Unbiased Platform-Level Causal Estimation for Search Systems: A Competitive Isolation PSM-DID Framework

2025-11-03

AI Summary

In search-based two-sided markets, causal evaluation of platform-level interventions is often biased due to cross-regional spillovers and network interference among users and merchants. To address this, we propose the Competing-Isolation PSM-DID framework: building upon propensity score matching (PSM) and difference-in-differences (DID), it introduces a competing-isolation mechanism that enforces mutual exclusivity among treatment units, thereby theoretically eliminating spillover-induced estimation bias. The method enables unbiased identification of causal effects on core metrics—including order volume and GMV—while jointly controlling for selection bias and inter-unit interference. We validate the framework on a large-scale real-world marketplace, demonstrating substantial reductions in both interference effects and estimation variance. Additionally, we release an open-source dataset to support reproducible research. Our key contribution is the first integration of competing isolation into the PSM-DID paradigm, yielding a rigorous and practical causal inference solution for platform interventions under network externalities.

Technology Category

• Machine Learning: Causal Learning
• Multiagent Systems: Mechanism Design
• Reasoning under Uncertainty: Causality

Application Category

• Economics, Online Markets and Human Computation: Economics and fairness of platforms and recommendation systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Security and Privacy: Large-scale security measurements

Abstract

Evaluating platform-level interventions in search-based two-sided marketplaces is fundamentally challenged by systemic effects such as spillovers and network interference. While widely used for causal inference, the PSM (Propensity Score Matching) - DID (Difference-in-Differences) framework remains susceptible to selection bias and cross-unit interference from unaccounted spillovers. In this paper, we introduced Competitive Isolation PSM-DID, a novel causal framework that integrates propensity score matching with competitive isolation to enable platform-level effect measurement (e.g., order volume, GMV) instead of item-level metrics in search systems. Our approach provides theoretically guaranteed unbiased estimation under mutual exclusion conditions, with an open dataset released to support reproducible research on marketplace interference (github.com/xxxx). Extensive experiments demonstrate significant reductions in interference effects and estimation variance compared to baseline methods. Successful deployment in a large-scale marketplace confirms the framework's practical utility for platform-level causal inference.

Problem

• Evaluating platform-level interventions in two-sided marketplaces with systemic effects
• Addressing selection bias and cross-unit interference in causal inference
• Providing unbiased estimation for platform-level metrics instead of item-level

Innovation

• Integrates propensity score matching with competitive isolation
• Provides unbiased estimation under mutual exclusion conditions
• Reduces interference effects and estimation variance significantly