An empirical analysis on entry and competition in the Dutch Real Estate brokerage Market

(1)

University of Amsterdam

Faculty of Economics and Business

An Empirical Analysis on Entry and

Competition in the Dutch Real Estate

Brokerage Market

Master Thesis for the Master in Economics

Major: Industrial Organization

Author:

Supervisor:

Niels van der Poel

Sander Onderstal

Student number:

Intern Supervisor:

6337473

Bastiaan Overvest

Submission Date:

Second Supervisor:

26-01-2012

Maarten Pieter Schinkel

(2)

Acknowledgements

The thesis in front of you is the result of many months of reading, researching, processing and writing. However I could not have done this without the help of some people I hereby wish to thank. First of all I would like to thank my family, friends and Paula for always supporting me. Secondly, I am grateful for the supervision given by Sander Onderstal, his input and quick reactions on my questions were very helpful. Thirdly, I want to thank the NMa for offering me the opportunity to do this internship. Next to this is also wish to express my gratitude to the colleagues of the Monitor Financial Sector, they supported me with lots of feedback and a great deal of new insights. Last but certainly not least I would like to thank my intern supervisor Bastiaan Overvest. Without the support and guidance of Bastiaan I would not have been able to write the thesis as it lies before you.

(3)

Disclaimer

(4)

1111

Introduction

The market for Dutch real estate brokers has gone through substantial changes in the last 20 years. The first major change occurred in 1994. This year the fixed commission rate for real estate brokers was abandoned. This created possibilities for real estate brokers not only to compete on quality but also on prices. In 2001 under the MDW project1_{, the market for real estate brokers was further deregulated. From this} moment on the profession of real estate agent was no longer protected and no more regulatory restrictions were posed upon entry. Up till that moment brokers had to be sworn into the profession. Besides these changes there was also another important factor that changed the market severely: the internet. Before the uprising of the internet an important advantage of the real estate broker was his insight on the market for houses. At that time this information was limited to real estate brokers what made them essential in the search for and sell of a house. However with the increasing influence of internet sophisticated housing search engines arose of which Funda is currently the most important. Through this change the information advantage of brokers largely disappeared and the focus of the broker shifted towards the remaining activities. It seems that these three changes should stimulate competition: more price competition, lower entry barriers and less market power for brokers. Indeed, a first look on this market does not give signals for competitive problems on this market. The number of real estate brokers increased substantially over the last decade. Although estimates differ, it seems that the number of brokers increased by almost 60% in the last decade. Next to this there was a small decrease in the commission rates in the past ten years. However in absolute terms the commission fee increased up till the financial crisis of 2008 (Business surveys NVM). Considering these mixed signals it is not really clear how competition in the estate brokerage market developed over the years. To illustrate this let us consider the US. In this market it is common to charge a commission fee of 6%. Although there is free entry and brokers can charge any price they wish there is little variation in the price charged (Hsieh and Moretti, 2003). Levitt and Syverson (2008) researched this phenomenon and found supporting evidence that this might be the outcome of (tacit) collusion among commissioning brokers. This example for the US shows that even if a market has a high level of entry it is possible that price competition is at a low level. This could also be the case for the Netherlands. Hsieh and Moretti (2003) find that substantial entry in the US brokerage market this did not lead to more price competition and that this entry was actually excessive and socially wasteful.

So, the question is whether the increased number of real estate brokers actually intensified competition. The purpose of this study is to get more insight in competition on the market for real estate brokerage. The focus in this case is on what the effects of entry and exit is on price competition in the Dutch estate brokerage market. The research question in this thesis is therefore:

How does entry and exit on the real estate brokerage market relate to price competition?

Many brokers and brokers associations claim that there is fierce competition on their market2_. However questions on the effects of entry in the Dutch market on prices have largely been unanswered in economic literature and other research. Yet this is an important question considering that the buying and selling of a house is one of the most important and expensive decisions in a person’s life. It is therefore

1_{MDW is a project to stimulate competition, deregulation and regulatory quality}

(7)

http://nieuws.nvm.nl/actual/juni_2011/nma-important that the brokerage market is functioning well. Because the majority of the activities of a broker are related to the selling of a house (Blauw, 2011), I will focus on the selling services of brokers and leave out other activities such as the buying and valuation of a house. The focus on the selling services has the advantage that it prevents the analysis from being too complex while the most important activities of realtors are covered.

