Supervisor: Johan Stennek Dental Care on Equal Terms?

(1)

Graduate School

Master of Science in Economics Master of Science in Finance Master Degree Project No. 2012:45

Supervisor: Johan Stennek Dental Care on Equal Terms?

An empirical study of price levels, consumption and competition on the Swedish dental care market

My Alnebratt and Sara Lyxell

(2)

Dental Care at Equal Terms?

An Empirical Study of Price Levels, Consumption and Competition at the Dental Care Market

Abstract

The Swedish dental care law states as an objective that dental care shall be easily accessible and given on equal conditions for the entire population. In this study we use data on consumption and actual prices charged for six different dental treatments for the years 2009- 2011, collected by The Swedish Social Insurance Agency, to investigate if this is the case.

Results show that dental care consumption differ significantly, where inhabitants in Småland and Stockholm consume about 80 percent more standard examinations than inhabitants in Norrland. We further show that there is non-trivial price disparity between areas; median prices of the most common treatment package differ as much as 30% between the areas at the 10th and 90th percentile. We also find strong evidence for competitive distortions in the market as price increases with the number of producers. Lastly, we find that the public dental services, whose prices are set by politicians at each County Council, have a price leading role in each county, which suggests that there is in fact a great deal of competition between private and public.

(3)

Acknowledgements

First of all we would like to thank our advisor, Professor Johan Stennek, for his great commitment and interest in our work. His ideas, comments and support have been invaluable to this study. We would also like to extend a warm thank you to Florin Maican who has put several hours of his valuable time into giving us highly qualified advice on the econometrics and methodology of this thesis.

Further, we owe much gratitude to Anna Svensson, Mikael Moutakis, Douglas Lundin and Lars Sjödin at The Dental and Pharmaceutical Benefits Agency, TLV, who offered valuable advice in the early stages of our work.

Finally, we would like to thank several people that have helped us greatly in providing the necessary data. Firstly, Magdalena Kubien and Fredrik Lindström at The Swedish Social Insurance Agency (Försäkringskassan) who was very helpful in providing us with our primary data. Secondly we are very greatful to Björn Gerdén at Experian Marketing Services, Lotta Leván at Cegedim Sweden AB, Johan Jakobsson at Postnummerservice Norden AB and Kerstin Wilson and Eva Alfredsson at Praktikertjänst AB for providing us with complementary data free of charge.

(4)

1. Introduction and Background

1.1 Introduction

Since the dental care reform in 2008, the Swedish government has increased its spending on dental care with 50 percent, to about five billion dollar a year (SSIA, 2012a). Dental care is also regulated by law, The Swedish Dental Services Act (1985:125), which states as objectives that dental care shall be easily accessible (3§) and given on equal terms to the entire population (2§). Despite this, little research has been done on dental care consumption, price levels, and the functioning of the market in Sweden.

In this study, we investigate this, and show that there are significant inequalities. Firstly, we can show that there is major disparity in consumption between different parts of the country.

Inhabitants in Småland and the Stockholm area consume almost 80 percent more standard examinations than inhabitants in the northernmost part of the country. This finding makes it important to study underlying reasons for the disparity, where we can see that population density is an important determinant of consumption. We can further show that also prices differ significantly. This could be seen as a violation of the law stating that dental care should be given on equal terms to the entire population.

Furthermore, we find strong evidence for competitive distortions in the market as price, in contradiction to predictions by general economic theory, increases with the number of producers. This suggests that dentists have the ability to charge a price above marginal costs and that an additional dentist will not increase price competition. The lack of competition is likely to be due to large information asymmetries in the market where patients have considerably less knowledge about both quality and price, than the dentists. These asymmetries make it difficult for patients to gather information and compare different practitioners. Better informed patients would therefore strengthen the consumer’s position on the markets which could increase competition in the sense that equilibrium prices would fall.

Finally, we can show that while a higher number of producers do not increase price competition significantly, the public dental services do in fact have a great impact on the overall price level on a market, implying that the market in a sense is more price competitive than previously assumed. A one percent increase in the public dental service price of a specific treatment will result in a 0.3 to 0.7 percent increase in the median price of private

(7)

firms in the same area. So while private dentists are able to charge higher prices than the public dental services, they are none the less constrained by the prices charged by the public.

This result is important as it implies that the prices set for the public dental services by politicians at each county council have greater effects than what would be expected at first sight. The differences are leveraged through its price leading effect on the overall market, helping create and uphold price differences across the country.

1.2 Aim and Contribution

The combination of government subsidies and free pricing makes the dental care market special. The market further distinguishes itself as public providers and private firms are competing with each other, yet their prices are set very differently. Private firms must be assumed to profit maximize while public practices do not have this agenda. Furthermore, the market is characterized by information asymmetries between patients and dentists which are likely to distort competition. (Konkurrensverket, 2004) All of these features make it interesting to study the market both from an efficiency as well as an equality perspective.

Significant variation in price can exist due to differences in demand and supply, as well as being the result of competitive distortions and inefficiencies in different markets (Cabral, 2000). As the dental care law states equality as an objective it is important to examine whether such differences exist and their cause.

The aim of this study is therefore to answer the following questions:

• Are there non trivial disparities in consumption and price in Swedish dental care, and how can they be explained?

• Are there significant competitive distortions which affect consumption and price in the Swedish dental care market?

We answer these questions by analyzing data on prices and consumption for four different treatment packages during the years 2009 to 2011, collected by the Swedish Social Insurance Agency (SSIA). Our data allows for a study which differs from previous studies analyzing prices and consumption on the Swedish dental care market as it contains information on the whole population. Second, our data on price is based on actual prices charged, as opposed to price list prices, which are likely to better reflect the conditions in the market. We further extend earlier research by including additional variables in our models, controlling for both

(8)

3 1. Introduction and Background

variation in market structure and demand. In addition, our study addresses the major concern of endogeneity when investigating price leadership, a problem which has been ignored by previous research. Lastly, no study of the price and consumption differences has been conducted since the introduction of the new state dental care financial support system in 2008, why this study provides valuable contribution.

The paper is organized as follows. Section one includes background and theoretical discussion. Section two present the data. Section three reports the results which show differences in consumption, as well as, the effects of demand, market structure and price leadership on price. Section four concludes.

