Estimating Switching Costs in Electricity Retailing

(1)

Estimating Switching Costs in Electricity Retailing

Empirical Evidence from Sweden

Emil Björkman

(2)

Abstract

In many situations, consumers face a cost for switching suppliers, either in monetary or in non-monetary terms. The cost associated with the switch is called switching cost and could lead to consumers that are locked-in to their current supplier. In the Swedish electricity retail market, evidence suggests that there are few switches of suppliers. Thus, one could possibly think of the switching costs to be reducing the number of supplier changes, working as a switching barrier.

Two different models are used to estimate the switching costs in the Swedish electricity retail market, Shy’s (2002) “quick-and-easy method for estimating switching costs” and Salies’ (2012) developed version of the model. The estimates vary greatly, depending on the choice of model. Some of the problematic assumptions of the original model are dealt with in the latter model, which is expected to make it more credible. The results indicate that there are switching costs present in the market, and that they are much higher for the consumers of large suppliers compared to the consumers of the smaller ones. Hence, it is more costly to make a switch from a large supplier than from a smaller supplier. The estimated switching costs for a household with an annual electricity consumption of 2,000 kWh (small apartment) amount for up to 600 SEK or 1,100 SEK, depending on which model one uses. In relative terms, the switching costs could account for up to 52 % or 94 % of the annual electricity price, respectively. Although switching costs are present and make it costly for changing supplier, they cannot explain the low number of switches completely on their own. The low consumer mobility in the market might to some extent be due to the switching costs but is also likely to be depending on the low savings from switching supplier.

The results in this study point in the same direction as the previous research in many aspects.

However, several problems with the models are pointed out in this study, which may affect their credibility. Some of these problems are possible to correct for, while some others are not. As a consequent, the models might be overestimating the switching costs, and therefore opens for alternative methods of estimating the switching costs. Not only is there a need for other models that can estimate the switching costs in electricity retail markets specifically, but rather for all different markets.

Keywords: Switching Costs, Undercut-Proof Property (UPP), Electricity Retailing

(3)

1. Introduction

In 1996, the Swedish electricity market was deregulated, so that the retail market was opened for competition (Swedish Energy Markets Inspectorate 2018). Previously, the Swedish consumers had to buy electricity from their local distributor. To increase the competition as well as making the electricity retail market more efficient, the deregulation was considered as necessary (Swedish Competition Authority 1996). An efficient retail market would lead to more options and presumably lower prices for the consumers. Whether the deregulation was successful or not can be discussed, and there are reasons to believe that the market is not as efficient as it should be. Some evidence suggests that even though there exist monetary benefits by switching supplier, which the consumers often are aware of, the change does not happen as often as one would expect. The lack of switches could possibly depend on the existence of switching costs in the market, where the consumers face a cost for changing supplier.

Switching costs are defined as the one-time cost for switching from one supplier to another (Burnham et al. 2003), and they could be divided into both monetary (as cancellation costs) and non-monetary (as time consumption) costs (Jones et al. 2002).

The switching costs complicate the supplier changes, where this particular type of costs have a lock-in effect of the consumers and may therefore counteract competition. In a market without any switching costs present, a change from one supplier to another would simply be a decision about price and quality. Further suppose a homogeneous good, where the quality is identical between the suppliers, and then it should only be the price that matters for the decision. As a result, in a perfect market, the price should be equal across all suppliers, more precisely equal to the marginal cost. However, markets are most often not perfect and with switching costs present, the consumer’s choice of supplier gets even more complicated. It is not only the different prices for the homogeneous good that matters, but also the switching costs affect the decision. Lots of evidence suggests that the cost of switching is an obvious obstacle preventing switches between suppliers, making it possible for the suppliers to set prices above marginal costs. Therefore, it is relevant to empirically investigate and estimate the switching costs in different markets with a homogenous good, such as the electricity retail market.

A highly homogeneous good that is a well-known example in economics is electricity. In its simplest explanation, it is a “movement and interaction of electrons” (Merriam-Webster 2021).

The electricity that a consumer purchases is identical, independent on which supplier that

(6)

6 provides the electricity or how it is produced. The costs of switching supplier can, however, make consumers locked-in to their current electricity retail supplier, where the consumers do not want to switch supplier because of the switching costs that they face. In the case with high switching costs, one will encounter a situation where consumers stay with the same supplier, even though some other electricity retail supplier offers electricity at a lower cost.

1.1 Previous research

In the middle 80´s, von Weizsacker (1984) was one of the first that noticed how switching costs, or “substitution costs” as he calls them, could theoretically lead to market power for the suppliers. There is later evidence that shows that switching costs have a positive relation with market power, which implies that the larger the switching costs are, the larger the supplier’s market power will be (Egarius & Weill 2016). By using a two-firm and two-period model, with either differentiated or homogeneous products, Klemperer (1987a & 1987b) suggests that switching costs affect competition negatively. For instance, he shows how the switching costs could yield monopoly power to the suppliers, where they could lock-in their current consumers and set a higher price in future periods. The prices in the first period although, could be either higher or lower in a market with switching costs present. A problem with the models that Klemperer uses is the fact that they are assuming an “end-of-the-world” scenario, where the suppliers only consider the first and the second period in their pricing behaviour. Therefore, Beggs and Klemperer (1992) extend the models to infinite-horizon models, i.e. models with an infinite number of periods. The results are nevertheless the same as in the previous models, suggesting a negative impact on competition and either higher or lower prices in the first period.

Furthermore, Klemperer (1995) examines the nature of the switching costs more closely, where he divides the switching costs into several parts and describes each part in-depth.

Even though there exists theoretical evidence for how the switching costs affect competition negatively and how they work as a lock-in mechanism, there is not as much empirical evidence.

Farrell and Klemperer (2007, p. 15) state that: “The empirical literature on switching costs is much smaller and more recent than the theoretical literature”. As the switching costs are not directly observable and require indirect methods to estimate them, it explains the relatively small empirical evidence. Hence, the need for a useful model that could empirically estimate the switching costs was great, and finally, Shy (2002) did create such a model. He refines a concept in form of the “Undercut-proof property” (UPP), which has its roots in the

“Undercut-proof equilibrium” (UPE) that Morgan and Shy (2000) propose. When the UPP is

(7)

7 satisfied, the suppliers’ prices are functions of the market shares and the unobservable switching costs. By using data for the observations of prices and market shares, it is then possible to estimate the switching costs of a representative consumer according to Shy’s model. The advantage of the model is that it is easy to use, which the name “A quick-and-easy method for estimating switching costs” clearly indicates (Shy 2002, p. 71). His model has been applied in many different empirical studies, where switching costs have been estimated in several different markets for homogeneous goods. Examples could be found in the airline sector (Carlsson & Löfgren 2006), the banking industry (Egarius & Weill 2016), the broadband industry (Krafft & Salies 2008), as well as in the telecommunications market (Lörincz & Nagy 2010).

