This paper attempts to shed light on the pricing behaviour of firms or sellers in market places with price competition. In particular, it aims at testing the mechanism that lies at the heart of one of the most famous models in economics, namely the Bertrand model of competition. This model suggests that firms' profits will be zero as soon as there are at least two sellers in the market. The model is built on the idea that consumers will go to the cheapest shop, even if the differences in prices are "infinitesimal". As a result, firms are tempted to undercut their opponents' price by a small amount and they do so over and over until they end up charging the cost at which they produce their products. However, the underlying assumption is at odds with the observation that consumers typically do not care very much about a single cent and the predicted outcome clearly goes against the fact that many prices end in 99c. If sellers really charged the price at which they buy or produce a good or a service, we should observe all possible price endings. Basu (2006) argues that the prevalence of 99 cent prices in shops can be explained by rational consumers who disregard the rightmost digits of the price. This "bounded rational" behaviour leads to an Bertrand equilibrium with positive mark-ups allowing firms to make at least small profits. We use data from an Austrian price comparison site, which bears the advantage that in such an environment other factors than price both can be observed online and play a minor role. Price comparison sites are therefore highly suited for our test. We find results highly compatible with Basu's theory. We can show that price points - in particular prices ending in 9 - are more frequent than other endings and have significant impact on consumer demand. This shows that the mechanism that was postulated by Bertrand is not in place. Consumers do not necessarily choose the cheapest price and they even less do so when the Euro digits of the price are the same (i.e. when the difference is small). Moreover, the nine-ending prices are sticky: neither the price-setter itself wants to change them nor the rivals do underbid these prices, if they represent the cheapest price on the market. This might either show that shops seem to believe that it is no use to only slightly undercut a rival's price if this doesn't trigger a change in the Euro, or it is an indication that there might be some sort of silent agreement to not undercut a price ending in 9. Our findings shed new light on the behaviour of consumers and firms, and propose that theoretical investigations on competitive behaviour should consider "discrete state spaces", i.e. markets where the smallest plausible price change is by a Euro or at least by a full Cent