1.3 The Swedish Dental Care Market - Deregulation, Subsidies and Information Asymmetries

As mentioned above, the Swedish dental care market is special in many ways. Thus, in order to understand the current issues it is important to have a grasp of the functioning of the market. In this section we aim to provide this, both in a regulatory as well as microeconomic point of view.

Market Structure

The dental care market is dominated by the public dental services, managed and run by each of Sweden’s 21 county councils, most often within the scope of the county administration.

About 40 percent of all adults, and 95 to 98 percent of all children, receive their dental care from the public dental services. (Folktandvården, 2012) The market shares vary greatly geographically however; in the cities private firms have a larger share while the public dental services more or less have monopoly on some smaller markets in sparsely populated areas.

(SOU 2007:19) There are about 7220 authorized dentists, of which 39 percent work for the public dental services, 47 percent at a private firm and 13 percent at specialist practices which can be either private or run by the county councils (Cegedim AB, 2012). The largest private player is Praktikertjänst AB, which was founded in 1977 and is run as a profit-driven limited company in a producer cooperative frame. Their shareholders are the dentists who are at the same time entrepreneurs running their own practices. Their market share is 30 percent of adult dental care and 20 percent of all dental care. (Praktikertjänst, 2010) Besides the public dental services and Praktikertjänst, the market is highly fragmented with mainly independent clinics.

(SOU 2007:19)

(9)

Until 1999, there was a price ceiling on dental care, which in practice resulted in all dentists charging the same price. The reason for the price deregulation was that the government was worried that the price ceiling was serving as a focal point for implicit price collusion, and it was believed that a removal would increase competition and consequently decrease prices.

(The Swedish Government, 1998) However, later studies confirm that this was a misconception as demand, although price elastic, did not contribute to increase price competition. Consequently, prices rose significantly with up to 16 percent during the first year of free pricing. The price increases further continued and between January 1998 and January 2006 the total price change was estimated to 71 percent. (SOU 2006:27) This further underscores the importance of studying this market as it could diminish misconceptions and help future decision making.

State Dental Care Financial Support System

The current state dental care financial support system was introduced in July 1, 2008. The reform had three main aims which were (1) maintaining good dental health for individuals with minor dental care needs, (2) making it possible for individuals with high dental care needs to get dental care at a reasonable cost, and (3) strengthening the patients’ position on the market. The third aim was to be fulfilled by a price comparison web page and a reference price list. The price comparison web page was introduced at the SSIA home page, and the idea was that dental practices would register their prices for different treatments. The site was a failure however, since dentists, despite the law requiring it, neglected to register their prices.

The page was subsequently closed but was previously (20 Jan 2012) reopened as a part of the health care information center web site, 1177.se. To date however, the problem seems to remain; few practices apart from the public dental services have registered their prices. The reference price list has two purposes, the first to set the compensation levels for dental care subsidies, but secondly to serve as another tool for price comparison. The list is set up and updated regularly by The Dental and Pharmaceutical Benefits Agency (TLV). (SSIA, 2011) The idea is that the first and second objectives will be achieved by a general dental care subsidy encouraging preventive action and regular visits, and by a high-cost protection scheme reducing the financial barriers to the consumption of necessary dental care. The general dental care subsidy is a benefit available to all adults from the age of 20 on a yearly basis. The subsidy is subtracted from the fee charged to the patient at the point of payment, and after reporting to the SSIA the amount is then reimbursed to the dentist. The grant is 300

(10)

SEK for patients aged 20-29 and 75+, and 150 SEK for patients aged 30-74, and can be saved for one year (a patient seeing a dentist every second year will thus be entitled to a 600/300 SEK grant). (SSIA, 2011)

The high-cost protection scheme has a threshold of 3000 SEK. It is based on the reference prices set by TLV, and not on the actual prices charged by dentists. This means that an increase in the dentist’s price - as long as it is higher than the reference price - is fully transferred to the patient whether he reaches the threshold for the high-cost protection or not.

Costs with a reference price above the initial threshold of 3000 SEK but below 15 000 SEK is subsidized with 50 percent of the reference price, while the share of the costs with a reference price above 15 000 SEK is subsidized with 85 percent. (SSIA, 2011)

Competition on the Dental Care Market

While one of the objectives of the deregulation of the dental care market in 1999 was to strengthen competition, there are still many distortions. On the demand side there are issues of information asymmetries which will be discussed in the next section. On the supply side, the competitive conditions between private and public producers are much-disputed, and both sides argue that there are competitive disadvantages on their part.

In order to have a well-functioning competitive market where one of the main producers is the public sector, it is vital that there are fair conditions between public and private players. If this is not the case, the competition may be distorted, which in turn may result in inefficiencies.

(SOU 2007:19) However, the conditions for the public dental services and private dental care producers differ fundamentally, due to regulatory reasons as well as differences in objectives.

According to the Dental Services Act, the county councils have special responsibilities such as planning the provision of dental care and offering good dental care for those who are inhabited in the county. Further, they are responsible for children’s dental care, specialist dental care and dental care for certain groups such as disabled and others with special needs.

The effects of such different regulatory conditions are distortions that are hard to both measure and compensate for. Examples of how this disfavors the public dental services are pointed out by the Swedish Association of Local Authorities and Regions (SKL) in a Government Official Report (SOU 2007:19). Firstly, public dental services may be disadvantaged as they are to follow the rules that apply to public administration. That is, the practices are subsumed conditions that aim to optimize overall county council operations, but

(11)

not necessarily the dental care. Secondly, the accessibility requirements entails running clinics in sparsely populated areas where the customer base is not large enough to cover costs, as well as to provide costly emergency dental care and premises suitable for disabled. Further, while private firms may turn down unprofitable patients or treatments, the public dental services have the ultimate responsibility to provide dental care to all its inhabitants.

However, in the same report, Privattandläkarna, the interest organization for private dental service firms, points out several competitive disadvantages met by their members. Some of them are the disadvantage of the public dental services having such a large part of the children’s dental care, which means many patients stay on as they move on to adult dental care. The county councils also have the ability to write off deficits, which gives them an advantage over private practices which must cover its costs in order to survive in the long run.

Thus the direction of the competitive distortion is somewhat unclear, and several studies point to the need for further regulations in order to completely even out competition. (SOU 2007:19)

Pricing on the Dental Care Market

A second fundamental difference between public dental services and private firms are their objectives. While the latter must be assumed to maximize profits the former could be said to maximize social welfare under given circumstances. This of course means that the premises for price setting differ. Private firms would, as all firms, seek to set prices that maximize profits; that is quantity times price minus costs. Doing so means taking many factors such as the price elasticity of demand, the level of market power, and the prices of competitors into account. The prices of the public dental services on the other hand, are decided upon by each county council, and according to the Local Government Act, they must be based on costs.

