Prices and price dispersion in internet markets : why is the internet market not as frictionless as expected? : an explanation by random search models

(1)

Bachelor Thesis Economics and Business – Economics

Faculty of Economics and Business

Prices and price dispersion in Internet markets

Why is the Internet market not as frictionless as expected? An explanation by

random search models.

Written by:

Nadia Vriend

Student number: 10411496

Supervised by:

Stephan Jagau

June 2016

(2)

- 2 -

Statement of Originality

This document is written by Student Nadia Vriend who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

- 3 -

1. Introduction

Over the last years, with the rise of the Internet, the way of how consumers gather information has changed. Consumers can go online and compare shops easily with each other. The rise of consumer information sites, from price-aggregation and comparison sites to product reviews has led to a large decrease in consumers’ costs of gathering information. This has implications for different market outcomes, and one of these outcomes is the effect on prices.

‘… industry titans such as Bill Gates, the boss of Microsoft, regale the world’s leaders with the promise of “friction-free capitalism” ’

The Economist, May 10, 1997

‘… the Internet offers the closest thing to a perfectly competitive market in the world today’

Thomas Friedman, 1999

In popular press as in the academic literature, people predicted that with the rise of the internet prices and price dispersion would decrease substantially and might in the extreme version lead to a perfect market. The availability of information about prices and products in combination with low costs of search would lead to a frictionless environment similar to the one that assumed in traditional economic models (Baye, Morgan, & Scholten, 2004). As a result, prices would fall till marginal costs and price dispersion would disappear altogether.

A number of years and studies later a drop in prices for some products has been observed, but prices in internet markets are generally well above marginal costs and there are different ranges of prices for even homogenous products, so neither has price dispersion disappeared in these markets. In most empirical studies, prices in Internet markets are lower compared with traditional markets or find that prices declined during the rise of the Internet. But this is not an unambiguous finding, several other studies observed higher prices in Internet markets compared with traditional markets or an increase in prices during the rise of the Internet. Moreover empirical studies find a substantial amount of price dispersion in Internet markets, and most of the studies did even find a higher price dispersion compared with conventional outlets. Not all studies compared price

dispersion with conventional outlets, a few found lower price dispersion in Internet markets but they all conclude that the Internet did not bring the near-perfect market many had expected. Why is the Internet market not as frictionless as it was predicted to become in the early days of the Internet?

Recent research on Internet prices has devoted a lot of effort understanding this complex issue. Theoretical and empirical studies about Internet markets come up with numerous candidate explanations as to how imperfections persist in Internet markets. In this thesis, the point of departure is the generally accepted fact that consumer search costs are lower online than offline. Several random search models are taken into consideration to provide explanations for the

observations that these lower search costs in Internet market do not bring these as close to perfect as was expected before the rise of the internet. A short consideration is given to the assumption of homogeneous goods as well – an assumption that will be shared by many of the models we consider.

(5)

- 5 -

The rest of this study is structured as follows: In section 2 we review some of the empirical literature that examines Internet markets. As will emerge from the review, the consensus in the empirical literature is that Internet markets are far from frictionless. In the following sections, we survey candidate explanations for the stylized fact from section 2, most notably the higher than expected observed price levels and the high price dispersion in Internet markets. Section 3 gives an explanation of why homogeneous goods can behave like heterogeneous goods in an Internet market and how this potentially influences prices and price dispersion. Section 4 and 5 -- the main part of the thesis -- examines random search models that help explain the observed frictions in Internet markets. In section 6 a short note is given to obfuscation strategy followed by a conclusion in section 7.

2. Empirical research in Internet Markets

After the explosive growth of Internet in the 2000’s, the Internet market became an attractive field to do research in. Many studies compared the efficiency of the Internet market to that of traditional markets. They all wanted to understand whether the Internet, this new retail platform, really provides increased efficiency as was commonly predicted before the rise of the internet. However empirical research in Internet markets did not find the expected outcomes. In this section some literature is reviewed and shows the different outcomes of empirical research.

The first empirical research on internet markets was done by Bailey (1998). He examines a data set which consists of more than 30.000 price observations from 52 different retailers for 337 distinct titles of books, compact discs and software. The observations were collected from February 1997 till January 1998. His data does not suggest that prices and price dispersion are any lower on the Internet market than on conventional markets. This finding, the author notes, is in contrast with his expectation that lower consumer search costs will perturb the current market and tend prices to move to a competitive equilibrium. However he also mentions that, at the time of his study, Internet commerce might not have been mature enough for prices to display convergence to general

equilibrium so that he is hesitant to draw conclusions as to whether internet markets would have less friction than conventional markets in the long run.

In a comparable study, Brynjolfsson and Smith (2000) examined prices for two

homogeneous products, books and CDs. They used a data set with over 8.500 price observations collected in 1998 and 1999. They found that prices were 9-16% lower than prices in conventional outlets, depending on whether taxes and shipping costs were included or not. They find an average price spread of 33% for books and of 25% for CD which was even larger than the price dispersion in conventional markets. However, weighting these prices by proxies for market share, they find that price dispersion is lower in the Internet market than in conventional markets. They conclude that there are less frictions in Internet competition than in conventional outlets, but that the Internet market is still far from a perfect market. Clay et al (2002) find similar results in their study of the online book industry. They observe the same average prices in online and offline book markets but find a substantially larger amount of price dispersion for online markets. Ancarani and Shankar (2003) did the same kind of research in books and CDs in the Italian market during 2002. They find that although listed prices tend to be lower online, prices including shipping turn out to be higher online. With respect to price dispersion they report that whether price dispersion is higher or lower

(6)

- 6 -

online or offline depends on the measure used. Internet retailers display a higher price spread, but a lower standard deviation. But the overall conclusion is clear, online markets are in general

inefficient. While there are particularly many examples of empirical research in online book markets because of the homogeneous nature of the product, research in other online markets has been conducted and has tended to show price drops relative to conventional markets as well. For

instance, Morton et al. (2001) estimated that consumers using an online service to help them find a car to purchase, paid on average two percent less than other consumers, Sengupta and Wiggins (2014) found price reductions of 11 percent in airline tickets because of online sales and Brown and Goolsbee (2002) estimated that because of price comparison sites, website prices of term life insurance policies decreased by 8 to 15 percent.

