ONLINE RETAIL: Evidence of firm collusion to raise prices

E-commerce is not as competitive as we would expect, according to research by Hampus Poppius, which links increasing interactions between retailers to price increases.

When firms compete in multiple product markets, the multiplicity of their contact can help them collude on higher prices. This is profitable for the firms but hurts consumers, which is why competition authorities by different means impede firms to collude. 

Economists have expected that e-commerce ought to be very competitive, resulting in low prices and other benefits to consumers. The new study shows that this might not be the case.

With data on all prices for consumer products sold online in Sweden, the author shows that prices increase when the retailers have contact in more product markets. An increase in the number of contact markets that represents the difference between two randomly picked retailers leads to a 7.5% price increase.

This means that online retail is not as competitive as economists once hoped, and that competition authorities should safeguard competition in this increasingly large industry with great care.

Previous research has found the same relationship between multimarket contact and higher prices in, for example, airline and banking. Those industries are typically very concentrated and expected to be able to avoid fierce price competition.

The online retail industry is quite different and generally assumed to be more competitive. Despite that, the effect found here is larger than most estimates in previous research.

The study uses a vast amount of data collected using a technique called web scraping. The author collected information about all the retailers and all the products they sell that are listed on the market leading price comparison website in Sweden. The data were collected at two points one year apart to analyse changes over time in relevant characteristics and prices.

The data include roughly 3,300 different retailers and about 900,000 different products from 950 separate product categories. The volume of data makes it possible to measure carefully in how many product markets all pairs of retailers interact, and how the number of contacts vary over time.

The time dimension makes it possible to account for a wide range of factors that could jeopardize the interpretation. In addition, because the data are so rich, the author can leverage new empirical methods that build on machine learning to estimate the effect of multimarket contact on prices for each retailer and product individually. It is thus possible to see more clearly when the retailers are able to facilitate collusion.

It turns out that only when the contact markets contain relatively few competitors can they facilitate collusion. This is to be expected by the theory about multimarket contact and collusion, which further bolsters the interpretation of these results.

The results are alarming because they show that even in a seemingly competitive industry do firms manage to soften competition and maintain high profits. The e-commerce sector is not spared from such traits, and this is not the only study to raise this alarm.

Relatedly, competition economists have recently paid more attention to measures taken by manufacturers to soften competition between their retailers. With the same data as use here, the authors separately studies how the manufacturers’ importance to the retailers induce the retailers to set the same prices.

These two studies together show that online retailing is not as competitive as we would like, which calls for more attention from the competition authorities.


Hampus Poppius /