This cost based pricing is not always straightforward however, and while some counties cross-subsidize between different treatments, others have cost-based pricing for each treatment. Further, the Local Government Act (Ds 2004:31) also enacts an equality principle meaning that the same prices must be charged in the entire county. Since different practices may have entirely different costs due to differences in rents, productivity, demand etc, this means that there may exist another form of cross-subsidization within counties, i.e. one geographical part subsidizes another. Thus the price does not always reflect the cost of production at a certain location. (SOU 2007:19)

(12)

A cost based pricing means that the public dental services are not subsidized, which is a prerequisite for a well-functioning and fair competition. However, a report compiled by The Swedish Agency for Public Management (Statskontoret) (2008) reveals that counties generally have insufficient disclosure of specific costs related to the provision of dental care.

The disclosure also differs greatly due to different operational models. A few county councils have their dental services operated as corporations, which gives better disclosure. Some have separated their dental services in units with their own balance sheet, and some run it more traditionally as part of the public administrations. The report concludes that the disclosure is generally insufficient to determine whether the public dental services are being subsidized or not. Thus, it could be argued that in lack of sufficient disclosure, politicians do not have enough information to ensure that prices perfectly reflect cost. That would mean that there may be political or structural differences between counties in price setting.

Information Asymmetry

The Swedish dental care market also embodies competitive limitations on the demand side.

General economic theory often takes its standpoint in efficient markets with perfect competition; however, few if any such markets exist. Markets do not function perfectly because of different kinds of market failures. One such failure is asymmetric information where information about the good differs between the producer and consumer. (See for example Frank (2009) for a deeper discussion on perfect competition and market failures.) In a well cited paper George Akerlof (1970) discusses how asymmetric information about quality can lead to under provision of high quality goods. When a consumer is unable to distinguish between high and low quality the producers will be unable to charge the high quality price why high quality goods will not be sold. Akerlof discusses a situation where the consumers have no information about the quality. There might be many situations where the lack of information is less extreme, but still important for the market structure. Satterthwaite (1979) argues that for so-called reputation goods price increases and quality decreases as market size and participants increase. He describes reputation goods as goods where the consumer’s ability to determine the quality is dependent on the person’s possibility to gather information from other consumers who have already consumed the product, such as friends and neighbors. Since it is harder to obtain information about a particular dentist if he or she only serves a fraction of the market, this possibility decreases as market size increases, why competition is lower in larger markets.

(13)

However, even if the consumers, through gathering information, might be able to decrease the information asymmetries, they will still impact the way the market work. The Swedish dental care market today is affected by difficulties for the consumers when gathering information about both quality and price. Firstly, there is inadequate information and knowledge about the difference in price and quality between different care givers, and to what extent the dental care financial support system will cover the costs. It can further be difficult for the patient to evaluate different treatment alternatives and their prices in relation to quality. Lastly, even after the treatment it is not certain that the patient will be able to assess the quality of the care and the appropriateness of the material used. (SOU 2007:19) All of these information asymmetries can be seen as market failures which will affect the competition and functioning of the market.

Market Power and Demand

Market power is defined as the ability to charge a price above marginal cost, which in turn will lead to a production below the socially optimal quantity. The level of market power will, amongst other things, depend on the price elasticity of demand. (Cabral, 2000) The combined inference from the previous studies on dental care demand gives that although patients are sensitive to price, they do not compare prices of different care providers, but instead stay with the same dentist. Grönqvist (2012) analyses the price sensitivity of dental care consumption in Sweden and concludes that the demand for a standard treatment is price elastic. Further, in a survey performed by FSI (Forskningsgruppen för Samhälls- och informationsstudier, 2005), 84 percent of respondents answered that they would not change to another dentist within reasonable distance to get a lower price. The foremost reason for this was the great trust they had for their dentist. These results indicate that the dentists will have market power as soon as trust is established. The price would therefore not affect the choice of dentist, but would still affect the choice of whether or not to consume dental care. These specific market conditions make it important to look at how and why price differs within the country, in order to evaluate if there is further evidence for market power.

(14)

9 2. Data

2. Data

In order to answer the research questions stated above, we will firstly look at the prevalence of consumption differences and their determinants, by regressing consumption on a set of demographic explanatory variables. Secondly, we will map price disparity and regress price on a set of demographic variables as well as variables describing market structure, allowing us to draw conclusions about the functioning of the market. Lastly, we will look at the interaction between different players on the market, by distinguishing between the public dental service and private firm prices and regressing the latter on the former together with a set of control variables.

2.1 Data description

Our price and consumption data covers Sweden over the period 2009-2011. Table 1 details the data sources and the variable definitions for the data used in our analysis.

Prices and Consumption

Our primary data are actual prices charged on dental care, reported to the SSIA, for the years 2009 to 2011, together with the number of treatments performed. The data is reported in connection with dentists requesting reimbursement from the dental care financial support system; either the general dental care subsidy or the high-cost protection scheme. While the latter is applicable for treatments exceeding 3000 SEK, it is aggregated and thus all treatments are usually reported.

Due to the Public Access to Information and Secrecy Act (2009:400), SSIA is not able to provide us with data on an individual level, in this case prices per firm or practice. Instead we are given aggregate data on the lowest level possible, which is two digit postal code area. It could be argued that municipality level would have been a more reasonable level; however, the SSIA records do not contain any information on municipality affiliation. There are 83 two digit postal code areas in Sweden with a large dispersion in size; the largest being the northernmost area (two digit postal code 98) at 47 000 km², and the smallest being central Stockholm (two digit postal code 11) at 51 km². The dispersion in population is significantly lower, however, with a range from 32 000 to 317 000 inhabitants, and with the largest areas generally having the lowest population. An overview of the areas with location, size and population as well as a map is given in Appendix 1 and 2.

(15)

Table 1. Definitions and data sources

Variable Mean (Std. dev)

where applicable Definition and Source Variables Testing for Consumption Differences

TR./1000 241 (48.7) Number of treatment 101 performed per thousand inhabitants, each year in each postal code area. Source: SSIA and Statistics Sweden.

