)

and ρ2E = ∂r , using a common heuristic based on central
1
Set Vc (i, j) = R q P (i + 1, j + 1) + (1 − q) P (i + 1, j) . difference method,
Set P (i, j) = max{Φ(S(i, j)), Vc (i, j)}.
If Φ(S(i, j)) ≥ Vc (i, j) set I(i, j) = 1, else set
I(i, j) = 0. ρ2J =
|OA (r + 0.01, t) − OA (r, t)|+|OA (r − 0.01, t) − OA (r, t)|
Step 5 ,
2
Forward recursion (probabilities of reachable
nodes and stopping distribution):
J = A, E.
Create an empty matrix π of size (n + 1) × (ns + 1)
and set π(0, 0) = 1. We use the values of ρ2A and ρ2E to obtain the effective
Create an empty vector Pf of length (ns + 1) and stopping time associated with the early exercise feature of
set Pf (i) = 0 for all i. an American option using (6). We also compute compute the
For i = 0, . . . , n − 1: stopping time using expectation of the classical deterministic
For j = 0, . . . , i:
If π(i, j) = 0, continue. fugit using (7). We obtain close results, as shown in Table
If I(i, j) = 1, set Pf (i) = Pf (i) + π(i, j) (stop here). II, which confirm the validity of the heuristics approached
Otherwise propagate: suggested by Taleb [9]. The results also show how τ ∗ changes
π(i + 1, j + 1) = π(i + 1, j + 1) + q π(i, j), with moneyness. As expected, a lower moneyness level implies
π(i + 1, j) = π(i + 1, j) + (1 − q) π(i, j). an earlier exercise time, i.e., the expected fugit is increasing
Step 6
in the moneyness. Figure 4 also demonstrates this graphically.
Maturity handling:
For j = 0, . . . , n, if π(n, j) > 0 and Φ(S(n, j)) > 0,
 8

S/K ρA ρE τ∗ = Ω τ ∗ = E[T̃ ]
0.8000 0.29931 0.4800 7.4833 6.6954
0.9000 0.3204 0.4353 8.8335 7.7924
1.0000 0.3047 0.3804 9.6117 8.8781
1.1000 0.2728 0.3230 10.1349 9.4785
1.2000 0.2348 0.2682 10.5033 9.9259
1.3000 0.1964 0.2190 10.7641 10.2288
1.4000 0.1611 0.1764 10.9622 10.4821
TABLE II
E FFECTIVE STOPPING TIME τ ∗ , IN MONTHS , FOR AN EQUITY PUT OPTION .

Fig. 5. Effective sopping time τ ∗ versus moneyness for an American currency

put

K S T σ rf r0 r̄ σr
100 80 1y 40% 10% 3% 4.18% 1.28%
TABLE V
I NPUT PARAMETERS FOR THE BASE CASE FOR CURRENCY CALL OPTION
UNDER STOCHASTIC LOCAL RATE

S/K ρA ρE τ∗ = Ω τ ∗ = E[T̃ ]
0.8000 0.1900 0.2226 10.24196 9.8729
0.9000 0.2719 0.3399 9.60074 9.2454
1.0000 0.3470 0.4722 8.8180 8.4378
1.1000 0.4011 0.6123 7.8608 7.4642
1.2000 0.4226 0.7548 6.7186 6.3336
Fig. 4. Effective sopping time τ ∗ versus moneyness for an equity put option. 1.3000 0.4013 0.8955 5.3769 5.0576
1.4000 0.3290 1.0322 3.8250 3.5660
TABLE VI
1) Currency Put: In this section, we consider a currency E FFECTIVE STOPPING TIME τ ∗ FOR AN A MERICAN CURRENCY CALL
OPTION
put with base parameter values similar to those in section ??,
with a foreign rate r2 = 2.8%. These parameters are given in
Table III for completeness.
K S T σ r2 r10 r¯1 σr1
100 80 1y 40% 2.8000% 1% 4.18% 1.28%
TABLE III
I NPUT PARAMETERS FOR THE BASE EXAMPLE ON INTEREST
STOCHASTICITY FOR A CURRENCY PUT

