What is the memoryless property of the Exponential distribution, and why does it matter?
An Exponential random variable satisfies P(X > s+t | X > s) = P(X > t): the remaining waiting time has the same distribution regardless of how long you have already waited. This is the only continuous distribution with this property, and it directly corresponds to a constant hazard rate — neither ageing nor improving over time.
How to think about it
The memoryless property is both the key strength and key limitation of the Exponential distribution. Understand it well before applying it to real-world waiting-time data.
Definition and CDF
X ~ Exponential(λ) has PDF f(x) = λ·e^(-λx) for x ≥ 0.
E[X] = 1/λ Var(X) = 1/λ²
P(X > t) = e^(-λt)The memoryless property — proof
P(X > s+t | X > s) = P(X > s+t) / P(X > s)
= e^(-λ(s+t)) / e^(-λs)
= e^(-λt)
= P(X > t)The elapsed time s vanishes entirely. The system has no memory of its past.
Worked example
A server processes requests at a rate of λ = 0.5 per second, so inter-arrival times follow Exp(0.5). A request arrived 4 seconds ago and another has not yet arrived.
What is the probability it arrives in the next 2 seconds?
P(X ≤ 2) = 1 - e^(-0.5×2) = 1 - e^(-1) ≈ 0.632The fact that 4 seconds have already elapsed is irrelevant — the answer is identical to asking about the first 2 seconds of any interval.
When memorylessness is appropriate vs not
Appropriate: Poisson process arrivals (network packets, call center calls), radioactive decay. The event mechanism genuinely has no wear or memory.
Inappropriate: machine failure that worsens with age (use Weibull with shape k > 1), human mortality (use Gompertz), customer churn after a honeymoon period (use mixture models).
The hazard rate h(t) = f(t) / P(X > t) = λ is constant for Exponential. A non-constant hazard rate is a direct signal that Exponential is wrong.