Variables Testing for Price Differences

PRICE 101: 668 (63.5)

202: 151 (42.9) 501: 2502 (293) 705: 1040 (72.3) 821: 67.2 (89.26) 824: 9194 (357)

Median price at each postal code area level, for either all public or all private practices.

Source: SSIA

d.PUBLIC Dummy variable taking the value one if the

median price refers to the median public price and zero if it refers to the median private price.

d.BIGCITY Dummy variable taking the value one if the

median price refers to an area within a big city, and zero otherwise.

d.TR.X Dummy variables taking the value one if the

price refers to treatment X, where X refers to any of the six different treatments 101, 202, 501, 705, 821 and 824, and zero otherwise.

d.YEAR_X Time dummy variables taking the value one

if the price refers to year X, where X refers to any of the years 2009-2011, and zero otherwise.

DT/100 000 73.0 (26.7) Number of dentists per 100 000 inhabitants.

Source: Cegedim Sweden AB

PTJ/PRIVATE 0.43 (0.21) Share of private dentists in an area affiliated to Praktikertjänst. Source: Praktikertjänst AB

DT/PRACTICE 2.67 (2.44) Average number of dentists at each practice, in either the private or public submarket.

Source: SSIA and Cegedim Sweden AB

(16)

11 2. Data

Variables Testing for Price Leadership

PUBLIC_PRICE Median price at postal code area level for

public practices. Due to endogeneity problems this variable is estimated with instrumental variables which are dummies for political color of the county council (Blue, Red or Mixed), county tax rate, and operational model. Source: SSIA, SKL, SCB

PRIVATE_PRICE Median price at postal code area level for

private practices. Source: SSIA.

d.TR.X Dummy variables taking the value one if the

price refers to treatment X, where X refers to any of the six different treatments 101, 202, 501, 705, 821 and 824, and zero otherwise.

d.YEAR_X Time dummy variables taking the value one

if the price refers to year X, where X refers to any of the years 2010-2011, and zero otherwise.

Demographic control variables

POP.DENSITY 268 (833) Population density, measured by the ratio of population to square kilometers, in the postal code area. Source: Statistics Sweden and Postnummerservice Norden AB

INCOME/CAP. 192 859 (22 320) Per capita income in 2011 in the postal code area. Source: Experian Marketing Services ETHNICITY 0.11 (0.056) Share of the population born outside of

Sweden, in the postal code area. Source:

Experian Marketing Services

EDUCATION 0.28 (0.093) Share of the population with post-secondary school education, in the postal code area.

Source: Experian Marketing Services

For each two digit postal code area our dataset contains the median price, as well as dispersion, for six different treatments over three years (2009-2011), and divided upon the county council public dental service and private practice median prices. The six treatments are listed in Table 2 together with their frequency. Since a patient seldom does one treatment in isolation, but instead as part of a series of treatments, we have chosen our treatments on the basis of treatment packages. The treatment packages have been defined in collaboration with Lars Sjödin, dentist at TLV, using previously defined treatment packages as basis (SSIA,

(17)

2010). We have taken into account the frequency with which each treatment is performed, aiming at investigating commonly used treatments due to its higher relevance and accuracy.

Further, we have chosen packages where a relatively low share of the price is due to material costs, minimizing the risk that price differences is due to the usage of different types of material across the country. The packages are listed in Table 3. While packages 1- 3 are estimated to be used by all age groups, number 4 is used most frequently by the upper age groups.

Table 2. Treatments with description and frequencies Treatment

code Treatment Reference price

2011

Frequency Jan-Nov 2011

101 Standard examination by dentist 660 2 470 000

202 Tooth cleaning and flour treatment 125 1 038 000 501 Root canal filling, one root canal 3 180 98 000

705 Filling, two surfaces 1 025 1 057 000

821 Prefabricated tooth, unit price 55 120 000

824 Partial prosthesis 9 400 17 000

Source: SSIA, 2012b, TLV, 2011

Table 3. Treatment packages Treatment

package nr

Treatment

codes Content Reference

price 2011

1 101+202 Standard examination and cleaning 785

2 101+705 Standard examination and filling 1 685

3 101+501+705 Standard examination, root canal filling and

filling 4 865

4 101+(3*705) +(6*821)+824

Standard examination, filling, prefabricated teeth

and partial prosthesis 13 465

Source: TLV, 2011

Our prices are from actual prices charged, not from price lists. We argue that this is the most relevant data, as it is not unusual for dentists to give discounts or for other reasons charge a price different from the listed prices. It should be noted, however, that since we use medians, these values usually correspond to the actual price lists when it comes to the county council public dental services.

(18)

13 2. Data Markets

When data is not gathered on areas which can be considered independent markets it is important to be careful both in the model estimation and variable selection, as well as when making inferences. In order to establish what would be a correct market specification in this case we would have needed individual data on practices as well as the home address of customers, which is unavailable to us. While the sizes of the two digit postal code areas can be considered reasonable, since inhabitants in sparsely populated areas generally travel longer to see a dentist than inhabitants in cities (Fernberg and Ordell, 2004), it is highly likely that people travel across the borders of our areas. The way the postal code areas are divided make this even more likely; in some instances such as for Gävle, Sundsvall and Norrköping, the city has its own two digit postal code area, while it is surrounded on all sides by another two digit postal code area. In these instances it is highly likely that many people living in the surrounding area work in the city or regularly visit it, and they are thus likely to visit the dentist there. Further, in the areas close to any of the three larger cities it is also likely that people travel into the city to visit a dentist, for the same reasons as above. We will thus be careful in considering our two digit postal code areas markets. When regressing on price, this is not as big of an issue since the price variables are constructed out of medians in each area, as opposed to aggregate figures. However, the results will still be interpreted cautiously.

2.2 Variable specification

Population Density

The population in each postal code area has been obtained for 2009-2011 from Statistics Sweden, and is measured at the beginning of the year. Data on the size in km² of each area have been obtained from Postnummerservice Norden AB. This data have been used to construct the variable for population density, Ln(POP.DENSITY), which is measured as the number of inhabitants per square kilometer. The variable accounts for demand differences that may depend on variation in population density, i.e. a longer distance to travel could reduce demand, and thus price should increase with population density. On the other hand, population density could affect market structure in the sense that scale efficiencies increase as the consumer base increases, or that larger markets increase competition. Both these effects lead to prices decreasing with population density.