While most of the empirical research found lower averages prices in Internet markets compared with conventional markets, there are also studies that find unambiguously higher prices in Internet markets. Erevelles et al. (2001) did research in another industry, the vitamin industry. They compared the prices of vitamins on the Internet with four other types of outlets: drug stores, discount retailers, warehouse retailers and supermarkets. They found out that the average price of vitamins was significantly higher on the Internet than for in the other market segments. In addition, price dispersion among Internet retailers was significantly higher than among traditional retailers.

In a research in S&P 500 index funds, Hortacsu and Syverson (2004) showed this

counterintuitive finding as well. They find that from 1992 until 2000, the number of mutual funds has been growing largely but at the same time fees have risen over time. The growth of highest costs funds fees was 30 times as great as the growth of lowest-cost funds fees. The highest-fee funds almost have the same market share as lowest-price funds, and as result, average prices increase as well in their dataset. All S&P 500 index funds have the same performance profiles, so differences because of the financial characteristics of these funds should be minimal.

3. Homogeneous goods?

The economic literature has traditionally focused on three explanations for price dispersion in markets: search costs, asymmetrically informed consumers, and product heterogeneity. We will focus on search costs in the next section, but before doing this we first examine the issue of product heterogeneity shortly.

In a perfect market, buyers and sellers have complete information about a product and it is easy to compare prices because all products are identical. The classic Bertrand Model of price competition suggests that undifferentiated retailers selling homogeneous goods in markets with perfectly informed buyers will charge prices equal to marginal cost (and, as a consequence, will not earn any profit) (Singh & Vives, 1984). One of the assumptions crucial to this result is the

homogeneous-goods assumption. As with all real-life markets, this assumption is violated for the Internet market. However, as we will see below, existing research suggests that product

heterogeneity takes some intriguing particular forms for internet markets that do not occur in traditional markets.

(7)

- 7 -

This statement sounds a bit weird in the first place. How is it possible to violate an assumption, which is just a fact about the good sold? There are heterogeneous goods and

homogeneous goods, and for a perfect market, one should look at homogeneous goods. But in the case of Internet markets, even homogeneous products behave like heterogeneous products. Brynjolfsson et al. (2010) suggest that while the underlying good is homogeneous, the final product is perceived as a differentiated product by the consumer. In this final product they include the bundled retailer services, because if a consumer buys a product on the Internet, services (direct or indirect) are automatically connected with the product.

To compare the Internet market with a perfect market, recent literature tried to eliminate the effects of heterogeneity a priori. These researches select products which are physically identical, as for example books or CDs. These kinds of products are totally identical and the marginal costs are the same as well across retailers (Brynjolfsson, Hu, & Smith, 2003). However this homogeneous product can be more than its purely physical characteristics. Aside from product heterogeneity there is retailer heterogeneity as well. A retailer that provides extra services might charge higher prices than his competitors. It is quite difficult, and might be even impossible, to eliminate these sources of heterogeneity out of a sample.

Brynjolfsson and Smith (2000) find that using service characteristics as explanation for the price dispersion is too easy. They observed in their data that services (return policies, title reviews ect.) offered by Internet retailers do either not significantly differ from each other or are negatively correlated with price, which means if retailers ask for higher prices, they offered less services. So observed services do not only explain price dispersion in their sample. Besides the visible services a retailer offers, there may be unobserved retailer characteristics as well. Brynjolfsson and Smith (2000) find evidence in their data that supports this prediction. They find that price differences between retailers do not significantly differ across titles. Price dispersion is not influenced by the product. Brynjolfsson and Smith (2000) suggest that price dispersion can be explained by

heterogeneity in (unobserved) retailer characteristics. Research of Chevalier & Goolsbee (2003) support this finding. In their study, Chevalier & Goolsbee (2003) use data on sales ranks of about 20.000 books from two leading online booksellers, Amazon.com and BarnesandNoble.com. Their result suggest price sensitivity at both retailers, but demand at BarnesantNoble.com was much more price-elastic than demand at Amazon.com. Amazon.com is the market leader with a high reputation, and because of this reputation Amazon.com is able to set higher prices without losing customers.

There are studies that have argued that trust is an important factor in online shopping (e.g. Hoffman et al, 1999; Pavlou and Gefen, 2004; Pavlou et al, 2007, Kim et al, 2012). Trust has always been an important factor in the buying process, but because of the impersonal and anonymous nature of the Internet market, the importance of trust is even higher for them (Pavlou and Gefen, 2004). Online sales often involve a delay between purchase and consumption when a product must be physically delivered. This means that shopping on the Internet comes together with the risk from the time lapse between purchase and delivery of the products. Consumers do not trust all websites enough to give them their personal information and money. So even homogeneous goods act like heterogeneous goods because these goods are usually bundled with ancillary services and the provision of these services might vary across sellers without being explicitly priced (Peitz &

Waldfogel, 2012). Sellers’ brand reputation might serve as proxy for the unobservable quality of such service provision.

