Abstract: Fixed points of the mean residual life (MRL) function are related to solutions of revenue maximization problems under stochastic demand. To derive economic interpretable conditions under which such fixed points exist and are unique, we study the decreasing generalized mean residual life (DGMRL) property. DGMRL distributions are closely related to increasing generalized failure rate (IGFR) distributions, which are extensively used in economic applications. As for IGFR distributions, we show that the moments of DGMRL distributions with unbounded support are linked to their limiting behavior at infinity. We derive closure properties of the DGMRL class that are important in economic applications and illustrate our results with numerical examples. While motivated by a specific stochastic revenue maximization problem, the present study contains results that are of independent interest.
Introduction
In the field of monopolistic pricing, a central objective for sellers is to identify the price $p$ that maximizes expected revenue, defined as $R(p) = p \mathbb{E}[D(p|\alpha)]$. Here, demand $D$ is contingent on a random variable $\alpha$, representing the buyer's type or the overall demand level. For this optimization problem to be tractable, the revenue function must typically be unimodal—possessing a single, well-defined peak.
Historically, economists and operations researchers have relied on the Increasing Generalized Failure Rate (IGFR) property to guarantee this uniqueness. While powerful, the IGFR condition is mathematically restrictive: it generally requires the demand distribution to be defined over a single, connected interval. In practical market scenarios, however, demand often manifests in distinct "regimes" or scenarios, such as a high-probability baseline demand coupled with a low-probability, high-impact "spike." Such distributions have disjoint support, which the IGFR framework cannot easily accommodate. This paper addresses this gap by introducing the Decreasing Generalized Mean Residual Demand (DGMRD) class.
The Fixed-Point Characterization of Optimal Prices
The authors utilize an additive linear demand model, where demand is expressed as $D(p|\alpha) = (\alpha - p)_+$. By drawing on reliability theory, they focus on the Mean Residual Demand (MRD) function, $m(p)$, which represents the expected additional demand given that the price has already reached threshold $p$.
A pivotal insight of the paper is that the seller's first-order condition for revenue maximization simplifies to a remarkably elegant fixed-point equation:
$$p = m(p)$$
This equation suggests that at the optimal price, the price itself equals the expected "leftover" demand. This provides a clear heuristic for price adjustment: if the current price is less than the expected residual demand ($m(p) > p$), the seller should increase the price; conversely, if $m(p) < p$, the price should be lowered.
Generalizing Reliability Classes
To ensure the existence and uniqueness of this fixed point, the authors define the Generalized Mean Residual Demand (GMRD) function as $\ell(p) = m(p)/p$. The DGMRD class consists of distributions where this GMRD function is non-increasing.
One of the most significant theoretical results is the mapping of DGMRD against existing reliability classes. As illustrated in the hierarchy below, DGMRD serves as a "bigger tent" that encompasses both IGFR and Decreasing Mean Residual Demand (DMRD) distributions.
Crucially, unlike IGFR, the DGMRD property can handle distributions with "gaps" in their support. This allows economists to model markets with distinct regimes—such as a modal demand level versus an extreme "black swan" event—without sacrificing the ability to find a unique optimal price.
Numerical Illustrations and Closure Properties
The paper provides several numerical examples to demonstrate how DGMRD behaves in scenarios where IGFR might fail. For instance, the Birnbaum-Saunders distribution is analyzed to show how its GMRD properties ensure a well-behaved revenue function even when traditional failure rate measures might be complex.
Beyond specific distributions, the authors establish that the DGMRD class is robust under common economic transformations. It is "closed" under positive scaling (changing currency units) and left-truncation (setting a minimum demand threshold). Furthermore, the paper links the finiteness of a distribution's moments (such as variance) directly to the limiting behavior of the GMRD function as the price approaches infinity, providing a bridge between statistical properties and economic outcomes.
Conclusion
By shifting the analytical focus from hazard rates to mean residual demand, this research offers a more flexible framework for stochastic revenue maximization. The fixed-point characterization $p = m(p)$ provides both a mathematical tool and an intuitive economic guideline for optimal pricing. While the current study focuses on linear demand models, the introduction of the DGMRD class paves the way for future research into more complex, non-linear demand structures and the behavior of mixture distributions in diverse market environments.
Summary updated