(19)

When investigating price disparity, we include a dummy variable called d.BIGCITY, taking the value one for areas in central Stockholm, Gothenburg and Malmoe. This to account for the disproportionately larger rents and other special effects present within the cities.

As have been described above, we need to be careful considering our areas to be markets, thus we will not use population as a variable. This is because the borders of the areas in this respect could be seen as more or less arbitrarily drawn. Thus, they could be redrawn without any loss of relevance; giving completely different values on population, why there is a risk of non-robust results if we use this variable.

Socioeconomic Variables

In order to control for differences in demand, we use data on income, ethnicity and education per postal code area. The data have been obtained from Experian Marketing Services, a marketing services company. From this data we construct Ln(INCOME/CAP) (income per capita) describing the average capital and employment personal income, EDUCATION describing the share of the population with post-secondary school education, and ETHNICITY which gives the share of the population born outside Sweden. These variables are included in order to account for differences in demand.

The effect of ETHNICITY on demand is uncertain. The extensive children’s dental services free of charge for individuals in Sweden could be expected to affect dental care demand in two ways. Firstly, it could promote a culture of visiting the dentist more regularly increasing demand for standard examinations. Secondly, after a good dental health has been established in a young age the demand for treating dental diseases on adults might decrease. Lastly, the effect of EDUCATION is also unclear. However, after controlling for the effects of income, a significant coefficient for this variable could imply that there are cultural differences related to education level.

Variables Testing for Differences in Dental Care Consumption

In the first section of our results, we investigate dental care consumption per capita. As a proxy for dental care consumption we use the number of standard examinations performed.

The variable TR/100 000 represents the number of standard examinations per 100 000 citizens performed by dentists at both public and private practices in each area.

(20)

15 2. Data

It is reasonable to assume that the majority of patients that consume any dental care will first do a standard examination. The rationale behind using this variable is that the standard examination is a dental health check whose main aim is to diagnose any dental issues, and not to any large extent fix possible dental health issues. Thus after deciding to take a standard examination, patients will generally go on to perform any additional treatments deemed necessary.

The issue of our areas not being well defined markets becomes more problematic when investigating dental care consumption, than when we regress on price. This since the variable describing treatments per population is unreliable due to patients systematically traveling across borders mainly from suburban areas into the cities. In order to solve this issue, we go through our data and delete the observations that appear to be biased through major outflows or inflows of patients: 16 areas in total.

Variables Testing for Price Disparity

In order to evaluate and explain price differences, we use the variable PRICE which is the median price at each postal code area for either all public or all private practices in an area, and for any of the years 2009-2011. In this regression model we thus treat the public firms’

median prices and the public dental services’ median prices equally. To control for any structural differences between the two we use the dummy variable d.PUBLIC, taking the value one if the price refers to a public dental service practice and zero otherwise. In order to account for differences between years and treatments we use dummy variables for each treatment and for each year (d.TR.202 etc., as well as d.YEAR09 etc.), omitting the most common treatment (101) and the most recent year (2011).

Apart from using the demographic control variables mentioned above, we construct three variables describing the dental care market structure differences between areas. The number of dentists in each area is used together with the population to construct the variable DT/100 000. The data on dentists have been obtained from Cegedim Sweden AB, a customer relationship management company within the health care industry. The data contains information on work place postal code as well as private or public affiliation for all qualified dentists in the country (7219 dentists). Further, we use a variable called PTJ/PRIVATE which is the number of dentists affiliated with Praktikertjänst to private dentists in the area. This data has been obtained from Praktikertjänst AB. For these variables, it is important to note that they are measures of individual dentists and not of practices in an area.

(21)

Lastly, DT/PRACTICE is the ratio of the number dentists in each area divided by the number of practices, for either the public or the private submarket. This variable thus describes the average size of each practice and is meant to capture effects of scale economies. The expectation is that the variable should have a positive effect on supply and thus a negative effect on price.

Variables Testing for Price Leadership

Our final regression entails testing for price leadership of the public dental services, by regressing the private median prices (PRIVATE_PRICE) on the public median prices (PUBLIC_PRICE). In this regression we thus divide the data on PRICE used in the previous regression into its private firm and public dental services components. We further include the dummy variables for each treatment and each year (d.YEAR09 etc. and d.TR.202 etc.) described above.

A major concern when studying the effect of public price setting is however the possibility of endogeneity within the model. This arises when public prices do not only affect private prices, but the causality is also reverse. In this model there is a possibility that public prices are indirectly affected by the market conditions and the prices set by private practices. For example, a lower price level among private practices in one area may put pressure on dentists and dental nurses to work more efficiently. This would lower the total costs of dental care in that county, which would, at least in theory, result in lowered prices. In the same way, if private firms charge low prices this could results in them having a relatively large part of the customers in a certain market. This would mean that the public dental service have fewer customers, which due to scale effects could mean that they will charge higher prices.

We circumvent the endogeneity problem through the use of instrumental variables (IV), correlated with public price but unrelated to private prices. The results of the IV regression will be reported together with the OLS results. The instruments used in the IV regression are political color, tax rate and operational model, which are likely to be correlated to the decision making in the county council and to the specific conditions in the county, and thereby also the dental care prices. The variable political color gives whether the county is governed by either the red and green (s-v-mp), the blue (m-fp-kd-c) or any combination across the block boundaries. A red government has a positive effect on price while a blue one has a negative effect. Tax rate is the county council tax rate, which has a mean of about 10,8, and a positive effect on price. The positive effect could be due to the fact that in counties with efficient

(22)

17 2. Data

county organizations, the tax rate can be kept lower, and as a result of the efficiency dental care is also cheaper. Finally operational model gives whether the public dental services in the county are run as a corporation or not – having it run as a corporation has a negative effect on price. This negative effect may also be due to higher efficiency in the public dental services run as corporations as opposed to the ones that are operated as part of the overall public administration. Further details on the IV estimation are given in Appendix 5.