Before going into detail on how I will answer my research question, I will go more into depth on the question itself. First of all what is competition? Competition has several dimensions. Usually competition on prices is assessed. However factors such as quality, quantity, product variety, innovation and capacity (maximum number of houses in portfolio) can be important factors too. In this thesis I especially focus on how entry and exit relate to price competition on the brokerage market. I do this by formulating two econometric models based on the theoretical model of Hsieh and Moretti. Using this model I will try to explain the number of brokers on the market and by doing so I infer the level of competition and efficiency on the estate brokerage market.

The set-up of the remainder of my thesis is as follows. In Chapter 2 I will offer a brief overview of the key characteristics of the brokerage market. In Chapter 3 I will review the economic literature on competition in the estate brokerage market. Based on the information of chapters 2 and 3, I will postulate a theoretical framework in chapter 4, which I will translate into two new econometric models. In chapter 5 I will go into detail on the dataset used for the empirical analysis. In chapter 6 I will report the results of the empirical analysis based on the econometric models postulated in chapter 4. Chapter 7 is the last chapter. In this chapter, I discuss the outcomes of my analyses and I present research recommendations.

(8)

2

22

2 Real estate brokerage in the Netherlands: A market over

Real estate brokerage in the Netherlands: A market over

Real estate brokerage in the Netherlands: A market overview

view

This chapter will supply a short overview of the key characteristics of the profession of a broker and developments on the Dutch real estate brokerage market in the last decade. In particular, I will give a first glance at competition in this market. In section 2.1 a short overview of the function of a broker is given. Subsequently in section 2.2, I will discuss the key characteristics and developments of the demand for and supply of real estate brokerage in the Netherlands. In section 2.3, I will further elaborate on pricing and pricing structures on the brokerage market. Finally, in section 2.5, I will conclude on the key characteristics of the Dutch estate brokerage market and discuss how this relates to competition on this market.

2.1

2.1 The

The function o

The

function o

function of a broker

f a broker

The primary goal of a broker is to mediate in the sell of a house. For this role as mediator the broker usually represents the seller or the buyer. This mediating involves several services that differ for a selling broker and a buying broker. The core services of a selling broker can be described as follows3_:

- Valuation of the house

- Determining the seller’s marketing strategy

- Advertising the house (including placement on the internet) - House tours for potential buyers

- Negotiations with potential buyers - Drawing of the purchase agreement - Conveyance of a house

Besides these core services there are realtors who provide additional services such as advice on mortgages and insurances, styling and other services related to the sell or buy of a house. However there is also a development of brokers who supply a minimal level of service to their customers at a lower price. For example house viewings are not part of the services of such a minimal service broker. Internet brokers are the best example for this new kind of realtor.

2.2

2.2 The demand

The demand for

The demand

for

for and supply

and supply of

and supply

of

of real estate brokerage

real estate brokerage

real estate brokerage services

real estate brokerage

services

It seems safe to assume that the key factor in the demand for a broker is the number of houses for sale. In turn the number of houses for sale depends on macro economic factors, such as the state of the economy, population growth and local demand and supply factors. In this it seems unlikely that the quality or price of brokers substantially influences the total number of houses on the market. The market for brokerage can therefore be considered as a secondary market of the housing market. As a consequence of this the total demand on the brokerage market is largely exogenous of the price and quality of the services of brokers.

(9)

http://www.hypodomus.nl/huis-verkopen/verkoop-huis-stappen/waardebepaling-In 2010 the total value of the real estate brokerage market was about 360 million Euros.4_{However at} its peak level in 2005 the market was worth roughly 664 million Euros. This volatility in the market size can largely be explained by the volatility in the number of houses sold, as can be seen in figure 1. The Dutch broker however has little influence on the total number of transactions. The market does however react on the changes in demand. When we take a look at the developments in this market one can see in figure 1 that up till 2004 a large part of the market size increase was driven by an increase in the commission fees charged by brokers. This coincided with more entry of brokers on the market, what in turn lowered the productivity in the period 1997-2004. After 2004 the volatility of the quantity of transactions increased, this played an important role in the increase of the market size up till 2006 and the sharp decrease in the market size after this year. When we consider the number of brokers one can see that to a large extent changes in the number of brokers coincides with changes in the market size. An explanation for this might be the low barriers to entry on this market. Due to the low barriers of entry potential excessive profits will quickly be dissipated by entry. Yet further and more substantiated empirical evidence is necessary to support this hypothesis. In chapter 4 I will present a theoretical framework that could explain the developments shown in figure 1, this model will also be used to further analyze competition and entry and exit in this market.