(8)

- 8 -

4. Consumer search models

4.1 Search costs on the Internet

After the short discussion of product heterogeneity we will focus in this paragraph on another important assumption for a frictionless environment, namely perfect information. Under perfect information all consumers and producers are assumed to have full knowledge of posted prices, preferences of market participants, product quality etc. Another component of this assumption is that consumer search costs should be zero. The following section focus on search costs and their role in Internet markets.

Search costs are the costs which consumers face when searching for products or services. Search costs include money a consumer spends on searching, opportunity cost of time and energy spend on searching, travelling costs. Economic theory predicts that consumers weigh the costs and benefits of search when making decisions. A consumer will stop searching if the expected price reduction of the next search is smaller than the costs of search (Stigler, 1961). Search frictions can explain why the law of one price fails and why retailers may have pricing power even in a market with homogeneous goods.

In a Bertrand competition model, which explains a perfect market, retailers are selling homogeneous goods with unlimited supply. Without market frictions, all consumers buy their products from the retailer with the lowest price. This makes a pure price competition where all firms need to set the same price which will be equal to marginal cost (Singh & Vives, 1984). If firms have higher marginal cost than others do, firms with higher marginal costs will exit the market because they make negative profit. This Bertrand competition does not hold for our well-known physical markets because of high search costs (Bailey, 1998). Consumers are not able to know all prices which various retailers quote at any given time. A consumer (or retailer) who wants to know the best price must visit various sellers, a phenomenon Stigler (1961) started to call ‘search’.

With the introduction of the Internet, searching for products and prices became much easier for consumers. On the Internet consumers can search for almost anything they want and find information about products and services without leaving their home. Especially with the service of shopbots, searching and comparing products became more efficient. At a shopbot site consumers can easily compare products among different retailers. Consumers can search for a product and the price and other characteristics will be displayed in a tabular format. In the coming paragraphs we are going to analyze random search models and examine if they can explain the observed frictions in Internet markets.

(9)

- 9 -

4.2 Stigler (1961)

The seminal random-search model is due to Stigler (1961). He presents a first ‘search-theoretic’ explanation for price dispersion. Search-theoretic models assume that consumers pay costs for every additional price quote they are obtaining. The search process in Stigler’s model is non-sequential. I.e., consumers decide prior to search how many searches they are going to make. After searching, they buy the product at the retailer who offers the lowest price among the observed retailers.

Stigler (1961) shows that even in markets with homogeneous goods and symmetric retailers, price dispersion can arise in equilibrium because of search costs. In his framework, “price dispersion is a manifestation – and, indeed, it is the measure – of ignorance in the market” (p.214). In the model of Stigler with a uniform distribution of asking prices by retailers, the number of buyers from a retailer increases at an increasing rate as his price is reduced. Moreover, with the uniform

distribution of asking prices, the number of buyers from a retailer increases with increased search if the price is below the reciprocal of the amount of search. Thus we should generally expect the high-price sellers to be small-volume sellers. Stigler argued that with greater amounts of search the competition of retailers will eliminate the profitability of quoting very high selling prices and having very low buying prices at the same time. It will make some of the extreme price bids impossible. On this score, the greater the amounts of search, the smaller will be the dispersion of prices and the lower will be the average price.

Stigler’s theory does not fully explain the prices and price dispersion in the Internet market. Assumed that lower search costs leads to increased search, Stigler’s model (1961) predicts that the Internet market will have lower prices and lower price dispersion than traditional markets. While most of the empirical studies found lower average prices in the Internet market, they did not find evidence for lower price dispersion. The model of Stigler (1961) is not able to explain this finding.

4.3 Clearinghouse models

Stigler assumes homogeneous consumers, which are evenly informed. In his model consumers pay extra costs of search for every price quote they obtain and follow a non-sequential search process. In this paragraph a new structure is introduced, in clearinghouse models consumers are not evenly informed, there will be informed an uninformed consumers and.

In clearinghouse models an information clearinghouse provides consumers a list with prices charged by different retailers in the market (Baye, Morgan, & Scholten, 2006). In the Internet market a shopbot or a price search engine may serve as a clearinghouse. Some consumers look at these price listings and buy from the retailer with the lowest price, other consumers are uninformed and let themselves be guided by other criteria like the brand of a product.

Models along the lines of the clearinghouse include Salop and Stiglitz (1997), Rosenthal (1980) and Varian (1980). In a clearinghouse, all prices are listed and the consumer is fully informed about the prices. Consumers have different costs of accessing the information provided by the clearinghouse, and if the expected gain of accessing the clearinghouse is larger than the costs, consumers will access the clearinghouse. Some consumers will access, some consumers will not access and as result there will be informed and uninformed consumers. This lead to the fact that for

(10)

- 10 -

some consumers, obtaining an additional price quote is costly, while other consumers are aware of all prices charged in the relevant market.

In this environment some retailers shall ask a low price for their product to attract consumers shopping at the clearinghouse, while others will post a high price and are only able to serve uninformed consumers who do not access the clearing house. This is how the model predict price dispersion. If no consumers have access to the clearinghouse, the equilibrium outcome is the same as in the Diamond paradox1. All firms will charge the monopoly price. And in contrast if all consumers are informed, in equilibrium the outcome would be the same as in Bertrand competition, a perfect market where prices will equal marginal cost.

If we apply the clearinghouse model in the Internet market, a shopbot may serve as a

clearinghouse. In a shopbot all prices are listed and consumers will be fully aware of all prices. With a relatively easy access to a clearinghouse in the Internet market, we expect a large part of informed consumers. If consumers are better informed we would expect increase price competition, a more elastic demand and average prices on the Internet should decline. In addition price dispersion in the Internet market will be explained due to the fact that there are still consumers who face significant search costs online. Even if the amount of uninformed consumers will decrease a lot, there will stay uninformed consumers. There may be some retailers who wants to attract this (small) part of consumers and set high prices. These outcomes are in line with many empirical studies that also find lower average prices in the Internet market than in traditional markets but found a substantial amount of price dispersion as well.