The model of Shy (2002) has also been used for estimating the switching costs in electricity retail markets. Salies (2005) applies his model to estimate the switching costs in the British electricity retail market, where he finds large switching costs for switching from an incumbent electricity retail supplier to an entrant. More recently, Magnani et al. (2020) use Shy’s model to estimate the switching costs in the Italian electricity retail market. Their results say that the consumers of the larger suppliers face larger switching costs than the consumers of the smaller suppliers. This makes it more expensive for the large supplier´s consumers to exit their contracts and switch to another electricity retail supplier, compared to the consumers of the smaller ones.

An interesting idea that they present, is that even though there are switching costs present in the market, there are lots of money to spare for the consumers. Magnani et al. (2020, p. 20) mean further that the “lack of consumer engagement is one of the biggest weaknesses of the liberalized retail electricity markets in European countries”.

There have been some developments of the original model proposed by Shy (2002).

Salies (2012) tries to improve the model by adding some more realistic assumptions to the original model. Even though he adds some more assumptions, the model still retains its simple construction and user-friendly application, only requiring data for the suppliers’ prices and market shares. This developed model has, unfortunately, not had the same impact in research as Shy’s model, even though it improves the model in several important ways.

It should be mentioned that there exist some alternative ways to estimate the switching costs, although they might be more complex. One alternative method is the “stated preference method”, where one uses the preferences of individual respondents about changing supplier.

Lee et al. (2006) use this particular method when estimating the switching costs in the Korean mobile telecommunications market. This approach implies that they use the preferences of

(8)

8 some randomly assigned respondents, obtained by specifically asking about their preferences regarding the mobile telecommunication. The respondents list which reasons that may, hypothetically, get them to change mobile operator. Then, they use these preferences to estimate a probit model, with the probability to switch supplier as their dependent variable. From this model, they can estimate the switching costs. It would certainly be possible to use this kind of approach in other markets than the mobile telecommunications, such as the electricity retail market. However, the problem with the stated preference method is the much more complex model, and the fact that it is based on a hypothetical setting. The respondents might answer the hypothetical question in some way but might act in a completely different way in reality.

Another big problem with the stated preference method is that it requires a comprehensive survey with many respondents.

If one looks more broadly at previous research about the switching behaviour of consumers regarding electricity retail suppliers and contracts, one will find a common thread in many of them. In a study by Ek and Söderholm (2008), they get an explicit answer for how many Swedish households that had switched electricity retail supplier or renegotiated their existing contracts during the last five years. According to their study, less than 30 percent had switched supplier and less than 25 percent had renegotiated their contracts. The switching behaviour for a homogeneous good should be stronger since there is some money to spare by actively compare electricity retail suppliers and contracts, even though the sizes of the potential savings might vary. As Goett et al. (2000) address, the inactive households lead to increased market power for the electricity retail suppliers, which can take advantage of their inactive behaviour. In a similar way, Vesterberg (2018) investigates the household´s switching behaviour in the Swedish electricity retail market. He notes that the consumer´s switching between electricity contracts is noticeable low, where he concludes that to increase the frequency of switching and thereby reducing the suppliers’ market power, switching barriers as switching costs must be decreased.

In short, the previous research gives an indication of how the switching costs in the Swedish electricity retail market might look like. The switching costs should reasonably be quite high, in line with research from other European countries, where a change from a large electricity retail supplier is associated with a higher switching cost than a change from a smaller one. In Sweden there is evidence that the switching between suppliers is shockingly low, which certainly could indicate that the Swedish consumers face high switching costs.

(9)

9

1.2 Purpose and delimitations

The purpose of this study is to estimate the switching costs in the Swedish electricity retail market, and to investigate some of the existing models for estimating switching costs. By only studying a selection of different electricity retail suppliers, for one specific electricity area and only during a specific year, the range of this study will be limited.

1.3 Disposition

This study will be disposed as follows:

Chapter 2 summarizes and describes relevant theory of switching costs.

Chapter 3 briefly describes the Swedish electricity market and the electricity retailing.

Chapter 4 explains the empirical methods that will be used for estimating the switching costs.

Chapter 5 introduces the data that is used in this study.

Chapter 6 presents the results of the estimations and some limitations of the models.

Chapter 7 discusses and analyses the results and its implications.

Chapter 8 concludes the study and its findings shortly.

(10)

10

2. Theory

This section defines switching costs and describes the basic theory of these costs, how to model them economically and how they affect welfare.

2.1 Definition of switching costs

The switching cost is the perceived (actual or experienced) cost for changing supplier of a homogeneous product, from the current supplier to a new one (Jones et al. 2002). Therefore, the switching costs make it costly, either in monetary or non-monetary terms, to change supplier. Switching costs is suggested to be “one-time” costs (Porter 1980, p. 10). As they are costly for the switching consumer, it results in weakened incentives for changing supplier.

Klemperer (1995) shows that it is possible to divide the switching costs into six separate groups, which are characterised by how they arise.

I. Compatibility with existing equipment

With technical products, it is reasonable that one needs other compatible products.

Think of computers with their charger, mobile phones with their protective case, printers with their colour cartridges.

II. Transaction costs

The costs of creating a new relationship with a new supplier, and of ending the current one. Think of costs for closing one bank account and open a new one, or it could be costly to terminate a long-term contract, as changing the long-term electricity retail supplier when having a fixed price contract.

III. Learning to use new brands

If the consumer is familiar with the use of some product, the consumer has incentives to continue to purchase that product. Even if there is some product with identical quality, one may just buy what one is used to. Think of buying a computer, which is functionally identical to some other computer, but with different operative systems. In that case, it is easier to buy a new one of the same brand because one already knows how to use it and does not need to learn something new.

IV. Uncertainty of new brands

Consumers tend to keep purchasing from the same brands, since they know what they get and what they can expect. Even if some other product is of same quality but cheaper, the consumer hesitates because of the uncertainty. Think of buying salt at the grocery

(11)

11 store, where there might be some cheaper salt that is as good as the one that the consumer usually buys. However, the uncertainty makes the consumer keep buying the same as before.