2.3

2.3 Commission and competition

Commission and competition

In the previous section, some signs were observed that possibly indicate a market where excessive profits are dissipated through entry. In turn this indicates that the market is competitive. Indeed, in standard industrial organization market models it is often shown that an increase in the number of players results in intensified competition. However did this increase also affect prices on this market? This section will go more into detail on the developments of the prices charged and the pricing structure in the brokerage market.

4_{According to CBS in 2010 126127 houses were sold, the average commission fee of NVM brokers was 3287 and according to a}

consumer survey (EIM, 2010) 87% of the transactions was carried out by a broker. This results in an estimate of the market value of 360 million euros.

Figure 1 - Development of key factors in the brokerage market

40 60 80 100 120 140 160 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Year In d ex ( 20 0 2= 10 0 )

(10)

First it is useful to elaborate more on how brokers charge consumers for their services. As in many countries (Delcoure and Miller, 2001) it is most common for Dutch brokers to charge a commission rate based on the selling price of a house. However there are an increasing number of brokers that charge a flat fee, independent of the housing price. Results from the consumer survey (EIM, 2011) show that the majority of the brokers (63%) uses a commission rate of the selling price, 21% uses a flat fee, 8% a combination of the two and the remainder of the respondents does not know the pricing structure of its broker. This shows that commission rates are still common practice among brokers. The average commission rate in the Netherlands varies between 1,5% and 2% (NVM, business survey, 2010: Mckinsey & Company, 2010). Compared to commission rates in other Western European countries this is relatively low. The average rate in Germany is for example 5.3%. However there are also countries where the average rate is lower such as England, where the average rate is approximately 1.5%.

Although the use of commission rates is common practice it is not clear why still so many brokers use commission rates. Literature does not suggest that there is a relationship between the price of a house and the costs of selling this house (Schroeter, 1987). Therefore in a competitive market the commission rates should be proportionally and inversely related to the housing prices. This would mean that if housing prices would increase by 1% the commission rate would decrease by 1%. Table 1 shows the average commission fee in different housing price categories.5

Table 1 – Average commission fee for different housing price categories

Price category house in Euros < 100.000 100.000-200.000 200.000-300.000 300.000-500.000 > 500.000 average Commission fee in Euros 1.619 2.442 3.365 4.594 7.422 3.528

The numbers in the table show that the average commission charged by an NVM broker sharply increases with the selling price of a house. This supports the idea that there is still a strong relationship between commission fees and housing prices. Assuming that these large differences in absolute fees are not driven by cost or effort differences it seems that the prices in the brokerage market are not on a competitive level.

Besides the pricing strategy of brokers it is also relevant to see the developments of the commissions over the last years. In figure 2 the developments of the average commission fee and selling price of houses is shown. It can be seen that the average housing price and average commission fee have been increasing in line with inflation. However the figure also shows that the fluctuations of the selling prices correlate with the average commission fee charged by NVM brokers. This supports the idea that pricing of brokers’ services is strongly related to the selling price of a house. However there are no signs that costs of a broker changed in a way that would reflect such a strong relationship between the commission fee and the housing. Considering the possible competitive pressure of entry it could even be expected that prices decreased in the last decade. Based on the strong relationship between the fee and the housing prices also in this case it seems that pricing on the brokerage market is not competitive. By using commission rates the fee of a broker depends on developments of the housing price. Because developments of the housing prices are

(11)

exogenous of services and fees, brokers are enabled to tacitly increase their fees. It seems that this increase is not noticed or countervailed by the consumer.

1000 1500 2000 2500 3000 3500 4000 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Year E u ro s 60 70 80 90 100 110 120 In fl at io n i n d ex

Average comission fee Average selling price x 100 Inflation index 2005=100

Figure 2 – Development of the average commission fee and the average selling price of a house

2.4

2.4 Conclusion

Conclusion

The goal of this chapter was to give an overview of the developments and key factors of the brokerage market. The overview supplied some contradicting characteristic of the market. First it is shown that there are high levels of entry on the brokerage market. Therefore, it seems that the market is competitive and possible excessive profits will be dissipated by entry. However the commission fee is typically an inflexible percentage of the house price, which is not in line with a competitive market. In a competitive market it is expected that prices depend on costs and competitive behavior between brokers and not on the average housing price. I will discuss these seemingly contradicting characteristics in the following chapters.