But we have to make notes to this model as well. Assumptions made in clearinghouse models do, as proved by empirical research, not hold for the Internet market. One of the

assumptions in clearinghouse models is that informed consumers buy their (homogeneous) product at the lowest priced retailer (Varian, 1980). But in empirical research Brynjolfsson et al (2010) observed click troughs of consumers visiting a shopbot site. They found that about half of the consumers did not click on the offer with the lowest price. So this assumption made by the clearinghouse models do not hold in the Internet market.

Compared with Stigler’s model (1961), the Clearinghouse models provide better

explanations for the observation that lower search costs in Internet markets do not bring these as close to perfect as was expected before the rise of the Internet. Clearinghouse models are by introducing informed and uninformed consumers able to explain the founded price dispersion in Internet markets. But the model has some shortcoming as well, as informed consumers do not always buy at the lowest price store, average prices in Internet markets are not as low as predicted by clearinghouse models. Furthermore opposite findings, studies that find unambiguously higher prices in Internet markets cannot be explained by this model.

1_{Diamond’s (1970) finding that in the presence of search costs the unique equilibrium will be that all retailers}

charge the same price, the monopoly price is known as the Diamond paradox. If a consumer have to pay a search cost, in order to observe a seller’s price, a unilateral price increase by one seller will not affect the probability that the consumer will pay to observe the seller’s price. On the same time will a small increase in price not affect a consumer’s decision to pay another search cost to get an alternative price quote. So sellers have unilateral incentives to raise prices. These incentives remain till the price is on monopoly level. There will be no search in equilibrium since consumers rationally expect the same price at each firm and there won’t be a reason for search.

(11)

- 11 -

4.4 Moraga-González, Sándor and Wildenbeest (2015)

All traditional models of search costs described above, are based on the facts that consumers are imperfectly informed, information is costly to obtain and retailers set prices to leverage on consumer heterogeneity in information and search costs. Another important assumption of these models is that search cost are sufficiently low. Because of these low search costs, models assume that consumers search at least once. But what if this is not true? What if search costs are sufficiently high and some consumers do not search ones. There may be consumers who found it not worth to search for a satisfactory product. In this paragraph a new structure is introduced were the

assumption of sufficiently low consumer search costs is released.

Moraga-González, Sándor and Wildenbeest (2015) introduce two margins which are effected by an increase or decrease in consumer search costs, the intensive and extensive search margin. The intensive search margin is affected by search intensity. All traditional search models take only this component into account, if search costs are decreasing, consumers who are searching can search more intense for the same costs. The intensive search margin will increase and may result in a more elastic demand and will lead to lower prices. The extensive search margin is affected by

participation. This is the component which is new in comparing with the traditional models. This margin measures the participation of consumers that are searching for the product, so which part of the population is searching. If search costs decrease, this may lead to an increase of the extensive search margin because of an increase of the participation of people who search once or more. The increase in the extensive search margin makes demand more inelastic. The demand will be more inelastic because after the decrease in search costs, people who did not search before, start searching now. As result a decrease in search costs affect how many consumers are searching, and more important who chooses to search. There are more people who have higher search costs and as consequence search little. The demand will be more inelastic. When looking at this component the price will increase. This means when search costs decrease there are two effects which both gave a different outcome on the price level. A decrease in search costs may result in higher or lower prices depending on which margin has more relative impact.

Moraga-González, Sándor and Wildenbeest (2015) found out that if search cost densities have a decreasing likelihood ratio property, a first order stochastic dominance decrease in search costs will lead to a higher average price in equilibrium. The search cost density in this case is log-submodular implying that lower search costs will decrease the relative part of consumers with low search cost. This means that a decrease in search costs will have a bigger effect on the extensive search margin than on the intensive search margin, the demand will be more in-elastic and prices will increase. Conversely, when search cost densities have an increasing likelihood ratio property, a decrease in search costs will lead to lower prices. The search cost density is then log-supermodular the unique equilibrium will be that all retailers charge the same price, the monopoly price such that the lower search cost will increase the relative population of consumers with low search costs. In this case a decrease in search costs will have a bigger effect on the intensive search margin than on the extensive search margin such that the demand will be more elastic and prices will decrease.

(12)

- 12 -

Compared to the previous models, the model of Moraga-González et al (2015) provides an explanation for the possibilities of either a decrease or increase in prices after a decrease in consumer search costs. Where the previous models could only explain a decrease in prices, the model of Moraga-González et al (2015) gives an explanation for the outcomes of empirical research were prices in Internets were lower than in traditional markets as well for studies where prices in Internet markets were higher than in traditional markets or observed increased prices during the rise of the Internet. For example Hortacsu and Syverson (2004) presented that prices of investment fees increased during the timeframe where Internet banking and online brokerages decreased

transaction costs, and searching for new investment opportunities became much easier. This

increase may be explained by the fact that more people were going to participate in the mutual fund market for the first time. These people did not have experiences with the mutual fund market and had higher search costs. Because of the large part of new investors, the extensive search margin had more influence on the outcome on price than the intensive search margin. Demand became more inelastic and this explains the increase in fund fees.

5. Consumer search for heterogeneous products

In the last section, search models helped explain the observed frictions in Internet markets. In these pure search models product homogeneity was assumed, but as observed in section 3, in Internet markets even homogenous goods behave like heterogeneous goods. This cause shortcomings in the above models, as for instance the clearinghouse models assumes that informed consumers buy their products at the lowest price store, but empirical research showed that this was not the case in Internet markets. In the following section models for heterogeneous goods are examined.