Similar to equity puts, we determine the effective stopping

time τ ∗ for the American currency put using (6) and (7). The
results are presented in Figure 5 for different moneyness levels.
We obtain again similar results as in the case of an equity put Fig. 6. Effective sopping time τ ∗ versus moneyness for an American currency
option confirming the validity of Taleb’s heuristic in (6). call

S/K ρA ρE τ∗ = Ω τ ∗ = E[T̃ ]
0.8000 0.4744 0.4876 8.4816 7.6977 C. Different Rate Dynamics and Moneyness Levels
0.9000 0.4380 0.4477 9.4292 8.7379
1.0000 0.3904 0.3960 10.0379 9.4770 In this section, we evaluate the hidden optionality of Amer-
1.1000 0.3371 0.3404 10.4487 9.9653
1.2000 0.2841 0.2859 10.7387 10.3291 ican equity puts, currency puts, and currency calls, under
1.3000 0.2349 0.2359 10.9521 10.5744 different rate dynamics. Similar to section VI, we start by
1.4000 0.1915 0.19215 11.1036 10.7799 considering a normally distributed interest rate, and then ex-
TABLE IV
E FFECTIVE SOPPING TIME τ ∗ VERSUS MONEYNESS FOR A CURRENCY PUT tend the analysis to a lognormally and Hull-White distributed
OPTION . rate.
1) Equity Put: In this section, we consider the American
2) Currency Call: We study a currency call option with equity put with the same parameters as before. We begin
parameters similar to those of the currency put above but with by extending the normal interest rate framework through
a foreign rate r2 = 10%. The observations we make here are additional experiments, using two other fixed values of S/K,
also applicable to call options with dividends, which have a and we present the corresponding results in Figure 7. We
similar pricing structure.1 The results in Figure 6, which again observe that, the lower the moneyness levels, the greater is the
validates the heuristic (6). optionality. Then, we consider a stochastic rate that is Hulll-
1 Optionality in the context of non-dividend paying equity calls is not
White and lognormally distributed, with the results reported in
relevant as it is not optimal to exercise these options before maturity, e.g. Hull Figure 8. We observe a behavior similar to that under a normal
[10]. distribution, as the standard deviation σr increases, the level
 9

Fig. 7. Optionality versus standard deviation for an equity put under a

normally distributed local interest rate for different moneyness levels Fig. 9. Optionality versus standard deviation for currency put option under
different rate dynamics with S/K = 1.00

of hidden optionality becomes more pronounced. Again, the

results show consistent monotone optionality values, πA , as a
function of the interest rate volatility.

Fig. 10. Optionality versus standard deviation for a currency put under a
normally distributed local interest rate for different moneyness levels

Fig. 8. Optionality versus standard deviation for a an equity put option under optionality metric as a function of the rate volatility. Then,
different rate dynamics with S/K = 1.00
we consider three different moneyness levels and repeat the
same experiments. The results in Figure 12 reveal again that
2) Currency Put: In this section, we further study the
deep in the money currency calls (with high values of the
optionality of the American currency put, with the base
moneyness S/K) exhibit higher optionality.
parameters listed in Table III. For S/K = 1 and under a
local rate, r1 which is (i) normally, (ii) log-normally and (iii)
Hull-White distributed, the optionality metric πA is computed D. Optionality of American vs European
similar to the case equity puts in section VI. We vary again As an alternative measure of hidden optionality, we can
the standard deviation of the interest rate σr and evaluate the compute the stochasticized American option value, OA(r),
e
resulting πA , which increases with σr , as illustrated in Figure under a normal interest rate, and compare it with the corre-
9. Moreover, we compute the optionality measure πA for three sponding European price, estimated as
different moneyness levels and we present the results in Figure Z
10. The results confirm once again that the deeper the option OE(r)
e = OE(r)frT (r)dr,
is in the money, the greater is the optionality. Dr
3) Currency Call: Next, we explore currency call options where frT (r) is the density of the interest rate at the option
using similar parameters to those in Section VII-B2, and under maturity, T . We then estimate
various local rate distributions. The results are shown in and
(2)
Figure 11 confirm a consistent monotonic behavior of the πA = OA(r)
e − OE(r),
e
 10