(12)

3333

Competition in brokerage markets: literature review on competition

in different countries

In assessing the Dutch brokerage market it is useful to also consider how brokerage markets work in other countries and what past research was done on this subject. I will use insights gathered in the literature in my empirical analysis. The remainder of this chapter consists of four sections. In section 3.1, I discuss competition on brokerage markets in countries similar to the Netherlands. Continuing on this I discuss literature that tries to explain the limited price competition on these markets (section 3.2). The literature that I address in section 3.3 focuses on the functioning of the market, especially the relationship between entry and exit and price competition is addressed. In section 3.4 I will conclude on the main findings in this chapter.

3.1

3.1 Competition in t

Competition in the estate brokerage market

Competition in t

he estate brokerage market

Similarly to what I found in the previous chapter, the OFT (2010) and FTC (2007) come to the conclusion that the estate brokerage markets in their countries are characterized by low barriers to entry, high levels of entry and inflexible commission rates. Based on this it seems that these markets are not functioning optimally. However Schnare and Kulick (2009) argue that in fact there is price competition on the US brokerage market. The authors conclude this on basis of empirical analysis conducted on the US brokerage market. They find evidence that the commission rate is not completely fixed in relation to the housing price. They also find that the commission rate is related to several factors in a way that reflects price competition, such factors are the ability of a broker and the strength of the market. The authors interpret these results as evidence for market-driven pricing. However in my opinion this is a rather subjective conclusion. This because I think that the results in this article should be interpreted as a market that is driven only very limited by prices. Therefore this indicates that price competition on this market is highly imperfect. To show this consider the following example. The authors find elasticities of the commission rate to housing prices that in general vary between -0.015 and -0.18. However in a competitive market it is expected that this elasticity is close to -1. This example shows that to some degree there is variation in commission rates however in my opinion the found results should be interpreted as an indication for limited price competition.

In 2007 the FTC analyzed the US brokerage market to assess the level of competition and the functioning of the market. This research was motivated by concerns on limited competition on prices among brokers. Empirical results from the different studies reported in the FTC (2007) report show that the average commission rate have been decreasing over the period 1991-2005. However the housing prices have been increasing much sharper resulting in a substantial increase of the commission fees in absolute values. According to the American Bankers Association if the market for brokers was competitive commission fees could fall as much as by half (FTC, 2007: 36). The strong relationship between housing prices and commission rates show that fees not so much depend on competition but more on the housing price. The OFT (2010) conducted a similar research to asses competition in the British market. They also find sticky commission rates and they observe that commission rates on this market are characterized by price points. On the back of the rising housing prices the absolute fees have been increasing in the UK, the OFT states that this resulted in excessive entry on the estate brokerage market. Based on the reports by the FTC and OFT it seems that price competition in the estate brokerage market in the US and UK is not functioning optimally.

(13)

Considering the similarities between the US, UK and the Netherlands it is useful to consider articles that go more into depth on explaining competition and the functioning of these markets.

3.2

3.2 Explaining

Explaining the lack of price competition

Explaining

the lack of price competition

In the previous section I established that price competition on the US and UK brokerage market is not optimal. In this section I will discuss some articles that give possible explanations for limited price competition on these markets.

The concentration of a market is often used as a tool to evaluate competition on a market. This is based on the idea that on concentrated markets the market power of incumbent players is possibly higher what could result in higher prices. Considering the sticky rates on the brokerage market it could be the case that this is due to high levels of concentration. Although this seems unlikely considering the high level of entry on this market and the large number of incumbents. Beck et al. (2010) study this hypothesis, using the Herfhindahl-Hirschman Index (HHI) the concentration of 90 markets in the US is studied. The outcomes of this analysis show only a few concentrated markets what leads the authors to conclude that in general the estate brokerage market is not concentrated. Based on these results it seems that the concentration of the estate brokerage market does not explain the lack of price competition on this market. However the article by Beck et al. (2010) does not substantiate the choice for the definition of the geographical market. If the authors defined the markets too wide this could lead to an underestimation of the real concentration on the market. Another possible improvement for the article is linking the average commission fee in a market with the corresponding HHI. This way it can be analyzed whether there is a relationship between the average price and the concentration level of a market. Considering the above it does not seem likely that the lack of price competition is due to concentrated markets, however further research is necessary to completely exclude this possibility.

