5. Results
5.1 Estimating switching costs using Shy’s model
Figure 10: Switching costs of consumers calculated using Shy’s model.
The blue graph shows that consumers have varying switching costs, but are all in an interval of 10%.
The switching costs of consumers of the largest banks seem to vary less than the switching costs of consumers of the smaller banks. Generally it seems as the largest banks serve consumers with the highest switching costs. The same picture can be seen if the market where only the six largest banks compete is considered. So according to the model, the largest banks serve the consumers with the highest costs of switching; similarly does the smallest banks serve the consumers with the lowest costs of switching.
From Figure 10 it is obvious that the two definitions of the market on which the banks operates yields very different results. Intuitively this seems reasonable, as the banks’ pricing methods should depend heavily on the firm5 they compete with or in the model framework, the one firm they fear will undercut them. The smallest firm in the market with the six largest banks have retail customer
5 Each firm only fears being undercut by one firm according to Definition 3.
deposits, which is used to calculate market shares, that is 250 times larger than the smallest firm in the model with all banks, so it is not surprising that the results are significantly different. The most obvious difference between the two models is the estimation of the switching costs of consumers of bank five and six. The main reason is that they go from being large banks in the model with all banks, and does not have an incentive to undercut the other banks, to being small banks in the model with only six banks and therefore have a larger incentive to undercut the other banks. The overall result that larger banks serve the customers with the largest switching costs, are equivalent for both market definitions.
Measuring the switching costs, as an interest rate can be hard to put in perspective, since the actual costs also depends on the unobserved principal and number of payments. Using an interest rate as the unit of measure can also be misrepresentative since switching costs is a onetime cost, while an interest rate is usually associated with repeated payments. It can however be related to the observed price of the bank. If the switching costs are measured as a percentage of the bank’s price, the
switching costs’ dependency on the principal and number of payments is eliminated. The measure is also independent of the price level and can be used if the price of the bank is not measured as an interest rate. The graph below illustrates the result.
Figure 11: Switching costs calculated using Shy’s model expressed as a percentage of the
price.
Figure 11 shows that in the model with all banks, the largest banks consumers’ switching costs are essentially the prices that the individual banks offer. This ratio decreases for the smaller banks, but the consumers’ switching costs are still above 50% of each bank’s price. This seems very high compared to other research, as well as general knowledge about markets. That the banks with the largest market share serve the costumers with the largest switching costs makes sense, as having consumers with high switching costs will for obvious reasons often result in a high market share.
Shy’s definition of market competition is not appropriate to use on markets where there are large differences in the size of the competitors, and therefore large differences in the competition between the individual competitors. The results of the model are highly dependent on the smallest firm in the market, as the smallest firm is the one most likely to undercut the price of the other firms. The other firms in the market are therefore essentially setting their price according to the smallest firm in the
market. If the smallest firm is significantly smaller than the larger firms, it is unrealistic that the larger firms will price according to the smallest firm, and the assumption is not suitable. The
problem arises because of the utility function of consumers and subsequently the demand structure of firms. The market shares are a function of the prices, and if a single firm undercuts the price of the others, and compensate the consumer’s switching costs, the firm will capture the entire market share. This is a simplified assumption, and is not in line with empirical observations. Instead of modifying the market assumptions, the demand structure of the model could be modified or the consumers switching costs are allowed to be asymmetric. For example could construction of a model where each firm face downward-‐sloping demand curves be constructed. In this thesis the underlying theoretical assumption of the model are intact, while the market assumptions are modified.
Two different modifications to Shy’s market conditions will be imposed. The purpose is to examine, if a more suitable assumption for the Danish banking industry can be found. The general idea is that the firms in the market do not set their prices according to the smallest firm in the market, but rather the firms with market shares similar to themselves. The assumptions are meant as a benchmark, and are not necessarily an improvement, or more correct. The first modification is that each bank only considers undercutting bank 𝑖+1, still using an index where banks are ranked by decreasing market share, i.e. each bank only consider undercutting the one bank with higher market share just above itself. In the second modification a moving average of both market shares and prices for bank
𝑖+3,𝑖+2,𝑖+1,𝑖−1,𝑖−2,𝑖−3, i.e. the three banks with market share just above or below bank 𝑖, is used as input in equation (4.10). For the three largest and smallest banks, where some of these do not exist, a moving average of the available banks are used. Below is a graphical presentation of the results.
Figure 12: Switching costs calculated using Shy’s model with modified market conditions.
The two modified estimations of switching costs are quite similar. They are both lower than the estimations using Shy’s assumption of undercutting on the market consisting of all banks. So if markets are defined as above, with different competitors for each bank, the switching cost of the consumers they serve are estimated to be lower. The market with the six largest banks cross both the new lines, implying that the switching cost variation of consumers is higher than in the two benchmark cases. Overall it seems that the four models yield estimated consumer switching costs with similar characteristics, but of different magnitude. Below is the switching costs given as a percentage of the each banks price illustrated.
Figure 13: Switching costs calculated using Shy’s model with modified market conditions
expressed as a percentage of the price.
The two estimations using Shy’s market assumption are quite different than the estimations with modified assumptions. Shy identify one bank that is most likely to undercut the others and the rest set their price relative to that bank, which usually leads to continuously decreasing switching costs of consumers. In the benchmark models the bank or banks that are most likely to undercut are different for each bank, which leads to switching costs of consumers that vary for each bank, and there is no uniform trend across the entire market. The tendency that large banks, serve consumers with high switching costs, does also seem to be existent with the modified market conditions.
The model is very simplified as it only considers prices and market shares, and makes some strong assumptions about which firms that potentially undercut each other. It is however an easy method to assess the level of consumer switching costs of each supplier in a market. Another disadvantage of the model is that switching costs are estimated on the basis of observed prices and market shares.
Prices and market shares will always be historical, and so will the consumer switching costs. This may not be an issue if switching costs of consumers are fairly constant over time, which is an interesting topic for further research.