V. Discount coupons and similar

There can be some monetary advantages of buying more from the same brand. Think of airline’s “frequent-flyer” programmes, offers of getting your sixth haircut or meal for free etc.

VI. Psychological costs

Even though the consumer has no monetary reasons to stay with the same brand, some psychological reasons could affect the switching behaviour. Think of brand loyalty and preferring the “safe” choice of sticking to the same brand. It could also be about loyalty to the local brand or supplier.

Furthermore, one could add one more important part to the switching costs, which is search costs (Shapiro & Varian 1998). For consumers, the search costs include psychological costs of changing habits, the time and effort to find a new supplier, and the potential risk of choosing a new, unknown supplier. It is mentioned that these costs may not be substantially large, but they still have an impact on the changes from one supplier to another. Especially in today’s information society, one can believe that the search costs are lower than ever before because of the modern technology. One can now easily compare many suppliers’ prices in just a few clicks on a computer or a smart phone. Although the search costs might be small, they tend to be perceived as higher than they actually are by the consumers (Giulietti et al. 2005, Swedish Competition Authority 2009). Note that the smaller the search costs, the higher the competition, which will lead to lower prices, ceteris paribus (Stiglitz 1989). Finally, it is important to remember that the consumers do not search for every potential supplier, instead they make sequential decisions if they will continue to search or not (Rothschild 1974).

The implication of these switching costs is that it gives the suppliers market power.

Klemperer (1995) explains in a simple theoretical way why that is the case. Think of a duopoly situation with two suppliers, 𝐴 and 𝐵, and only one time period. Both suppliers sell the same homogeneous product. Consumers that have bought from one of the suppliers before have a switching cost of 𝑆 to switch to the other supplier. To steal supplier 𝐵’s consumers, supplier 𝐴 must cut its price, so it is at least 𝑆 lower than the price of 𝐵. If 𝑆 is substantially large enough, supplier 𝐴 will make a loss if it must decrease its price for all its existing consumers. That would

(12)

12 happen if the profit lost from cutting the price is larger than the profit increase from the new consumers that are stolen from 𝐵. In such a case, 𝐴 would prefer to keep its current consumers and act as a monopolist for its own customer base, which is the same for 𝐵 as well. In this simple theoretical case, the firm with the largest consumer group will then make the largest profit, which shows the importance of large market shares.

2.2 Two-period switching costs model

Klemperer (1987b & 1995) presents a two-stage model, which more formally could show how the consumers do not face any switching costs in the first period, while they face such costs in the second period. These switching costs bind them to the supplier chosen in the first stage.

Let 𝜋_𝑡^𝐹 be the profit of supplier 𝐹 in period 𝑡, and 𝜎_𝑡^𝐹 be the market share of supplier 𝐹 in period 𝑡, for 𝑡 = 1, 2. Then, supplier 𝐹´s discounted profit is:

𝑉^𝐹 = 𝜋₁^𝐹 + 𝛿𝜋₂^𝐹(𝜎₁^𝐹) (1)

The variable 𝛿 ∈ [0, 1] is a discount factor. In the first period, supplier 𝐹 wants to maximize the discounted profit with respect to the price in time period 1, 𝑝₁^𝐹. Firm 𝐹´s first order condition for an equilibrium is:

𝜕𝑉^𝐹

𝜕𝑝₁^𝐹 = ^𝜕𝜋¹^𝐹

𝜕𝑝₁^𝐹+ 𝛿^𝜕𝜋²^𝐹

𝜕𝜎₁^𝐹

𝜕𝑝₁^𝐹 = 0 (2)

By looking closer at the different derivatives, one can assume that 𝜕𝜎₁^𝐹⁄𝜕𝑝₁^𝐹 < 0, since the supplier’s market share is supposed to decrease when the price in the first period increases.

Also, it is possible to assume that 𝜕𝜋₂^𝐹⁄𝜕𝜎₁^𝐹 > 0, so that the profit in the second period increases when the supplier’s market share in the first period increases. These two assumptions would imply that the last term is negative, 𝛿^𝜕𝜋²^𝐹

𝜕𝜎₁^𝐹

𝜕𝑝₁^𝐹 < 0. Then, it must be true that 𝜕𝜋₁^𝐹⁄𝜕𝑝₁^𝐹 > 0, if 𝜕𝑉₁^𝐹⁄𝜕𝑝₁^𝐹 = 0 is going to hold. More precisely, this implies that 𝑝₁^𝐹 is lower than the price that gives 𝜕𝜋₁^𝐹⁄𝜕𝑝₁^𝐹 = 0, i.e. the price is lower than in the case without switching costs present. This can be seen as a proof for that switching costs in the second period affect the first periods prices, where they will be lower in the case with future switching costs.

Some examples in the real life could indicate that this competition about consumers in the first period exists. Special offers for first-time purchasers are usually available in lots of different industries. First-time offers for cell phones and streaming services are both examples of how suppliers set a lower price in the first period. The goal is to obtain higher market shares and increase the prices in later periods when the customer base is larger.

(13)

13 Note that the assumption of 𝜕𝜋₂^𝐹⁄𝜕𝜎₁^𝐹 > 0 does not necessarily always apply. Suppliers with small market shares have stronger incentives to set low prices and get new consumers than a supplier with a large market share. The reason for this is that the supplier with a small market share is better off by getting more consumers and give up some profit from the existing ones.

This could result in a situation where a supplier has a negative relationship between market share and profit. If it is increasing its own market share, it means that it is decreasing some others market shares, so that they get more aggressive in future periods. In this case it must be true that 𝜕𝜋₂^𝐹⁄𝜕𝜎₁^𝐹 < 0, which implies that 𝜕𝜋₁^𝐹⁄𝜕𝑝₁^𝐹 < 0. As a result, 𝑝₁^𝐹 is higher than the price that gives 𝜕𝜋₁^𝐹⁄𝜕𝑝₁^𝐹 = 0. The prices are then larger in the first period then in a situation without switching costs present, which depends on the fact that some suppliers are avoiding gaining market shares. This shows how the prices in the first period might either be higher or lower when switching costs are present in the market. It depends on the sign of the derivative of 𝜕𝜋₂^𝐹⁄𝜕𝜎₁^𝐹, i.e. the relationship between first-period market shares and second-period profits.

