Complex Menus in Discriminatory Pricing

summary

The gist

Complex Menus in Discriminatory Pricing shows that optimal menus are generically infinite in multi-dimensional screening problems when complexity costs are present but small The results suggest that…

In short

The study investigates optimal menus in multi-dimensional screening problems where complexity costs exist. It finds that optimal menus are generically infinite unless complexity costs are very large. This suggests that simple pricing schemes might be driven by high complexity costs, and technological progress lowering these costs could lead to more product variety, like in cloud computing.

Key concepts

Optimal Menus
These are the best possible sets of options a seller can offer to maximize revenue when selling different products to different types of buyers. The paper shows that under certain conditions, the best menu size is often infinitely large.
Complexity Costs
These are costs associated with creating and managing a menu of options. Crucially, these costs decrease as the menu gets bigger but grow without bound as they get smaller. The paper links small complexity costs to infinitely large optimal menus.
Genericity
This refers to the idea that most scenarios or distributions of buyer types are 'generic' and not special exceptions. The theorem proves that for generic type distributions, every optimal pricing mechanism will result in an infinite menu size, making this a robust finding.

Terminology used across episodes

This episode discusses

The paper

Complex Menus in Discriminatory Pricing · Read on arXiv

ENS Paris-Saclay · Caltech · University of Zurich

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: I'm Ruth.

Warren: And I'm Warren, here with Iris.

Ruth: We're looking at the paper: "Complex Menus in Discriminatory Pricing".

Warren: Here's the gist of the paper. Complex Menus in Discriminatory Pricing shows that optimal menus are generically infinite in multi-dimensional screening problems when complexity costs are present but small The results suggest that cloud computing might well be a harbinger of things to come Core Theoretical Findings The paper establishes that optimal menus are generically infinite in generic multidimensional screening problems, even when efficient contracts are very simple This complexity arises from screening considerations alone, not from non-linearities in production costs or preferences The results extend beyond the canonical linear multi-dimensional screening model Specifically, for generic type distributions, revenue-maximizing menus cannot be achieved by simple screening contracts such as two-part tariffs because the necessity of infinite menus implies that optimal mechanisms must divide the market into infinitely many segments The Role of Complexity Costs Complexity costs are defined as costs that decrease the menu size, but grow without bound as these costs decrease The results show that optimal menus are generically infinite if complexity costs are small These results imply that simple pricing schemes observed in practice might be driven by relatively large complexity costs rather than the absence of them The paper notes that technological progress lowers these costs, which is expected to increase the extent to which sellers price discriminate through vertical differentiation Friction Effects on Menu Size The paper explores four reasons why adding new options might be costly or difficult, including setup costs, buyer navigation costs, technological constraints, and variable costs For generic type distributions, the optimal menu size grows without bound as these frictions vanish Specifically, menu-size costs are likely to fall due to technological progress in information technology Cognitive costs can also increase the optimal menu size, as a seller must leave every buyer who inspects the menu enough expected surplus to cover the increasing cognitive cost Genericity and Robustness Theorem 1 proves that for every k ∈ N, the set of type distributions for which every optimal mechanism has menu size greater than k is strongly generic This means that for generic type distributions, every optimal mechanism has infinite menu size The proof involves constructing a perturbation of an existing finite menu that yields strictly higher revenue under a signed density, demonstrating the open and dense nature of the set P>k Corollary 5 shows that for generic type distributions, the optimal menu size along any sequence of allocation spaces (Qn)n∈N tends to infinity as mesh(Qn) → 0 Implications for Practice

Title and authors: Ruth: The title itself, "Complex Menus in Discriminatory Pricing," sets up the tension between variety and cost. It immediately makes you wonder how far a company can push its product line before it becomes too complicated to manage or for buyers to deal with.

Warren: And the authors are tackling this by looking at multi-product or multi-attribute cases, which is what we see everywhere in real business today.

Iris: I’m thinking about how this relates to the other work we've seen, like those models where firms are making choices about course allocation or resource distribution through markets. Does this paper apply those same screening principles?

