A statistical analysis of product prices in online markets

TL;DR

This paper analyzes online product prices, revealing they behave like random walks with higher probabilities of price changes after multiple events, but lack long memory and follow an exponential distribution, contrasting with asset prices.

Contribution

It provides the first empirical analysis of product price fluctuations in online markets, highlighting key differences from asset price behaviors and suggesting reasons for these differences.

Findings

Product prices behave like random walks.

Price change distribution is close to exponential.

No long memory property in volatility.

Abstract

We empirically investigate fluctuations in product prices in online markets by using a tick-by-tick price data collected from a Japanese price comparison site, and find some similarities and differences between product and asset prices. The average price of a product across e-retailers behaves almost like a random walk, although the probability of price increase/decrease is higher conditional on the multiple events of price increase/decrease. This is quite similar to the property reported by previous studies about asset prices. However, we fail to find a long memory property in the volatility of product price changes. Also, we find that the price change distribution for product prices is close to an exponential distribution, rather than a power law distribution. These two findings are in a sharp contrast with the previous results regarding asset prices. We propose an interpretation that these differences may stem from the absence of speculative activities in product markets; namely, e-retailers seldom repeat buy and sell of a product, unlike traders in asset markets.