Going back to the assumption that 𝜕𝜋₂^𝐹⁄𝜕𝜎₁^𝐹 > 0, it is important to remember that it does not imply that the suppliers compete more than without switching costs. Instead, one can say that it means that they compete more than if they would not take future profits into account. Also, the consumers are supposed to be rational. If they know about the switching costs and how the suppliers behave, they are supposed to be aware of the risk to potentially be locked-in by the suppliers in future periods. Rational consumers do further know that the supplier that takes the lowest price will get a large market share and raises its price in the future. Therefore, the consumer might avoid the cheapest supplier in the first period, which would lead to less competition in the first period compared to the case without the presence of switching costs.

2.3 Multi-period switching costs model

The two-period switching costs model is useful in examining how firms lock-in the consumers, where new consumers enter in the first period and are locked-in to that supplier in the second period. However, this model does not answer the question what happens when new consumers are mixed up with old ones. In such a case, there are new consumers entering every single period and suppliers cannot discriminate between new and old consumers. The suppliers will then face a problem, lower their prices for all consumers and attract new ones, or keep high prices for the existing consumers that are locked-in. For such a question, one need to look at a more dynamic switching costs model for multiple time periods. Farrell and Shapiro (1988) propose such a model, where they look at a duopoly case with multiple periods. They show

(14)

14 how the supplier with the larger market share will be less interested in setting low prices and getting new consumers. This is because the larger supplier will maximize its total future discounted profit by setting a higher price for its existing consumers. Therefore, the supplier with the lower market share sets the lowest price and get the new consumers.

Beggs and Klemperer (1992) show in a similar manner, with other and less narrow assumptions than previous authors, that the results from the two-period model also applies to the case with multiple time periods. For simplicity, one can yet again look at one of Klemperer’s models.

Klemperer (1995) develops a multi-period model in a different and more basic way, where he simplifies and suppose a constant number of consumers in the market, without any new consumers entering the market or old ones leaving the market. As in the two-period model, he assumes that firm 𝐹 will maximize its future discounted profit, 𝑉_𝑡^𝐹, in time period 𝑡:

𝑉_𝑡^𝐹 = 𝜋_𝑡^𝐹 + 𝛿𝑉_𝑡+1^𝐹 (𝜎_𝑡^𝐹) (3)

Once again, it is possible to see that the value of time period 𝑡 + 1 is depending on the market share in previous period, time period 𝑡. By maximizing the price in period 𝑡, 𝑝_𝑡^𝐹, one obtains the following first-order condition:

𝜕𝑉_𝑡^𝐹

𝜕𝑝_𝑡^𝐹 = ^𝜕𝜋^𝑡^𝐹

𝜕𝑝_𝑡^𝐹+ 𝛿^𝜕𝑉^𝑡+1^𝐹

𝜕𝜎_𝑡^𝐹

𝜕𝑝_𝑡^𝐹 = 0 (4)

If one looks closer at the different derivatives, one can assume that a lower current price raises the suppliers current market share, 𝜕𝜎_𝑡^𝐹⁄𝜕𝑝_𝑡^𝐹 < 0. Then assume that 𝜕𝑉_𝑡+1^𝐹 ⁄𝜕𝜎_𝑡^𝐹 > 0, which would imply that the last term is negative, 𝛿^𝜕𝑉^𝑡+1^𝐹

𝜕𝜎_𝑡^𝐹

𝜕𝑝_𝑡^𝐹 < 0. For the first-order condition to hold in that case, so that 𝜕𝑉_𝑡^𝐹⁄𝜕𝑝₁^𝐹 = 0, it must be true that 𝜕𝜋_𝑡^𝐹⁄𝜕𝑝_𝑡^𝐹 > 0. This says that the firm sets a price lower than in a case where the future profit does not matter. However, it does not say anything about how the pricing is compared to a case without switching costs. In a case where such costs are present in the market, the supplier’s demand is likely to be more inelastic.

As previous mentioned, the supplier faces a problem, either it lowers its price, gains a larger market share and increases its future profit, or it charges a high price to maximize its profit from the already locked-in consumers.

To see whether the switching costs give higher or lower prices for the consumers, it is possible to rewrite supplier 𝐹’s total future discounted profit function as an explicit function of its rival’s price:

𝑉_𝑡^𝐹 = 𝜋_𝑡^𝐹(𝑝_𝑡^𝐹, 𝑝_𝑡^𝐺) + 𝛿𝑉_𝑡+1^𝐹 (𝑝_𝑡^𝐹, 𝑝_𝑡^𝐺, 𝑝_𝑡+1^𝐹 , 𝑝_𝑡+1^𝐺 ) (5)

(15)

15 The rival firm 𝐺’s price is written as 𝑝_𝑡^𝐺 and 𝑝_𝑡+1^𝐺 . For this explicit function, the first-order condition for firm 𝐹 is:

𝜕𝑉_𝑡^𝐹

𝜕𝑝_𝑡^𝐹 = ^𝜕𝜋^𝑡^𝐹

𝜕𝑝_𝑡^𝐹+ 𝛿 (^𝜕𝑉^𝑡^𝐹

𝜕𝑝_𝑡^𝐹+^𝜕𝑉^𝑡+1^𝐹

𝜕𝑝_𝑡+1^𝐺

𝜕𝑝_𝑡^𝐹 ) = 0 (6)

The effect of switching costs on prices arises in two different ways. Firstly, consumers that already purchased from a supplier are locked-in to that supplier to some degree. If suppliers only cared about current profits, they could set higher prices in a case where the consumers face switching costs. It is the same effect as the one that was shown in the second period of the two-period model, which is a less elastic demand, so that 𝜕𝜋_𝑡^𝐹/𝜕𝑝_𝑡^𝐹 is larger at any price at the relevant range. This implies that the supplier must increase its price to get equation (6) in balance. Secondly, the suppliers know that lower current prices will increase their future profits through their increased market shares, 𝜕𝑉_𝑡+1^𝐹 /𝜕𝑝_𝑡^𝐹 < 0. It is the same effect that was generally dominant in the first period in the two-period model. This implies that the supplier must decrease its price to get equation (6) in balance.

As one can see, these two effects pull the price in opposite directions, where the dominant effect will determine whether the effect of switching costs will increase or decrease the price that the suppliers set. In a simple model without discounting, these effects might balance each other.