Ruth: It does, but it shows that the need for complex menus isn't just about having many options; it’s a structural result of how buyers choose between different contracts or products.

Warren: And they build on previous examples, showing that even when you start looking at more complex bundles—where each item or attribute is either in or out—simple menus aren't enough to maximize revenue.

Iris: So the point isn't just that complex menus are sometimes useful for revenue maximization, but that the need for them is generic across different types of buyer populations.

Ruth: Precisely. The paper argues that finding these infinite menus isn't an exception; it’s the norm in many screening environments when you account for complexity costs.

Warren: And the authors show this holds even when you introduce those complexity costs, as long as they are small enough.

The paper's summary: Ruth: So, let’s look at what the paper actually summarizes. It boils down to this: if complexity costs are small—meaning it’s cheap for buyers and sellers to deal with more variety—then optimal menus will just keep getting larger and larger indefinitely.

Warren: They contrast this with the classic results from Stokey, Myerson, and Riley and Zeckhauser, which in single-dimensional screening models suggest selling just one variety is always optimal.

Iris: That’s a big jump from their intuition. I wonder what variation is driving this divergence—is it really the multi-dimensional nature of the screening that matters here?

Ruth: It is. The paper shows that the results aren't dependent on non-linearities in production costs or preferences; this happens purely because of how buyers screen themselves based on these attributes.

Warren: And they demonstrate this by constructing examples where simple bundles—where each good or attribute is just included or excluded—don't lead to the maximum revenue.

Iris: So, even when you try to keep the menu structure simple, it fails to capture the optimal strategy in these multi-dimensional settings.

Ruth: Right. And they show that for generic buyer populations, the known instances where complex menus are needed for revenue maximization are actually just typical situations rather than rare exceptions.

Warren: This really changes how we think about optimal product lines when you move beyond simple one-to-one screening scenarios.

The paper's improvements: Ruth: Now, the authors suggest a few ways this research can be applied practically or theoretically. One big thing is modeling market response to technology by including complexity costs as a dynamic function.

Warren: That means if technology makes those complexity costs fall, we should expect the optimal menu size to decrease, but not necessarily end up at a fixed point. It grows without bound as those costs shrink.

Iris: That gives us something concrete for forecasting product variety growth in areas like cloud computing, which is exactly where that technological progress is happening.

Ruth: Right. And another improvement they suggest is using AI agents to navigate these large menus more efficiently, because AI tools can actually reduce the buyer-side costs of evaluating complicated options.

Warren: I agree on the AI part. If buyers can use better tools, they might be willing to deal with a bigger menu than we currently think they are.

Iris: That leads me to another point: system design should dynamically adjust menu complexity based on buyer cognitive costs, because the paper shows that those cognitive costs actually increase the optimal menu size.

Ruth: So sellers can manage this by using pricing or by offering tailored subsets of buyers, essentially trying to control how much complexity a specific group of people deals with.

Warren: And for digital goods applications specifically, they suggest training AI models on the linear monopoly model where extensive product differentiation is just a result of monopoly power, not necessarily technological progress.

Conclusion: Ruth: So to wrap up this look at "Complex Menus in Discriminatory Pricing," the core implication is that we need to stop assuming simple menus are sufficient when screening involves multiple attributes and low complexity costs.

Warren: The paper shows that for generic type distributions, the optimal menu size tends toward infinity as you reduce those complexity costs, which is a huge statement about market structure.

Iris: I just want to add that this result is robust because it doesn't rely on non-linearities in production or preferences; it stems purely from the screening mechanism itself.

Ruth: That’s right. And they show this holds for every k, meaning no matter how big you set a threshold for the menu size, there are always distributions that require an even bigger one.

Warren: It suggests that pricing algorithms in dynamic environments should be optimized to balance revenue maximization against those menu-size costs because small menu-size costs generically lead to large optimal menus.

Iris: I think the biggest practical point is that we need to model how falling complexity costs translate into future variety growth predictions for sectors like cloud computing.

Ruth: That’s it for this paper, "Complex Menus in Discriminatory Pricing." We’ll be looking at other interesting work next week.

More episodes

← Home