Good Data and Bad Data: The Welfare Effects of Price Discrimination

2025-02-05

ACM Conference on Economics and Computation

AI Summary

This paper examines the welfare implications of monopolistic price discrimination enabled by consumer data. We develop a theoretical model featuring endogenous market segmentation under residual uncertainty and adopt a weighted total surplus—i.e., a convex combination of consumer and producer surplus—as the welfare metric. We introduce the novel concept of “surplus monotonicity” to characterize the monotonic effect of information refinement on welfare, and derive a general necessary and sufficient condition that reduces welfare evaluation across multiple demand curves to a closed-form test involving only two basis demand curves. Rigorously delineating the boundaries under which information is universally beneficial or harmful, we establish that data collection enhances social welfare if and only if demand satisfies three conditions: domain overlap, two-basis separability, and hyperbolic monotonicity. Our results provide an operationally tractable theoretical benchmark for data regulation.

Technology Category

• Game Theory and Economic Paradigms: Imperfect Information
• Reasoning under Uncertainty: Decision/Utility Theory
• Search and Optimization: Other Foundations of Search & Optimization

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Data transparency and provenance
• Responsible Web: Ethical and legal aspects of web-scale data analysis, uses and collection practices

Abstract

The rise of big data technologies, allowing firms to collect detailed consumer data to estimate their willingness to pay, has reignited the longstanding debate on the welfare implications of price discrimination. A significant difficulty in regulating data collection practices is that it is close to impossible to perfectly monitor and control how firms use consumer data, which makes highly targeted regulation impractical. Often the relevant question is if data collection should be permitted, without knowing how much information the firm already has nor how much additional information it might be able to collect. To answer this question, we develop a model to study endogenous market segmentation by a monopolist subject to residual uncertainty. There is a given set of consumer types each represented by a downward-sloping demand curve specifying the distribution of consumers' valuations of that type. The seller has access to some information structure that maps types to signal realizations, allowing her to segment the market and charge a profit-maximizing price for each segment. Types in our setting represent everything that is possibly knowable, e.g., a complete profile of consumer characteristics, and a segmentation reflects what the seller actually knows, e.g., perhaps only consumer locations. Within this framework, we examine how additional information impacts welfare through a pair of opposing properties. Information is "monotonically bad" if every refinement of any segmentation reduces weighted surplus — a convex combination of consumer and producer surplus. Information is "monotonically good" if weighted surplus is higher for any refinement. We refer to these properties as surplus-monotonicity properties. Our main result characterizes each surplus-monotonicity property and has two parts. The first part of the result reduces the problem, for an unrestricted set of demand curves, to one with only two demand curves. It says that surplus-monotonicity holds if and only if three conditions are satisfied. First, the demand curves cannot be too far apart in the sense that the optimal monopoly price of each of them is in the interior of any other's domain of prices. Second, the set of all demand curves is decomposable into at most two basis demand curves. Third, the two basis demand curves satisfy the surplus-monotonicity condition themselves. Given the reduction in the first part of the result, the second part then identifies a closed-form expression that characterizes when a given pair of demand curves satisfies surplus-monotonicity. The full version of the paper can be accessed here: https://nimahaghpanah.com/pdfs/gooddata.pdf

Problem

• Studying welfare effects of monopolist price discrimination using consumer data
• Developing market segmentation model with residual uncertainty characterization
• Analyzing monotonic and non-monotonic welfare impacts of data usage

Innovation

• Model of market segmentation with residual uncertainty
• Characterization using two demand curves
• Bounds on welfare effects of information