(2)
Fig. 13. πA versus moneyness for equity put options under a normally
distributed interest rate

Fig. 11. Optionality versus standard deviation for a currency put option under
different rate dynamics with S/K = 1.00

(2)
Fig. 14. πA versus moneyness for currency put options under a normally
distributed interest rate

Fig. 12. Optionality versus standard deviation for a currency call under a
normally distributed local interest rate for different moneyness levels

as a measure of the gain from the hidden optionality of the

(2) (2)
American option. We compute πA for various moneyness Fig. 15. πA versus moneyness for currency call options under a normally
distributed local interest rate
levels for (i) an equity put, (ii) a currency put, and (iii) a
currency call. We present the results in Figures 13-15. The
(2)
results show again consistent positive πA values, indicating see Eq. 3 as π∆a could be immediately computed by moving
that the American option is less vulnerable to interest rate either rate.
stochasticity than the European one. In addition, we observe
that the lower the moneyness level, the greater is the option-
R EFERENCES
ality (and implicitly the robustness) of the American option
over the European one. Such results have been observed, [1] Taleb, N.N. and Douady, R. (2013). Mathematical definition, mapping,
and detection of (anti) [Link] Finance.
sporadically, in the literature, e.g., [18] and [12].
[2] Taleb, N.N., Canetti, E., Kinda, T., Loukoianova, E., and Schmieder, C.
(2018). A new heuristic measure of fragility and tail risks: application
VIII. D ISCUSSION to stress testing. International Monetary Fund.
[3] Bachelier L. Théorie de la spéculation. Ann Sci Éc Norm Supér.
This paper uncovered significant hidden convexity in Amer- 1900;17:21–86.
ican options and presented a technique for pricing and un- [4] Dupire B. Pricing with a smile. Risk. 1994;7(1):18–20.
covering hidden risks so far ignored in the literature. While [5] Dupire B. A unified theory of volatility. In: Derivatives pricing and
credit exposures. London: Risk Publications; 1997.
our approach was near-exhaustive, by focusing on one single [6] Derman E, Kani I. The volatility smile and its implied tree. Quantitative
interest rate, even simpler techniques can be applied on the fly, Strategies Research Notes. New York (NY): Goldman Sachs; 1994.
 11

[7] Hull J, White A. The pricing of options on assets with stochastic

volatilities. J Finance. 1987;42(2):281–300.
[8] Gatheral J. The volatility surface: a practitioner’s guide. Hoboken (NJ):
John Wiley & Sons; 2006.
[9] Taleb N. N. Dynamic hedging: managing vanilla and exotic options.
New York (NY): John Wiley & Sons; 1997.
[10] Hull JC. Options, futures, and other derivatives. 10th ed. Harlow:
Pearson; 2017.
[11] Garman MB, Kohlhagen SW. Foreign currency option values. J Int
Money Finance. 1983;2(3):231–7.
[12] Medvedev A, Scaillet O. Pricing American options under stochastic
volatility and stochastic interest rates. J Financ Econ. 2010;98(1):145–
68.
[13] Stoer J, Bulirsch R. Introduction to numerical analysis. 3rd ed. New
York (NY): Springer Science & Business Media; 2013 Mar 9. Chapter
3, Topics in integration.
[14] FRED, Federal Reserve Bank of St. Louis, 2025. Market
yield on U.S. Treasury securities at 1-year constant matu-
rity, quoted on an investment basis [DGS1]. Available at:
[Link]
[15] Hassan NE, Maddah B. Power approximation for pricing American
options, 2026. Int Trans Oper Res. 2024 Jan;33(1), :117–1-42.
[16] Cox, J. C., Ross, S. A. and Rubinstein, M., 1979. Option pricing: a
simplified approach. Journal of Financial Economics, 7(3), pp.229–
263.
[17] Black, F. and Scholes, M., 1973. The pricing of options and corporate
liabilities. Journal of Political Economy, 81(3), pp.637–654.
[18] Garman, M. B. and Kohlhagen, S. W., 1983. Foreign currency option
values. Journal of International Money and Finance, 2(3), pp.231–237.