Another explanation that can clarify the differences in rates and fees might be that this signals quality. However Panle and Pathak (2010) analyze the US brokerage market and show that higher commission rates have very little impact on the likelihood of selling a house and the selling time of a house and even no significant impact on the selling price of a house. Based on these results Panle and Pathak (2010) conclude that the prices charged by brokers signal very little quality. This remark by the authors raises the question whether there is any competitive base for the commission rate system that is custom in estate brokerage markets. The stickiness of commission rates causes that there are large difference in absolute fees between different housing price categories. In a competitive market the price of a service is based on costs, however it seems unlikely that differences in costs can explain the large differences in fees (Schroeter, 1987; Cavelaars, 2003).

In the previous two paragraphs I discussed possible explanations for the lack of price competition in the US and UK market. These explanations were based on the idea that the market structure of the brokerage market could possibly result in rigid commission rates. However it is also possible that (tacit) collusion is responsible for the lack of price competition on the estate brokerage market. Levitt and Syverson (2008) supply evidence that this might be the case for the US brokerage market. An important characteristic of the US market is that the transaction of a house generally involves a selling and a buying broker. This interdependency between brokers is a key factor in sustaining collusion. The authors argue that this dependency can be used by brokers to discipline discount brokers. Buying brokers can steer away from houses that are sold by discount brokers. In doing so the broker gives the discount broker a competitive disadvantage in the selling of a house. Using this as a stick to punish deviators, brokers can possibly sustain

(14)

a collusive equilibrium. The authors find supporting evidence for this hypothesis. By doing empirical analysis they find that the average selling time of a discount broker is significantly higher compared to full fee brokers. At the same time houses sold by discount brokers sell at similar prices compared to normal brokers. This leads the authors to conclude that consumers on average could save $5000 on brokerage services. It is however questionable whether this possible collusive outcome could be sustained in the Dutch estate brokerage market. This is because the interdependency between brokers is substantially lower for the Netherlands. A large part of the Dutch housing buyers does not use a buying broker. Considering this, the “stick” which brokers can use to discipline discount brokers might be too small to sustain a collusive outcome.

3.3

3.3 Price competition and entry

Price competition and entry:::: explaining the functioning of the market

Price competition and entry

explaining the functioning of the market

In the previous sections it was shown that price competition on the US and UK brokerage market is limited and that the standard literature cannot give a conclusive explanation for this observation. A model that can possibly explain the functioning of the brokerage market comes from Hsieh and Moretti (2003). The authors base their theoretical framework on the seemingly contradicting characteristics of this market: the rigid commission rates and the low entry barriers on this market. Using these features the authors formulate an entry model. This model show that the lack of price competition results in a market where entry and exit is the mechanism through which the market reacts on exogenous changes, such as an increase in the average housing price. Based on this theoretical model the authors conducted several empirical analyses. The findings of these analyses show that higher housing prices result in a higher number of brokers while at the same time the productivity of a broker decreases. Next to this they also find that there is no relationship between income and housing prices. These three findings are consistent with the proposed model by the authors. Consequence of these findings is that entry on the US brokerage market is excessive and socially wasteful. Hence, extra income through higher prices is dissipated by entry. As a result of this, increases in commission fees do not benefit anyone, especially consumers. Besides this excessive brokers could add value to economic prosperity by being active in alternative employment opportunities.

Panle and Pathak (2011) continue on the results found by Hsieh and Moretti and try to quantify the social costs of excessive entry in the brokerage market. In doing so they compute several counterfactuals. The first counterfactual assumes a regulated decrease in the commission rate. Using a micro-founded model the authors find that if the average commission rate would be lowered by one half, entry would fall by one third, productivity of active brokers would increase by 73% and the likelihood of selling a house increases by 2%. These changes add up to huge social benefits of almost 3 billion dollars. Second counterfactual considers a market structure that is based on the costs of selling properties that potentially could lead to more flexible commissions. Using a conservative measure of costs the authors find that there would be 24% fewer agents and a productivity increase of 31%. The results of the authors show that higher social welfare could be achieved when commission rates in the US would be less rigid. This would especially benefit consumers while at the same time remaining brokers do not have to be worse off.

Han and Hong (2008) also point out the inefficiencies of excessive entry. The authors discuss two cost inefficiencies: loss of economies of scale and non price competition. The first inefficiency is based on the idea that the average costs of a brokerage firm decreases with the number of transactions. Considering that excessive entry will lower productivity this will result in an average cost increase per transaction. The second source of inefficiency is found in the social wastes of non-price competition. Because the commission rates in the US tend to be rigid there is little price competition. However with the increase in the number of

(15)

often comes with socially wasteful costs such as soliciting and marketing. For both the mentioned sources of cost inefficiencies the authors find empirical evidence to support their hypotheses. The results of the article show that a 10% increase in the number of brokers results in a 4.8% increase in the average costs per transaction.