5.1 Anderson and Renault (1999)

Anderson and Renault (1999) developed a classic model of price competition in the presence of search costs and product differentiation. They show how product differentiation resolves the Diamond Paradox. Anderson and Renault (1999) combine two general models for their theory. The general model of search costs (for homogeneous goods) combined with the general model for differentiated products. The general model for search costs, as the model of Stigler, provides a framework were the equilibrium prices increase if search costs increases and prices will fall if the number of firms increase.

In models of product differentiation (without consumer search), equilibrium prices increase with the degree of consumer taste for diversity (Anderson et al, 1992). The intuition behind this model is that a greater taste for diversity will lead to more market power for firms because of more intense preferences. The outcome of the product differentiation model will change with the introduction of consumer search costs. In the first place, when there are low preferences for diversity, the result will be the same as in the Diamond paradox, because consumers have no

incentive to search. As the preferences for diversity rise, there will be consumers who check multiple retailers if the first retailer they check is not what they like. This means that consumers have started to search, which brings retailers in competition and equilibrium prices will decrease. But if there will be even more intense preferences and this is high enough that a sufficient number of consumers will

(13)

- 13 -

start searching it will give a different outcome. The situation will be back to the model of product differentiation were consumers are perfectly informed and the equilibrium price will increase with the taste variety as in the models of product differentiation described above.

Brynjolfsson et al. (2010) concluded the following out the model of Anderson and Renault: ‘We can infer that for any given level of search costs, consumers should become less price sensitive and search more as their taste for variety increases. For any given level of variety, consumers should search more as their search costs decrease.’ (p.10).

With the model of Anderson and Renault a new explanation for friction in the Internet market is given. Comparing with the previous models, Anderson and Renault (1999) introduced that prices also depend on the degree of consumer taste for diversity. This model may give explanation for friction in the Internet market in the following way: With the rise of the Internet, search costs decreases and consumers were able to search more. But at the same time, as explained in section 3, trust in retailers became more important. The importance of trust in retailers may increase the taste for variety and make consumers less price sensitive. This gives as result that lower search costs should decrease prices, but because of increased taste for variety, prices will increase at the same time. Summing up this two factors together may give a proper explanation why prices did not decrease as much as expected and price dispersion did not disappear.

5.2 Bakos (1997)

By introducing the model of Bakos, the explanation of Anderson and Renault by increased degree of consumer taste for diversity will be extended. A framework from Bakos, where search cost for prices and search costs for product information are separated, is used to divide consumers in two different types. Consumers who are price sensitive and consumers who are not price sensitive and search for other product characteristics as for instance brand.

Bakos (1997) was the first one who modeled the role of buyer search costs and analyzed the impact of reducing this search costs in the context of an electronic marketplace. Bakos (1997) explained the electronic market place as an interorganizational information system that allows buyers and sellers to exchange information about prices and product offerings. He analyzed markets for homogeneous goods and concluded that a decrease in search costs may destabilize a

monopolistic equilibrium and the seller profits may disappear. But as stated before, homogeneous products act like heterogeneous products so we will focus on Bakos’ analysis on the role of search costs in market with heterogeneous goods.

Bakos (1997) introduced search costs in a market with heterogeneous products and consumer tastes by introducing search costs in a spatially differentiated market. He used the same ‘unit circle’ setting of Salop (1997) and introduced search costs in this model. In a market place with differentiated products, a buyer have to pay search costs to find product information and the price of the product offered by the retailer. After each search a consumer have to decide to search further or buy the product. Bakos (1997) showed that electronic market places reduced the search costs of buyers to obtain price and product information, which would slowly result in decreased price premiums and seller profit margins. Consumers who have access to the Internet become more demanding, and are willing to make fewer compromises concerning their ideal product. Buyers will be better off because of lower prices, being better informed and the ability to see more product offerings for the same price of search (Bakos, 1997).

(14)

- 14 -

In the conclusion above, the two types of search costs, costs for price and information have been bundled together. This is typical in search costs literature, for example the model of Anderson and Renault (1999) described above, assumed the same. This assumption is intuitive, if people go to a shop, or in our case search on the internet, they obtain price and information at the same time. But in another setting it may be possible to separate this two costs. Bakos (1997) give the example of phoning a store to obtain the price and visiting the store afterwards to obtain product information. But another possibility to use this model is for instance if people are not price sensitive, and the only costs they are facing are costs for searching for information. And of course the other way around, when consumers are more interested in prices and only pay costs for price searching.

Bakos (1997) modelled this by introduce three types of costs, access costs to visit a seller, costs to acquire price information and costs to acquire information about product characteristics. In this model he assumed that a buyer must incur access costs and after that the consumer will face search costs for price and/ or information. He concluded that lower search costs for price

information resulted in increased competition among retailers and lower prices. But in contrast with this finding, he showed that reducing search costs to obtain product information resulted in

increased monopoly power and profits for retailers. The intuition behind this finding is that if

product information is easy to obtain, and a consumer know which product fits him the best, further incentive to search is reduced. Each buyer becomes more captive of the retailer who offers the best fit because not purchasing from that retailer is certain to increase the buyer’s ‘fit costs’. Fit cost is a term which Bakos (1997) introduced to represent a consumers utility loss for buying a good different from their ideal one. Thus this analysis suggests that if consumers give more importance to product information instead of price information, retailers may maintain higher prices.

