Pricing Patterns in Online Fashion Marketplaces: Evidence from Amazon and Temu

This thesis empirically examines pricing patterns across two online marketplaces operating under fundamentally different business models: Amazon and Temu. While existing literature has extensively documented price dispersion in digital markets and the role of platform architecture in shaping pricing outcomes, no study to date has directly compared product-level price distributions between these two platforms within a controlled product category. This study addresses that gap through a cross-sectional analysis of 550 product observations — 275 per platform — collected within the women's tank tops category on Amazon.it and Temu.com, restricted to a price range of €1 to €35. Descriptive analysis reveals a raw price gap of €11.63, with Amazon products exhibiting a mean price of €17.51 compared to €5.88 on Temu. Amazon prices are substantially more dispersed, with a standard deviation of €6.14 against Temu's highly compressed distribution of €1.30. Additionally, 70.5% of Amazon listings employ psychological pricing endings compared to only 5.1% on Temu, consistent with the behavioral pricing literature. The empirical strategy employs two OLS hedonic pricing specifications using the natural log of price as the dependent variable. The first specification controls for product rating and standardized review volume; the second excludes review volume to assess sensitivity to the potential endogeneity of reviews. Across both specifications, the platform coefficient β₁ is negative, statistically significant at the 1% level, and economically large. After controlling for observable product characteristics, Temu products are associated with prices approximately 69.8% to 70.1% lower than comparable Amazon products. The stability of β₁ across both specifications — with a difference of only 0.0106 log points — suggests the estimated platform price differential is not substantially driven by the endogenous relationship between review accumulation and unobserved product quality. A median regression robustness check confirms the finding is not driven by extreme price observations, yielding a consistent platform coefficient of −1.1478 (−68.3%). These findings are interpreted as conditional associations rather than causal effects, as the cross-sectional design does not permit causal identification. The results are consistent with the theoretical expectation that Temu's Consumer-to-Manufacturer supply chain model eliminates intermediary markups, while Amazon's platform architecture — characterized by seller fee absorption, Buy Box algorithmic enforcement, and behavioral pricing mechanisms — sustains a structurally elevated price level. This thesis contributes the first product-level empirical comparison of pricing between Amazon and Temu, situates this comparison within a cross-border e-commerce context, and engages seriously with the methodological challenges of endogeneity and causal identification inherent in observational marketplace data.