When we regress Ln(PRIVATE_PRICE) on Ln(PUBLIC_PRICE) using OLS we include the variable Δ(PUBLIC_PRICE) which was introduced by Eriksson (2004) and represents the price change between two years. We calculate it from our primary data as the percentage change of the sum of county council prices for our six treatments between year 2009 and 2010 as well as 2010 and 2011. The rationale for using it is that it addresses the possibility that there are demand or cost differences that are common for both county councils and private firms, such as differences in wages, rent, income or education, which make them adjust prices in the same way. If this would be the case, prices for all treatments will vary in a similar way for both public and private firms in a county and a regression of Ln(PRIVATE_PRICE) on Ln(PUBLIC_PRICE) would be highly significant but it would not actually say anything about Ln(PUBLIC_PRICE)’s casual effect on Ln(PRIVATE_PRICE). To control for this possibility we thus include Δ(PUBLIC_PRICE), since a county council with high costs would increase all of its prices. Using it allows us to isolate the effect that a price increase in one treatment by the public dental services has on the private mean price of the same treatment.

By including it, we thus control for the alternative explanation of common cost or demand differences that leads private and public firms to adjust their prices in the same way.

However, when the regression is estimated using an instrumented variable for Ln(PUBLIC_PRICE) the Δ(PUBLIC_PRICE) is not included in the model. This as we do not expect general changes in demand or cost to affect the instrumental variables, which represent the political and organizational structure, and would therefore be redundant in this regression.

(23)

3. Results

3.1 Accessibility and Consumption

As mentioned above, the Swedish Dental Services Act states that dental care shall be easily accessible and given on equal terms for the entire population. While equal terms may entail both equal quality as well as prices, accessibility could be seen as both distances to the nearest dental care practice as well as the average waiting time for a visit, etc. However, we argue that an interesting starting point is to evaluate differences in consumption across the country.

While demand may vary across the country, it is not very likely that the actual dental care needs vary significantly. Thus, looking at consumption differences allows us to see the real effects of the current dental care policy and market dynamics.

Further, the government spending on dental care has increased drastically since the introduction of the dental care reform in 2008. With such high spending, it is interesting to evaluate consumption differences, as unequal consumption patterns would suggest that this major investment does not redound equally to the entire population.

In order to evaluate consumption differences, we look at the number of standard examinations performed by dentists per 1000 inhabitants. In Table 4 below, we have aggregated the consumption in our 83 postal code areas into ten larger geographical areas, which both minimizes the risk of patient inflow and outflow that would bias our results, and also provides a better overview of the situation. The areas are specified in Appendix 3.

It is clear that the number of standard examinations, which serves as a proxy for general dental care consumption, differ greatly between the areas. North Norrland is the area where the inhabitants consume the least dental care; only on average 161 standard examinations per 1000 inhabitants each year. This is followed by Värmland with 186 and South Norrland with 192. The inhabitants in Småland have the highest consumption, with on average 292 standard examinations per 1000 inhabitants each year, followed closely by the inhabitants in the Stockholm area and East Svealand with 289 and 288 respectively. The variation is striking;

the difference between the areas with the lowest and the highest average consumption is 81 percent. This means that if we assume that the average Swede visits the dentist for a standard examination once every second year (according to SOU 2007:19) a majority of patients have an interval of 1-2 years between examinations, while some go more seldom), then almost 60

(24)

19 3. Results

percent of the inhabitants in Småland see a dentist regularly while only just over 30 percent of the inhabitants in North Norrland do. While lifestyle and food culture differences may account for a slight difference in need, they are not likely to account for differences of this magnitude. Thus, it seems that dental care is not equally accessible, either economically or geograpically, to the entire population.

Table 4. Number of standard examinations per 1000 inhabitants

Area 2009 2010 2011 mean

1 Stockholm 289 292 284 289

2 Skåne 249 260 282 264

3 Småland 287 295 293 292

4 Upper west coast 260 256 264 260

5 North east Götaland 275 271 271 272

6 Värmland 187 189 182 186

62 Gotland 238 192 183 204

7 East Svealand 276 293 294 288

8 South Norrland 187 193 195 192

9 North Norrland 158 166 161 161

A clear pattern in the table above is that consumption seems to be positively related to population density, with the sparsely populated areas in the northern part of Sweden having the lowest consumption and vice versa. This leads us to the question of whether and how consumption varies with population density, as well as with other socioeconomic variables, such as income, ethnicity and education.

In order to investigate this further we use OLS to estimate TR/1000_𝑖 = β₀+ β_𝐷 𝐷𝑒𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑐𝑠_𝑖+ ε_𝑖

where TR/1000_𝑖 is the number of standard examinations performed in the ith area, and 𝐷𝑒𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑠_𝑖 is the matrix of our demographic control variables. Some of the independent variables are taken in logarithms, as many of them have both large and skewed dispersions.

The regression results are reported in Table 5. Due to collinearety issues, we obtain insignificant results for Ln(INCOME/CAP) when including Ln(POP.DENSITY), thus we include two different regressions.

(25)

Table 5. Regression results (standard errors in parentheses) TREATMENT/1000

(1) (2)

Ln(INCOME/CAP) -0.0960 (0.0617)

0.137**

(0.0617) Ln(POP.DENSITY) 0.0530***

(0.00845) Ln(POP.DENSITY)^2 -0.00481***

(0.0117)

ETHNICITY -0.0777

(0.152)

EDUCATION 0.211**

(0.0958)

Constant 1.23*

(0.740)

-1.44*

(0.751)

R2 ADJ 0.2916 0.0209

NOBS 184 184

***significant at the 1 percent level, **at the 5 percent level, and *at the 10 percent level.

In Column 1 of Table 5 we can see that the coefficients for Ln(POP.DENSITY) and Ln(POP.DENSITY)² are highly significant, with the first one being positive and the second one negative. This implies thatconsumption increases with population density, at a decreasing rate. This could imply that geographical accessibility influences consumption; in an area with a high population density inhabitants can be expected to have a lower average distance to a dental clinic, and will thus be more likely to visit it. While the county councils through the public dental services are obligated to offer dental care accessible to all inhabitants (in other words they must have clinics in sparsely populated areas despite the customer base being too small to cover costs) it is still likely that the average distances to the nearest clinic will differ significantly between the areas. While we cannot find any significant results for ETHNICITY, the coefficient for EDUCATION is positive and significant at the five percent level. The size of it indicates that a one percent increase in the share of inhabitants with post-secondary school education will be associated with a 0.002 unit increase in TREATMENT/1000; the effect is thus very small. The effect is not surprising but none the less interesting to have in mind as it is present after including income in the regression, indicating that it is not the higher income associated with a higher education that leads to increased dental care consumption.