However, in a more realistic model, there exist some additional effects in the case with the presence of switching costs. One example is that the discount factor, 𝛿, reduces the incentives to attract new consumers relative the suppliers´ willingness to exploit their current customer bases. Furthermore, the rational consumers do know that the low prices they face at the start will increase in the coming time periods, so that the consumers have a less elastic demand. The suppliers do also know that if they set higher prices today, their rivals will become less aggressive in the future, which they will find positive.

2.4 Welfare effects of switching costs

By using the previous examined models, Klemperer (1995) argues for different effects on the welfare. The most obvious is the market power that the suppliers can get from the switching costs, having the opportunity to lock-in consumers. This market power could theoretically be used to extract the consumer surplus and obtain monopoly profits from the customer base, resulting in socio-economic efficiency losses.

(16)

16 In the long run, the possibility to lock-in consumers will presumably lead to higher prices, possibly even monopoly prices in the extreme case. However, the effect on prices is not self-evident. Farrell and Klemperer (2007) say that the prices do not necessarily need to be higher in the case with switching costs present. The effect on prices is rather ambiguous and seem to depend on the supplier’s balance between low and high prices. An interesting conclusion about the effect of switching costs on prices is presented by Fabra and García (2012).

They suggest that the switching costs’ effect on prices depend on the distribution of the market shares of the suppliers. With markets characterized by large spreads of the market shares, the switching costs tend to increase prices as the suppliers behave less aggressive in the competition. In the opposite cases, when the market shares are more evenly distributed between the suppliers, switching costs rather lead to lower prices as the suppliers compete more aggressively. Looking at the empirical evidence, it seems like they differ. Some suggest that prices tend to be higher in markets where switching costs exist (Farrell & Schapiro 1988, Beggs & Klemperer 1992, Padilla 1995), while some say that the opposite is true, i.e. that switching costs decrease prices (Dubé et al. 2009, Doganoglu 2010, Rhodes 2014).

By getting the consumers locked-in to a specific supplier, the switching costs contribute to a reduced consumer mobility, with less changes of suppliers. The switching costs and the market power can also create entry barriers that make it difficult for new suppliers to entry the market.

Not only do the prices need to be lower than the rivalry suppliers’ prices, but it must also subsidize the switching costs for the switching consumers. (Klemperer 1995)

In summary, the switching costs have some different effects on welfare. The effect on the prices is ambiguous, where they could either increase or decrease the prices. It seems that the switching costs create barriers for the consumers as well as for the entrant suppliers. They make it costly for the consumers to switch supplier and make it hard for the entrant suppliers to compete for consumers. Furthermore, the switching costs make it possible for the suppliers to obtain monopoly profits from their customer bases, implying socio-economic welfare losses.

In the long run, the switching costs could in worst case lead to markets that are less competitive and with increased prices. Therefore, from a socio-economic point of view, it is crucial to minimize the switching costs in a market to reduce the welfare losses.

(17)

17

3. The Swedish electricity market

3.1 Description of the market

As previously mentioned, the Swedish electricity retail market was deregulated in 1996 (Swedish Energy Markets Inspectorate 2018). This meant that the consumers were free to choose their own electricity retail supplier, and not forced to have their local electricity distributor. From a proposition of the Swedish government (Prop. 1991/92: 133), some overarching goals of the deregulation could be identified. Primarily, the government wanted to increase the competition and make the electricity trading more efficient, with more options and lower prices for the consumers as a likely result.

There is an important distinction that must be drawn for understanding the Swedish electricity market. That is, the difference between the electricity network and the electricity retailing, i.e. the actual electricity. Swedish consumers can only choose the supplier of the electricity retailing. The electricity network is owned by various electricity network companies, which have exclusive rights to the electricity distribution within their geographical area. However, the consumer can freely choose which company it wants to buy the actual electricity from.

(Vattenfall 2021)

Since the deregulation in 1996, the physical trade of electricity is organized by the marketplace Nord Pool, where the Nordic and Baltic countries are connected at the same place. At Nord Pool, the electricity prices are determined, and they depend on the cost of producing the last kilowatt hour (kWh) of electricity to meet the demand.¹ The financial trade of electricity, however, is mainly taking place on Nasdaq Commodities. In that place, the electricity retail suppliers can trade long-term contracts and price hedging opportunities for days, weeks, quarters, or years. (Swedish Energy Markets Inspectorate 2017)

More recently, in 2011, Sweden was divided into four different electricity areas (The Swedish Consumer Energy Markets Bureau 2020a). As demonstrated in Figure 1 (el.se n.d.), the numbers go from 1 to 4 and start from the north. This means that the north part of Sweden has electricity area number 1, while the south part has area number 4, with area number 2 and 3 in between.

1 When talking about electricity prices and consumption, one usually talks in terms of electricity in kWh.

(18)

18 Figure 1. The Swedish electricity areas

Electricity area 1 and 2 do have a surplus of electricity, where more electricity is produced than consumed in these areas (Swedish Energy Markets Inspectorate 2017). In area 3 and 4 there are a deficit of electricity on the other hand, so that the production of electricity is smaller than the consumption. The electricity prices may deviate between the areas, depending on the supply and demand in the specific electricity area. For example, Vattenfall’s variable prices varied greatly in August 2020, where the variable prices for electricity areas 1 and 2 were about 0.15 SEK/kWh, while the variable prices for area 3 and 4 were 0.35 SEK/kWh and 0.42 SEK/kWh, respectively (Vattenfall 2020). This example is certainly an extreme case, but either way it emphasizes the importance to look at prices for the same electricity area when comparing the prices of different suppliers. Otherwise, there is a great risk that a misconception will be obtained.

In Sweden, a large proportion of electricity is produced from renewable energy sources, such as water, wind and the sun (Swedish Energy Agency 2020). As can be seen in Figure 2 (Holmström 2021), the share of electricity produced from water and wind accounts for 62 % of the total electricity production in 2020. The nuclear power accounts for 30 %, while other sources as thermal power and sun power (diminutive) accounts for the rest of the electricity production in Sweden.