Brynjolfsson et al. (2010) found in their research that search intensity is not correlated with greater price sensitivity. They observed that consumers who search more, care relatively more about non-price factors like service characteristics, brand and reputation. Brynjolfsson et al. (2010)

analyzed the use of an Internet shopbot for books. They used data from a prominent online

comparison-shopping service (‘DealTime’) over a period of 12 months. After searching for a product, they could observe which actions a consumers took, did a consumer for instance click on lower screens by scrolling down, or did they re-sort the data by other characteristics as for example shipping time? They explored consumer heterogeneity and its implications for consumer behavior, based on observable behavior of consumers across different offers and screens.

When a consumer visits DealTime they choose a specific book offer and prices for different retailers will be calculated. Up to ten offers fit on a single screen, and the offers are ranked by total price (shipping and taxes are included) from lowest to highest. In their data they saw different ranges of consumer behavior. They saw consumers clicking on the first offer, and consumers who first click on multiple screens before deciding on which offer to click. This suggested heterogeneity in preferences and search costs. Brynjolfsson et al. (2010) introduced three kinds of consumers. First screen consumers that only inspect the first screen, low screen consumers who clicked at least once on offers in lower screens and sorting consumers who resorted the orders in another ordering than price. Brynjolfsson calculated price elasticities for the different kind of consumer types, and found difference elasticities across groups. First screen consumers present the highest price elasticities with a median between -6 and -14. In terms of Bakos’ model, this are consumers who are price sensitive and acquire search costs to obtain price information. Which are in this case very low, because of the use of the Shopbot. In contrast, low screen and sorting consumers had low price elasticities with a median less than one. These consumers are not very price sensitive, other product

(15)

- 15 -

characteristics like brand and service level play an important role to search for better options. In terms of Bakos’ model this are consumers who acquire search costs to obtain product information.

Because of a decrease in search costs consumers are able to search more for the price of products but for product and retailer information as well. Out of the observation of Brynjolfsson et al. (2010) and theory of Bakos, we may conclude that price dispersion in the Internet may arise because of different consumer types. For instance high reputational sellers or sellers with high service characteristics can raise their prices and still attract consumers; consumers who search for product/retailer information. Low reputational sellers have to decrease their prices to survive and will serve the other type of consumers, consumers who are searching for prices. As result there will be price dispersion in Internet markets and average prices do not necessary decrease.

Thus the outcome of empirical studies that prices in the Internet are not as low as expected and that some prices did even increase during the rise of the Internet is explained in a different way than Moraga-González et al. (2015) did. Where Moraga-Gonzalez et al. explained the fact by the ratio of the intensive and extensive margin, explanation based on the models of Anderson & Renault (1999) and Bakos (1997) is given by consumer preferences for other product characteristics than price. Where both aspects seem plausible, I would expect that the explanation by consumer preferences would play the biggest role. As mentioned in section 3, trust is an important factor in Internet markets and this is reflected in markets for almost every product or service sold on the Internet. The explanation by the intensive and extensive margin assert for a smaller part of the Internet market. For products as books and CDs this explanation is not plausible since people had already a relative easy access to these products in traditional markets. For this kind of products the extensive search margin cannot explain why prices are not that low as predicted. However, the explanation by extensive and intensive margin may give insights for markets on the Internet which were difficult to access in traditional markets, as the example given of the investment funds.

6. Obfuscation strategy

In section four and five, we examined explanations by random search models after a reduction in consumer search costs. But there are also studies that mention that ‘the effect of the Internet on search frictions is not so clear-cut’ (Ellison & Ellison, 2009). They found that advances in search technology because of the internet are accompanied by investments from retailers in obfuscation. Bakos (1997) already mentioned that it may be interesting for retailers to compensate for the lower search costs in Internet markets by making it increasingly difficult for buyers to compare the price of alternative product offerings.

Retailers have incentives to make it hard for consumers to compare product characteristics and prices. This is called obfuscation strategy. These practices can frustrate consumers’ price search, because they raises the cost of learning about each firm’s offerings and forces more consumers to conduct firm-by-firm searches, and as result it will reduce price competition. In addition another way of obfuscation are the sales of ‘add-ons’. The prices for these add-ons are high and can raise a retailer’s profit.

Ellison and Ellison (2009) examined an environment with a dominant role for price search, they observed small retailers selling computer parts through Pricewatch.com. Pricewatch.com is an

(16)

- 16 -

Internet price search engine which act like a clearinghouse. It compose a large number of small, minimally differentiated retailers who are selling computer parts. The retailers receive most of their customers through Pricewatch. If consumers search on Pricewatch they can search for a certain product and retailers will be listed sorted on price (from low to high). After clicking on one of these offers, the consumer will be leaded to the website of the retailer. Here the consumer will see the selected product, but the retailer is going to show more products as well. Retailers try to convince the consumers that another, more expensive product of higher quality, will be better than the product advertised on Pricewatch.com.

A consumer choose a retailer from the list, after clicking on a retailer, it will see more product options

All retailers advertise the low quality products, they have it in stock and sell it. But the low quality price is an advertising tool as well, it behaves like a loss-leader. Retailers try to convince the consumers that other, more expensive products of higher quality, are better than the product advertised on Pricewatch.com. Low quality products are dominating the sales, but there are

substantial sales for medium and high quality products that increase profits lots. Add-on pricing have been done by retailers forever. But before the Internet, this was very expensive, a sales men should convince you at shop, this men has to be well trained and it cost lots of time. Now on the Internet, the only thing retailers need is a website design.