(26)

21 3. Results

Looking at Column 2 in Table 5, we can see that the coefficients for Ln(INCOME/CAP) is significant at the five percent level. The sign of the coefficient imply that the consumption of standard examinations increases with income. The size of the coefficient is small however; a one percent increase in Ln(INCOME/CAP) results in a 0.0014 unit increase in TREATMENT/1000. The positive relation is an expected result; the higher the disposable income of an individual, the more likely he or she will be to consume an examination; in other words we observe a positive income elasticity on an aggregate level. The findings are also supported by Grönqvist (2012) who uses data on individual consumption to investigate the income elasticity closer. He finds that an increase in disposable income by one percent is associated with a 0.04 percent higher likelihood to see a dentist.

From the regressions above we can thus conclude that the large differences in dental care consumption can be partly explained mainly by population density and to a smaller extent by income per capita and education, where it seems inhabitants in areas with high population density, income and education level consume more dental care. However, our regression model only explains 29 percent of the variation, as shown by the Adjusted R-squared of 0.29.

Another factor that we hypothesize to affect consumption and thus would increase the explanatory power of the model is price. Grönqvist (2012) studies the price elasticity of demand in the Swedish dental care market and finds that demand is price elastic. He finds that a one percent increase in price decreases consumption of treatment 101 by 1.4 percent. For eight other treatments tested no statistically significant decrease in demand was found. The results may well reflect special demand conditions on the dental care market, where the consumer decides whether to consume dental care or not, instead of deciding of which amount he or she will consume. Thus the price of dental care seems to affect foremost the probability of visiting the dentist rather than of which amount to consume.

Due to the potential endogeneity problems of including price in the regression, we need to instrument for it. We use rent level, political color of the majority in the county council, tax rate and operational model as instruments for public price in each area. However, the instruments are valid but weak, and do not yield any significant or interpretable results why we choose to only include the regression in Appendix 4. The lack of significant results is likely to be due to model misspecification rather than absence of a real effect. What our model likely fails to take into account is a cultural component that plays a major role in determining dental care habits. The effect of this component on consumption would work in the opposite

(27)

direction to price, thus controlling for it would likely yield a negative effect of price on consumption. Exploring these cultural differences further is however beyond the scope of this study.

Going back to the equality aim of the Swedish dental care law, large price disparities may in themselves be seen as a violation. However, the fact that consumption is price elastic gives that price differences will also affect consumption, why it increases the importance of studying price disparity within the country.

3.2 Price Disparity

This section will show that there is large price disparity between different areas in Sweden which suggest that dental care is not given on equal terms for the entire population. We find that both socioeconomic variables as well as variables explaining demand have significant effect on prices. The most striking result is that price competition does not seem to increase with number of practitioners as predicted by general economic theory. On the contrary, price increases with the number of dentists per person, why it is reasonable to assume that competition is distorted.

Large differences

Table 6 reports summary statistics of prices for the four treatment packages analyzed in this study. It is evident that there are large price disparities for all four treatment packages between different areas within the Swedish dental care market. Treatment package 1 which represents a standard examination and cleaning, has the highest variation where price differs 32 percent between the 10^th and 90^th percentile among public prices and correspondingly 17 percent on the private side. As the treatment packages become larger and more expensive the percentage differences decrease compared to treatment package 1. However, due to the higher prices these differences get more substantial in terms of actual patient cost. The table shows that public prices are generally lower than private prices for all treatment packages, as the prices are lower for the public submarkets, both on the median and the 10^th percentiles.

However, for treatment package 1 and 2, the public postal code areas on the 90^th percentile charges more than the private areas on the same percentile. Further, for three out of four treatment packages the public prices show the greatest variation. The Local Government Act states that the public prices should be cost based why these differences in theory should

(28)

23 3. Results

reflect cost differences over the country. However, as previously discussed there might also be systematical differences in the price setting.

Table 6. Treatment package dispersion for public and private prices, for the year 2011.

Treatment package 1

Treatment package 2

Treatment package 3

Treatment package 4 Percentiles Public Private Public Private Public Private Public Private

10th 775 790 1665 1675 4114 4485 13245 13445

25th 775 810 1665 1685 4385 4558 13277 13690

Median 782 850 1685 1760 4475 4655 13550 13913

75th 833 880 1784 1841 4560 4744 13620 14265

90th 1025 925 1935 1879 4565 4820 13883 14589

Diff. 90th – 10th 32% 17% 16% 12% 11% 7% 5% 8%

These differences show clearly that dental care is not given on equal terms for the entire population. Further, as dental care consumption is price elastic one can conclude that this disparity will lead to large differences in consumption, which is in line with the results above.

These results make it important to evaluate why and how prices differ in order to get a better understanding of the underlying reasons for this disparity.

Regression analysis

We will now through regression analysis explain the price differences shown above. When regressing for price disparity in different areas our dependent variable is the median price per treatment and year in each postal code area, divided upon public and private prices. The dependent as well as several of the independent variables are taken in logarithms, as many of them have both large and skewed dispersions.

The estimated OLS equation is

Ln(PRICE)_i = β₀+ β_D 𝐷𝑒𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑠_𝑖+ β_c 𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑜𝑛_𝑖 + β_d 𝑑𝑢𝑚𝑚𝑦_𝑖+ ε_𝑖

where Ln(PRICE)i is the ith observation of either public or private median price in a postal code area and 𝐷𝑒𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑠_𝑖 is the matrix of exogenous explanatory variables describing differences in demand across areas. 𝐶𝑜𝑚𝑝𝑒𝑡𝑖𝑡𝑖𝑜𝑛_𝑖 is the equivalent matrix of explanatory variables describing variation in market structure between areas. 𝑑𝑢𝑚𝑚𝑦𝑖 is the matrix of dummy variables used to explain price differences between treatments, public and private prices, and over years.