(19)

19 Figure 2. Renewable energy production in Sweden

3.2 Switching costs in the electricity retail market

In a report from the Swedish Competition Authority (2018), it is stated that only ten percent of the Swedish households change electricity retail supplier annually. It further mentions that the small price differences are likely to be one possible explanation to the low percentage of supplier changes. However, Oberoende Elhandlare (2021) disagrees and says that the competition situation is far from optimal, claiming that the Swedish electricity retail market does not function well at all. The organization lists the need for improved competition as the most important question to solve in the future. According to Oberoende Elhandlare, one explanation to the bad function of the market is, among other things, the financial barriers to switch supplier, such as switching costs

As can be seen, there are different points of view regarding the competition in the electricity retail market. However, one concluding remark that is applicable to the opinions of both parties mentioned above, is that there are few consumers that change electricity retail supplier. Whether it depends on small price differences or switching barriers (as switching costs) for the consumers is not obvious though.

From the different types of switching costs that was presented earlier in section 2.1, the switching costs in the Swedish electricity retail market can be divided into a couple of different groups:

Water Wind Nuclear Other

(20)

20 Transaction costs

There are some costs that arise in connection with a change of electricity retail supplier. If the consumer has a fixed price contract, the transaction cost is the cancellation fee for terminating the contract prematurely. However, when having a variable price contract, there are no transactions costs for switching electricity retail supplier.

Uncertainty of new brands

The consumers may be uncertain about the services of the other suppliers, while they know what they can expect from their current electricity retail supplier. This uncertainty may affect the consumer and hence the switching behaviour.

Psychological costs

Some psychological costs can also affect the decision of switching supplier. One such is brand loyalty, where the consumer can have built up a sense of security for the brand. It could potentially also be a loyalty to the neighbourhood, where the consumer wants to support the local electricity retail supplier. Moreover, since the consumer needs to pay for both the electricity network and the electricity retailing, the consumer might prefer to get both these costs on the same bill to minimize the number of separate bills.

Search costs

In order to switch electricity retail supplier, the consumer will most likely face a search cost, which is the cost of searching for alternative suppliers. These costs should be relatively small, as it exists several price comparison sites that can compare the electricity retail suppliers’ prices for the specific consumer and its household. However, they might be overestimated by the consumers themselves (Giulietti et al. 2005, Swedish Competition Authority 2009).

(21)

21

4. Empirical model

In this section, the two different models will be presented in-depth, starting off with the original one from Shy (2002) and followed by the developed model of Salies (2012).

4.1 Shy´s model

Shy (2002) created an empirical model to estimate the switching costs, which he means is a

“quick-and-easy method for estimating switching costs” (p. 71). In its simplicity, it can estimate the implicit switching costs by only using prices and market shares of the different suppliers in a market for a homogeneous good. As the switching costs are not directly observable, and it is hard to find relevant data for measuring these, Shy´s model is of interest since data on prices and market shares are easily accessible.

4.1.1 Assumptions of the model

Consider a duopoly market with the two suppliers A and B, that are selling a homogeneous product for the prices 𝑝_𝐴 and 𝑝_𝐵, respectively. One can assume that in this time period, period 𝑡, all consumers have bought from one of the suppliers before. The cost of switching supplier can be denoted by 𝑆 and is assumed to be positive, 𝑆 > 0. The utility for the consumers that have purchased from supplier 𝐴 is 𝑈_𝐴, and similarly 𝑈_𝐵 is the utility for the ones that purchased from supplier 𝐵. The utility of each consumer can be written as:

𝑈_𝐴 ≡ {−𝑝_𝐴 staying with brand A

−𝑝_𝐵− 𝑆 switching to brand B 𝑈_𝐵 ≡ {−𝑝_𝐴− 𝑆 switching to brand A

−𝑝_𝐵 staying with brand B (7)

Furthermore, let 𝑛_𝐴 and 𝑛_𝐵 be the number of consumers purchasing from supplier 𝐴 and supplier 𝐵, respectively. 𝑁_𝐴 and 𝑁_𝐵 are the number of consumers that bought from each supplier before, i.e. in period 𝑡 − 1. Equation (7) implies that:

𝑛_𝐴 = {

0 if 𝑝_𝐴 > 𝑝_𝐵+ 𝑆 𝑁_𝐴 if 𝑝_𝐵− 𝑆 ≤ 𝑝_𝐴 ≤ 𝑝_𝐵+ 𝑆 𝑁_𝐴 + 𝑁_𝐵 if 𝑝_𝐴 < 𝑝_𝐵− 𝑆 𝑛_𝐵 = {

0 if 𝑝_𝐵> 𝑝_𝐴 + 𝑆 𝑁_𝐵 if 𝑝_𝐴 − 𝑆 ≤ 𝑝_𝐵≤ 𝑝_𝐴 + 𝑆 𝑁_𝐴+ 𝑁_𝐵 if 𝑝_𝐵 < 𝑝_𝐴− 𝑆

(8)

(22)

22 By assuming that the suppliers’ production costs are zero, so that they have a marginal cost of zero, 𝑀𝐶_𝐴 = 𝑀𝐶_𝐵 = 0, their profits are given by:

𝜋_𝐴(𝑝_𝐴, 𝑝_𝐵) = 𝑝_𝐴𝑛_𝐴

𝜋_𝐵(𝑝_𝐴, 𝑝_𝐵) = 𝑝_𝐵𝑛_𝐵 (9)

Note that both 𝑛_𝐴 and 𝑛_𝐵 are given by equation (8).

The important assumptions that Shy makes will be crucial for the entire model to hold. To clarify which these assumptions are, they are listed below:

I. Homogenous good II. Positive switching costs III. Symmetric switching costs² IV. Zero marginal costs

V. Constant number of consumers in the market VI. No price discrimination

4.1.2 Undercut-proof property (UPP)

From a Nash-Bertrand equilibrium, an important property could be obtained. This property is called the Undercut-proof property (UPP) and explains how suppliers consider switching costs in their price setting. The property will be described in detail henceforth.

The non-negative pair of prices 〈𝑝_𝐴, 𝑝_𝐵〉 is the Nash-Bertrand equilibrium in price competition.

Both suppliers choose their prices, given the other supplier’s price, to maximize their profits.