Ellison and Ellison (2009) used data from Pricewatch and one particular firm who gave insight in all their data (about sales, prices, and costs of sales etcetera). If they only took the low quality products into account, they could say that the Internet brought the effects predicted in the early days of the Internet. The low quality products are facing extremely high elasticity of demand and as result very narrow profits. The market for low quality goods almost behaves like Bertrand suggested. But Ellison and Ellison (2009) found evidence that the low quality product is effectively used as an advertisement tool. They found out that controlling for a site’s medium and high quality prices and other variables, the site sells more medium and high quality products when it has a higher position on Pricewatch’s list. It is unlikely that the rank of the low quality good on Pricewatch is correlated with a rank for a retailers medium and high quality prices because they saw that ranks on Pricewatch changed frequently, where medium and high quality prices are left unchanged for a substantial period of time. The markups for low quality goods are sold for a slightly loss or a small profit, but the markups for medium and high quality goods are sufficiently higher and make up for the loss made by low quality products. Elasticity for medium and high quality products are

significantly lower than for low quality products. This presents the fact that it is more difficult to search for them and give evidence for search frictions on the Internet.

(17)

- 17 -

With the outcomes of this research, evidence is found that only looking at the consumer search side of the Internet market may not give us a complete story. There is a balance of power, because of the Internet search became easier for consumers, but obfuscation became easier and cheaper for retailers as well. There are lots of other markets were we see this, or other obfuscation strategies for example in the Airline Industry, were add-on pricing is very well-known concept as well.

7. Conclusion

In the early days of the Internet people speculated about a perfect market. The Internet would reduce search costs, and in popular press as in the academic literature people predicted that prices and price dispersion would decrease substantially and might in the extreme version lead to a perfect market. Out of empirical research we concluded that a perfect market is far away, and the Internet market is not as frictionless as expected.

Several random search models gave multiple insights and provided explanations for the friction in Internet markets. The theory of Stigler (1961) predicted lower prices and price dispersion after a decrease in search costs because of increased price competition. Most empirical research found lower prices in Internet markets compared with conventional markets, but a decrease in price dispersion is generally not supported.

An explanation for the large price dispersion in Internet markets came from the

clearinghouse models which introduce informed and uninformed consumers. A shopbot may serve as a clearinghouse and consumers who access the shopbot will be fully aware of all prices. With a relatively easy access to a clearinghouse in the Internet market, a large part of informed consumers is expected. If consumers are better informed we would expect increase price competition, a more elastic demand and explains why prices in Internet markets decline on average. The existing price dispersion in Internet markets could be explained by the fact that there are still consumers who face high search costs online and some retailers choose to serve this part of consumers and are able set high prices.

With the model of Moraga- González et al (2015), a new structure was introduced were the assumption of sufficiently low consumer search costs (and as result consumers search as least once) is released. Where the previous models could only explain a decrease in prices because of lower search costs, the model of Moraga- González et al (2015) is able to explain increase in prices as well. Morgaga-González et al (2015) introduced two margins which are affected by a decrease in search costs, the intensive- and extensive search margin. The intensive search margin is affected by search intensity, the component which all previous models only took in account. The extensive search margin is affected by participation and is newly introduced. A decrease in search costs affect the intensive search margin in a way that people are going to search more, demand becomes more elastic and this will lead to lower prices. However, a decrease in search costs results in a more inelastic demand when looking at the extensive search margin. People who did not search before start searching now and these people have relatively high search costs. When looking at this component prices will increase. As result a decrease in search costs may result in higher or lower prices depending on which margin has more relative impact.

(18)

- 18 -

The first models considered gave some good intuitions about the effects of a decrease in search costs but all assumed product homogeneity. However as emerged from empirical research, studies find that even homogeneous goods acted like heterogeneous goods in the Internet market because of observable and unobservable retailer characteristics. Trust in a retailer is an important factor in e-commerce and mainly because of this factor we are not able to treat a products like a book, as a homogeneous product. Because of this finding, models for heterogeneous goods were examined as well.

Anderson and Renault (1999) combined the general model of search costs, with the general model for differentiated products. Anderson and Renault (1999) introduced that prices also depend on the degree of consumer taste for diversity. Their model outlined that consumers will search more as their search costs decrease and this can raise price competition, but at the same time as their taste for variety increases, consumers will search more for their best fit (for example for a retailer they trust) and become less price sensitive. Together this may explain why prices in Internet markets did not decrease, or did not decrease as much as expected.

By introducing the model of Bakos, the explanation of Anderson and Renault by increased degree of consumer taste for diversity is extended. A framework from Bakos, where search cost for prices and search costs for product information are separated, is used to divide consumers in two different types. Consumers who are price sensitive and other consumers who are not price sensitive and search for other product characteristics as for instance brand. Out of this interpretation of Bakos’ model, price dispersion in Internet markets may be explained by different consumer types. Some retailers will serve price sensitive consumers, and will face because of low search costs increased price competition and quote low prices. Other retailers serve the type of customers who are not price sensitive and search for other product characteristics, reducing search costs to obtain product information result in increased monopoly power for retailers and higher prices. Together this will imply that price dispersion in Internet markets may be higher than in conventional markets and average prices do not have to drop as much as speculated with the rise of the Internet.

Looking back at the explanations given by random search models, we may conclude that there is none unambiguous reason of why the Internet market is not as frictionless as predicted. Expected is that consumer preferences for other product characteristics than price play the biggest role for the friction in Internet markets. Prices in Internet markets may not have been decreased as much as expected after a reduction in search costs because of the fact that consumers became less price sensitive and searched more for product information than for price information. But in markets which were difficult to access in traditional form, the friction may be explained by the new

participating consumers who faces relative high search costs and make the total demand more inelastic.

The last thing to mention is that looking at the consumer side of the market gives not the whole story of the Internet market. Consumer gained power because of the Internet but retailers did as well. Search costs may not decrease that much in all markets on the Internet because of

obfuscation strategies of Internet retailers. All these things together makes the Internet market a very complex market and gives never ending opportunities for further research.