(29)

Table 7. Regression results (standard errors in parentheses) Ln(PRICE)

Ln(INCOME/CAP.) 0.0452*

(0.0273) Ln(POP.DENSITY) 0.0132***

(0.00281)

ETHNICITY -0.164**

(0.0664)

EDUCATION -0.158***

(0.0448) Ln(DT/100 000) 0.0288***

(0.0103)

PTJ/PRIVATE 0.0488***

(0.0114)

Ln(DT/PRACTICE) 0.0123

(0.00639)

d.BIG.CITY 0.0201*

(0.0110)

d.PUBLIC -0.0801***

(0.00875)

d.TR.202 -1.51***

(0.00753)

d.TR.501 1.32***

(0.00752)

d.TR.705 0.445***

(0.00749)

d.TR.821 -2.30***

(0.00757)

d.TR.824 2.62***

(0.00753)

d.Y09 -0.0854***

(0.00535)

d.Y10 -0.0359***

(0.00532)

Constant 5.89***

(0.335)

R²ADJ. 0.9950

NOBS 2879

***significant at the 1 percent level, **at the 5 percent level, and *at the 10 percent level.

(30)

25 3. Results

Competitive Distortions, Information Asymmetries and Product Differentiation

By looking at the coefficients of the variables describing market structure in Table 7, inferences can be drawn about the level of competition in different areas and how it affects the price level. The fact that the coefficient for Ln(DT/100 000) is positive and highly significant is a striking result. We can thereby infer that the market conditions are such that another dentist entering the market will not put downward pressure on the price, but instead an additional dentist will increase price. In fact, the coefficient of 0.0288 gives that a one percent increase in Ln(DT/100 000) will increase price by around 0.03 percent. Since the number of dentist per 100 000 inhabitants varies from 26 to 198 this will have great impact. It is also important to stress that this effect is found after controlling for variables such as income, population density and a dummy variable to control for extreme values in the big cities.

The result contradicts general economic theory which predicts that larger markets with higher number of producers should be characterized by higher competition, in the sense of lower equilibrium prices. As our results show that price instead increases with the number of dentists we can infer that competition is distorted in the Swedish dental care market. This is an important result as it suggests that inefficiencies in the market increase prices. The possibility for firms to charge a price above marginal cost could arise both due to that the consumers, through product differentiation, are willing to pay more for the product or that they are unaware of the higher price due to information asymmetries. (Cabral, 2000)

The dental care market is characterized by severe information asymmetries as the dentist will have large information advantages, both when it comes to quality and prices, as well as about possible treatment options. These asymmetries will have a negative effect on price competition as the patient will have less knowledge of other dentists’ prices and options, and could thereby explain the positive effect of dentist per person on price. Better informed patients would therefore increase the consumers’ position on the market and should consequently increase the price competition. As mentioned in Section 1.3, the price comparison web page introduced in 2009 aimed to increase the knowledge about prices and thus decrease the asymmetries. However, as it has not yet been a success stronger efforts to promote information to patients could have a positive effect on the competition.

Furthermore, the positive effect of entry on price may very well be explained by product differentiation in the form of trust. Dental care consumers are unwilling to switch dentist due

(31)

to the trust built between dentist and patient. This reluctance to change dentist gives the care giver market power. The high level of product differentiation thus makes it possible for dentists to charge monopolistic prices. Hence, the product differentiation will lead to that less than the optimal quantity is consumed, especially since demand is price elastic (Grönqvist, 2012). Moreover, the product differentiation is an interesting issue from two perspectives.

Firstly, if the actual quality differs widely between dentists this should raise concerns about unequal dental care. On the other hand, if the product differentiation is created through trust between patient and dentist, but not on actual quality differences, this raises a greater concern about the information asymmetry in the market.

The positive effect of dentist per person on price could also support the theory on reputation goods discussed by Satterthwaite (1979) who argues that for reputation goods price increases with market size and participants. Such a mechanism might be present in the dental care market, where information asymmetries affect competition negatively. Grönqvist (2006) and Eriksson (2004) both test and find partial support for the reputation good effect. Grönqvist uses prices from 2004 and 2005 from a sample of dentists and regresses on number of dentists and population on municipality level. He finds that for some treatment packages both variables have significant effects on price. Eriksson (2004) uses data from 1998 and 1999 on prices for a sample of dentists, which he regresses on population on municipality level. Both these studies differ in that that they use price list prices and not data on actual prices charged.

Furthermore, they only use a sample of the population, while our data contains information on all practices. We further extend the analysis by the inclusion of additional variables other than income describing demand and market structure.

The effect that a higher number of producers increases prices could also be argued to be due to the presence of scale efficiencies or to be the result of higher prices in cities where commuting increases dental consumption and thereby dentist per capita. However, as we include Ln(DT/PRACTICE) in our model the economies of scale effect should be accounted for. Further, the results are statistically significant after accounting for income, population density and after including a dummy for the big cities. Thus the positive coefficient for DT/100 000 should not be viewed as due to either of these effects.

Private and Public Prices

Our data is collected for public and private median prices separately and we use the dummy variable d.PUBLIC to capture possible systematic price differences between the two. As

(32)

27 3. Results

shown in Table 7 we obtain a highly significant negative coefficient for d.PUBLIC, which concludes that public practices charge less than private practices in general. Our results gives that the public prices are overall about eight percent lower than the private prices. However, these differences are in reality not consistent for all areas, but the difference will be larger for some areas while public will charge the highest price in other. Table 8 depicts the share of areas where the public price median is higher, lower or equal to the private price median. As can be seen, in around 70-80 percent of the postal code areas, the public median price is lower than the private median price.

Table 8. Relation between public and private median prices Treatment

package 1

Treatment package 2

Treatment package 3

Treatment package 4

Private median higher 77% 69% 78% 79%

Public median higher 19% 29% 21% 17%

Same price 4% 2% 1% 4%

The fact that private dental care in general is more expensive than public dental care is an interesting finding, but it might not be in violation with the equality condition expressed in the Swedish dental care law. As long as public dental care is still accessible for all citizens they can choose the public dental care provider. We have seen that public practices are present in all of the postal code areas, why price differences between private and public practices might not in itself be a major equality concern. However, if this price difference reflects quality differences between public and private practices this raises concern about inequalities in the provision of dental care. In a survey from 2010 consumers answered about their level of satisfaction of different services. The satisfaction level was very high for both public and private dental care, but the private patients were about seven percentage points more satisfied.

(SKI, 2010) However, it is important to note that this study did not say anything about the actual quality given. As we do not have data on quality a further analysis of this subject is beyond the scope of this study.

Praktikertjänst

When we include a variable for the portion of the private dentists that are affiliated to Praktikertjänst we obtain a coefficient that is positive at a high significance level. This gives that the Praktikertjänst market share is positively correlated with the median price in the area.

This result could have different explanations. Firstly, Praktikertjänst dentists could for various

Supervisor: Johan Stennek Dental Care on Equal Terms?

Graduate School

Master of Science in Economics Master of Science in Finance Master Degree Project No. 2012:45