Supplier 𝐴 can set its price to maximum 𝑝_𝐴 = 𝑝_𝐵+ 𝑆, otherwise it will lose its 𝑁_𝐴 consumers and end up with zero consumers. Similarly, if supplier 𝐵 wants to keep its customer base of 𝑁_𝐵, it can set a maximum price of 𝑝_𝐵 = 𝑝_𝐴+ 𝑆. Such an equilibrium does not exist according to Shy. The problem is that these two price equations are inconsistent, and therefore, the price pair 〈𝑝_𝐴, 𝑝_𝐵〉 cannot be a Nash-Bertrand equilibrium. This can be illustrated with a contradiction proof:

Proof 1. Supplier 𝐴 can set a higher price, 𝑝_𝐴^′ = 𝑝_𝐵+ 𝑆 = 𝑝_𝐴+ 2𝑆 without losing any of its consumers.³ In the same way, supplier 𝐵 can raise its price to 𝑝_𝐵^′ = 𝑝_𝐴 + 𝑆 = 𝑝_𝐵+ 2𝑆, and also keeping its consumers.⁴ Since 𝑝_𝐴^′ > 𝑝_𝐴 and 𝑝_𝐵^′ > 𝑝_𝐵, both suppliers could increase their

2 𝑆 is the same for both suppliers, no matter in which direction the consumer switches.

3 Note that 𝑝𝐵+ 𝑆 = 𝑝𝐴+ 2𝑆 since 𝑝𝐵 = 𝑝𝐴+ 𝑆.

4 Note that 𝑝𝐴+ 𝑆 = 𝑝_𝐵+ 2𝑆 since 𝑝_𝐴= 𝑝_𝐵+ 𝑆.

(23)

23 prices without losing any consumers. The supplier’s profit would then become larger and the pair of prices, 〈𝑝_𝐴, 𝑝_𝐵〉, cannot be a Nash-Bertrand equilibrium. ∎

Although the Nash-Bertrand equilibrium does not exist in these circumstances, there is one important property of that type of equilibrium that is satisfied in the present model, the UPP.

This property can be used for predicting the prices that the suppliers will set in situations where switching costs are present. By explaining the undercutting property with two different general definitions, the property of interest will be presented.

Definition 1. Supplier A is said to undercut supplier B, if it sets its price to 𝑝_𝐴 < 𝑝_𝐵− 𝑆, so that supplier 𝐴 “subsidizes” the switching costs for the consumers of supplier 𝐵. If supplier 𝐴 undercuts supplier 𝐵, then supplier 𝐴 will sell to all consumers according to equation (8), 𝑛_𝐴 = 𝑁_𝐴 + 𝑁_𝐵 and 𝑛_𝐵= 0. The opposite yields for supplier 𝐵, so if it undercuts supplier 𝐴, then it will get all consumers, while supplier 𝐴 will lose all its consumers. ∎

Definition 2. A pair of prices 〈𝑝_𝐴^𝑈, 𝑝_𝐵^𝑈〉⁵ satisfies the UPP if:

(a) Given 𝑝_𝐵^𝑈 and 𝑛_𝐵^𝑈, supplier 𝐴 chooses the highest price, 𝑝_𝐴^𝑈, subject to:

𝜋_𝐵^𝑈 = 𝑝_𝐵^𝑈𝑛_𝐵^𝑈 ≥ (𝑝_𝐴 − 𝑆)(𝑁_𝐴+ 𝑁_𝐵) (10) (b) Given 𝑝_𝐴^𝑈 and 𝑛_𝐴^𝑈, supplier 𝐵 chooses the highest price, 𝑝_𝐵^𝑈, subject to:

𝜋_𝐴^𝑈 = 𝑝_𝐴^𝑈𝑛_𝐴^𝑈 ≥ (𝑝_𝐵− 𝑆)(𝑁_𝐴+ 𝑁_𝐵) (11) (c) The distribution of the consumers, 𝑛_𝐴^𝑈 and 𝑛_𝐵^𝑈, is determined in equation (8). ∎

At the price pair 〈𝑝_𝐴^𝑈, 𝑝_𝐵^𝑈〉, both supplier A and B get the highest possible profits. Neither of the suppliers would be able to increase their profits by increasing or decreasing their prices. As a consequent, none of the suppliers will find it profitable to undercut the other supplier. It means that the inequalities (10) and (11) must bind. i.e. they can be treated as equalities. When no supplier finds it profitable to undercut the other, the suppliers’ customer bases will remain constant, so that 𝑛_𝐴^𝑈 = 𝑁_𝐴 and 𝑛_𝐵^𝑈 = 𝑁_𝐵. It is then possible to solve for 𝑝_𝐴^𝑈 and 𝑝_𝐵^𝑈:

𝑝_𝐴^𝑈 =^(𝑁^𝐴^+𝑁^𝐵^)(𝑁^𝐴^+2𝑁^𝐵^)𝑆

(𝑁_𝐴)²+𝑁_𝐴𝑁_𝐵+(𝑁_𝐵)²

𝑝_𝐵^𝑈 =^(𝑁^𝐴^+𝑁^𝐵^)(2𝑁^𝐴^+𝑁^𝐵^)𝑆

(𝑁_𝐴)²+𝑁_𝐴𝑁_𝐵+(𝑁_𝐵)² (12)

Some important conclusions could be drawn from this. By substituting 𝑝_𝐴^𝑈 and 𝑝_𝐵^𝑈 from equation (12) into equation (8), one obtains 𝑛_𝐴^𝑈 = 𝑁_𝐴 and 𝑛_𝐵^𝑈 = 𝑁_𝐵. This means that the

5 Note that the superscript U indicates that it satisfies the Undercut-proof property.

(24)

24 suppliers set as high prices as possible, given that the other supplier does not find it profitable to undercut. When the suppliers’ prices satisfy the UPP, they will keep selling to the same consumers as in the previous time period, 𝑡 − 1, which explain why 𝑛_𝐴^𝑈 = 𝑁_𝐴 and 𝑛_𝐵^𝑈 = 𝑁_𝐵. In Figure 3, the concept of the Undercut-proof equilibrium (UPE) is illustrated graphically (Morgan & Shy 2000). It is a further extension of the UPP and shows how both suppliers are constrained in their price setting, so that the other supplier cannot benefit from undercutting. At the point where the two lines cross each other, at 𝑝_𝐴^𝑈 and 𝑝_𝐵^𝑈, they maximize their prices without any of them finding it profitable to undercut the other. In that specific point, where both suppliers satisfy UPP, the UPE is located.

Figure 3. Undercut-proof equilibrium

To move on, one can assume more than two suppliers, having 𝐼 different suppliers. Each supplier is individually indexed by 𝑖 = 1, 2, … , 𝐼 and set its price to 𝑝_𝑖, providing 𝑁_𝑖 consumers.