(19)

- 19 -

8. Bibliography

Ancarani, F., & Shankar, V. (2003). Symbian: Customer interaction through collabroration and competition in a convergent industry. Journal of Interactive Marketing, 17(1), 56-76. Anderson, S., & Renault, R. (1999). Pricing, Product Diversity, and Search Costs: A

Bertrand-Chamberlin-Diamond Model. The RAND Journal of Economics, 30(4), 719-735.

Bailey, J. P. (1998). Intermediation and Electronic Markets: Aggregation and Pricing in Internet

Commerce. Massachusetts: Ph.D. Dissertation.

Bakos, J. (1997). Reducing Buyer Search Costs: Implications for Electronic Marketplaces.

Management Science, 1676-1692.

Baye, M. R., Morgan, J., & Scholten, P. (2006). Handbook on economics and information systems. Amsterdam: Elsevier.

Baye, M., Morgan, J., & Scholten, P. (2004). Price Dispersion in the Small and in the Large: Evidence from an Internet Price Comparison Site. The Journal of Industrial Economics, 52(4), 463-496. Baylis, K., & Perloff, J. (2002). Price Dispersion on the Internet: Good Firms and Bad Firms. Review of

Industrial Organization, 21(3), 305-324.

Brown, J. R., & Goolsbee, A. (2002). Does the Internet make markets more competitive? Evidence from the life insurance industry. Journal of Political Economy, 110(3), 481-507.

Brynjolfsson, E., & Smith, M. (2000). Frictionless Commerce? A Comparison of Internet and Conventional Retailers. Management Science, 46(4), 563-585.

Brynjolfsson, E., Dick, A., & Smith, M. (2010). A Nearly Perfect Market. QME, 8(1), 1-33.

Brynjolfsson, E., Hu, Y., & Smith, M. (2003). Consumer Surplus in the Digital Economy: Estimating the Value of Increased Product Variety at Online Booksellers. Management Science, 49(11), 1580-1596.

Chevalier, J., & Goolsbee, A. (2003). Measuring Prices and Price Competition Online: Amazon.com and BarnesandNoble.com. Quantitative Marketing and Economics, 203-222.

Clay, K., Krishnan, R., Wolff, E., & Fernandes, D. (2002). Retail Strategies on the Web: Price and Non-Price Competition in the Online Book Industry. The Journal of Industrial Economics, 50(3), 351-367.

Diamond, P. A. (1970). A Model of Price Adjustment. Journal of Economic Theory, 156-168. Ellison, G., & Ellison, S. F. (2009). Search, Obfuscation, and Price Elasticities on the Internet.

Econometrica, 77(2), 427-452.

Erevelles, S., Rolland, E., & Srinivasan, S. (2001). Are prices really lower on the Internet? An analysis of the vitamin industry. Working Paper, University of California, Riverside.

(20)

- 20 -

Hann, I., & Terwiesch, C. (2003). Measuring the frictional costs of online transactions: The case of a name-your-own-price channel. Management Science, 49(11), 1563-1579.

Hoffman, D., Novak, T., & Peralta, M. (1999). Building consumer trust online. Communications of the

ACM, 42(4), 80-85.

Hong, H., & Shum, M. (2006). Using price distributions to estimate search costs. The RAND Journal of

Economics, 37(2), 257-275.

Hortacsu, A., & Syverson, C. (2004). Product Differentiation, Search Costs, and Competition in the Mutual Fund Industrie: A Case Study of S&P 500 Index Funds. The Quarterly Journal of

Economics, 119(2), 403-456.

Kim, H.-W., Xu, Y., & Gupta, S. (2012). Electronic Commerce Research and Applications. Elsevier, 241-252.

Moraga-González, J., Sándor, Z., & Wildenbeest, M. (2015). Prices and Heterogeneous Search Costs.

CEPR Discussion Paper, 1-37.

Morton, F. S., Zettelmeyer, F., & Silva-Risso, J. (2001). Internet Car Retailling. The Journal of Industrial

Economics, 49(4), 501-519.

Orlov, E. (2011). How does the Internet influece price dispersion? Evidence from the airline industry.

The journal of industrial economics, LIX(1), 21-37.

Pavlou, P., & Gefen, D. (2004). Building effective online marketplaces with instituation-based trust.

Information systems research, 15(1), 37-59.

Peitz, M., & Waldfogel, J. (2012). The Oxford Handbook of the Digital Economy. Oxford: Oxford University Press.

Reinganum, J. (1979). A simple model of equilibrium price dispersion. The Jorunal of Political

Economy, 87, 851-858.

Salop, S., & Stiglitz, J. (1977). Bargains and Ripoffs: A Model of Monopolistically Competitive Price Dispersion. The Review of Economic Studies, 44(3), 493-510.

Sengupta, A., & Wiggins, S. N. (2014). Airline Pricing, Price Dispersion, and Ticket Characteristics On and Off the Internet. American Economic Journal: Economic Policy, 6(1), 272-307.

Singh, N., & Vives, X. (1984). Price and Quantity Competition in a Differentiated Duopoly. THE RAND

Journal of Economics, 15(4), 546-554.

Stigler, G. J. (1961). The Economics of Information. Journal of Political Economy, 69(3), 213-225. Varian, H. (1980). A Model of Sales. The American Economic Review, 70(4), 651-659.

Walter, Z., Gupta, A., & Su, B.-c. (2006). The Sources of On-Line Price Dispersion Across Product Types An integrative View of On-Line Search Costs and Price Premiums. International Journal

(21)

Prices and price dispersion in internet markets : why is the internet market not as frictionless as expected? : an explanation by random search models

Bachelor Thesis Economics and Business – Economics

Faculty of Economics and Business