Suppose furthermore that the suppliers are indexed in order of their market shares, so that supplier 1 has the largest market share and supplier 𝐼 has the smallest one, 𝑁₁ > 𝑁₂ > ⋯ > 𝑁_𝐼. By assuming that the supplier with the smallest market share has the greatest incentive to undercut the other suppliers and gaining new consumers, one can suppose that the other suppliers will set their prices so that supplier 𝐼 not will undercut them. This gives that all the other suppliers, 𝑖 ≠ 𝐼, will maximize 𝑝_𝑖 such that:

𝜋_𝐼^𝑈 = 𝑝_𝐼^𝑈𝑁_𝐼 ≥ (𝑝_𝑖− 𝑆_𝑖)(𝑁_𝑖+ 𝑁_𝐼) (13)

Similarly, supplier 𝐼 assumes that it is the primary target of supplier 1, the supplier with the largest market share. Hence, it sets its price, 𝑝_𝐼, such that supplier 1 finds it unprofitable to undercut supplier 𝐼:

𝜋₁^𝑈 = 𝑝₁^𝑈𝑁₁ ≥ (𝑝_𝐼− 𝑆_𝐼)(𝑁₁+ 𝑁_𝐼) (14)

(25)

25 As prices and market shares are directly observable, one can obtain the unobservable switching costs by solving the equations (13) and (14) in the equality case, giving:

𝑆_𝑖 = 𝑝_𝑖^𝑈 − ^𝑁^𝐼^𝑝^𝐼^𝑈

𝑁_𝑖+𝑁𝐼

𝑆_𝐼 = 𝑝_𝐼^𝑈− ^𝑁¹^𝑝¹^𝑈

𝑁1+𝑁𝐼 (15)

Equation (15) is the formula that Shy (2002) proposes to estimate the implicit switching costs.

It is important to remember that the switching costs are individual in real life, depending on the specific consumer’s preferences, especially when it comes to the non-monetary parts. For instance, one consumer might value their time higher than another. Similarly, one consumer might be more risk averse, so that it will be more likely to avoid the uncertainty of a new supplier compared to another consumer. Therefore, one should recall that the estimated switching costs obtained by Shy’s model are estimations for a representative consumer, or more precisely for the average consumer.

4.1.3 Criticism to Shy’s model

Shy (2002) assumes that all suppliers except for the one with the smallest market share, set their prices so that the smallest supplier cannot undercut it. This is an assumption that builds on the fact that the smallest supplier should be the most aggressive to gain new consumers. However, this must not be the case in reality. Instead, it should be a more reasonable assumption that a supplier sets its price so that no other supplier will undercut them. This implies that supplier 𝑖 maximizes its price, 𝑝_𝑖, given:

𝜋_𝑗^𝑈 = 𝑝_𝑗^𝑈𝑁_𝑗 ≥ (𝑝_𝑖 − 𝑆_𝑖)(𝑁_𝑖 + 𝑁_𝑗) (16) Note that 𝑖 ≠ 𝑗. By solving for the switching cost for supplier 𝑖, 𝑆_𝑖, one will obtain:

𝑆_𝑖 = 𝑝_𝑖^𝑈 − ^𝑁^𝑗^𝑝^𝑗

𝑈

𝑁𝑖+𝑁𝑗 (17)

This gives the same formula as Shy proposed, but in a more generalized way. One can proceed by looking at an actual example of when this generalization of Shy´s model performs better.

Example 1. Suppose three suppliers in the market, where every supplier has a marginal cost of zero. The mission is to solve the price of supplier 1, given following market shares and prices:

Supplier 1: 𝑁₁ = 20 and 𝑝₁^𝑈 = ? Supplier 2: 𝑁₂ = 18 and 𝑝₂^𝑈 = 12 Supplier 3: 𝑁₃ = 16 and 𝑝₃^𝑈 = 14

(26)

26 Furthermore, one can assume that the consumers that recently purchased from supplier 1, face a switching cost of 10, 𝑆₁ = 10. Then one can try to solve for the price that supplier 1 takes without being undercut by any of the other two suppliers. By using the generalized formula derived above, see equation (17), one has:

𝑆₁ = 𝑝₁^𝑈 − ^𝑁^𝑗^𝑝^𝑗

𝑈

𝑁1+𝑁_𝑗 for 𝑗 = 2, 3

Then, it is possible to rewrite this equation and solve for 𝑝₁^𝑈: 𝑝₁^𝑈 = ^𝑁^𝑗^𝑝^𝑗

𝑈

𝑁1+𝑁_𝑗+ 𝑆₁ for 𝑗 = 2, 3

By substituting the number of the example into the equation above, one will get:

𝑝₁^𝑈 =^18×12

20+18+ 10 ≈ 15,7 for 𝑗 = 2 𝑝₁^𝑈 =^16×14

20+16+ 10 ≈ 16,2 for 𝑗 = 3

This gives an answer to the maximum price that supplier 1 can set, without being undercut by supplier 2 or 3. From this example, one can see that the lowest price out of the two, i.e. the price that limits the price setting for supplier 1, is the equation for 𝑗 = 2. Hence, supplier 2 is the one that limits supplier 1 and its price setting.⁶ This shows how Shy´s model could overestimate the switching costs, since it would say that the price of supplier 1 is obtained by the equation for 𝑗 = 3. In more general terms, this proves that it must not necessarily be the supplier with the smallest market share that limits the other suppliers in their price setting. ∎ Furthermore, in the model that Shy (2002) presents, one of the underlying assumptions is that the switching costs only take positive values. In the beginning of section 4.1.1, 𝑆 was assumed to be positive, 𝑆 > 0. However, it should not be impossible that the switching costs take negative values in practice, which would imply that the switch of supplier would increase the consumers utility. This could be a result of some consumers that may be dissatisfied with their current supplier or those who can gain a lot in terms of price differences.

Another problem that should be addressed is the assumption of zero marginal costs, 𝑀𝐶 = 0. For most products, including electricity, this assumption is quite unrealistic.

To include marginal costs in the model would be easily done, but then it would include a variable that is difficult to collect data for. That may very well be the reason to why Shy chooses to exclude that variable, simply by assuming no marginal costs for all the suppliers.

6 The smallest 𝑝1𝑈 is the price that limits supplier 1’s price setting. Since 𝑝1𝑈 for 𝑗 = 2 < 𝑝1𝑈 for 𝑗 = 3, supplier 2 is the one that limits the price setting of supplier 1.

Estimating Switching Costs in Electricity Retailing