3.4

3.4 Conclusion

Conclusion

In chapter 2 it was already shown that the Dutch estate brokerage market is characterized by low barriers to entry, high entry levels and seemingly uncompetitive prices. The literature discussed in this chapter shows similar characteristics for the US and the UK. Economic research into these countries shows that entry does not have to be a signal for a competitive market and that in combination with rigid commission rates this entry can even be excessive and socially wasteful. Considering the similarities with the Dutch estate brokerage market it is interesting whether this also holds for the Dutch brokerage market. Using the theoretical framework postulated by Hsieh and Moretti (2003) this will be further analyzed in the next chapters.

(16)

4

4 Theoretical framework

Theoretical framework

Theoretical framework and methodology

Theoretical framework

and methodology

In this chapter I develop a model that builds on the theoretical framework of Hsieh and Moretti (2003). As the literature suggests, entry and exit of brokers is possibly the mechanism through which the market reacts on exogenous shocks (Hsieh and Moretti, 2003; Han and Hong, 2008; Panle and Pathak, 2011). The framework suggested by Hsieh and Moretti (2003) is consistent with this observation and postulates the number of broker as the endogenous factor in the estate brokerage market. This is remarkable considering that often prices or quantities are postulated as the endogenous factor in theoretical models. Whether this hypothesis holds can only be shown by empirical analysis. In this chapter I propose an economic model that will serve as the basis for the empirical analysis. I will do this by elaborating more on the theoretical framework of Hsieh and Moretti (section 4.1) and subsequently postulating two new econometric models (section 4.2). In section 4.3, I will discuss the assumptions and the possible effects of violations of this assumption. Continuing on the findings in the previous sections I will postulate three hypotheses and elaborate what empirical outcomes would be in line with these hypotheses (section 4.4). In the last section (4.5) I will conclude on the general findings of this chapter.

4.1

4.1 The

The theoretical framework

The

theoretical framework

The following elaboration of the theoretical framework draws heavily on the article by Hsieh and Moretti (2003).

Consider an individual that wishes to enter a local market as a real estate broker. Assuming that there are no fixed costs or other substantial entry barriers this person will become a broker when the expected income (

π

j) in region

j

exceeds the alternative wage (

w

j) in region

j

. The expected income,

assuming that all brokers are symmetric, is equal to total market income divided by the number of brokers in the market (

b

_j). The total market income is simply the number of transactions (

S

_j) times the commission rate (

c

) and the average selling price of a house (

P

j). Using this we can find that an individual is indifferent

between employment in estate brokerage and employment in an alternative sector when:

j j j j

w

b

S

P

c

=

⋅

(1)

The intuition from equation (1) is that brokers will only enter this market when the income of being a broker is larger than the alternative wage opportunity. The market is in equilibrium if the income of a broker is equal to the alternative wage. This model implies that no supra normal profits can be made, in the absence of entry barriers or other frictions.

To better understand the model suppose that equation (1) is not in equilibrium and the income of a broker exceeds the alternative reservation wage ( j

j j j

w

b

S

P

c

>

⋅

). In this case it will be attractive to become a broker considering the excessive profits in this market. Therefore brokers will enter, and as a response to the competitive pressure of (possible) entry incumbents on the market could lower their commission rate. A lower commission rate may restore the market back in equilibrium.

(17)

In this model only the commission rate and the number of brokers serve as mechanisms through which the market reaches equilibrium. The other variables in the model, number of transactions, average housing price, and reservation wage, are assumed to be exogenous factors. A discussion on the assumptions will follow in section 4.3.

Using the feature that the model responds through either entry and exit or through the commission rate implications on price competition can be inferred. To show this consider the following example. The average housing price increases by 10%, as a result ceteris paribus the income of a broker increases by 10%. This supra normal income cannot be sustained long, either the increased profits will be dissipated by entry or the commission rate will have to be lowered. In case of perfect price competition prices are at marginal cost level, an exogenous increase of the housing price will therefore be corrected by a proportional decrease in the commission rate. In case there is no price competition incumbent brokers do not react on the exogenous increase of their fee. This lack of price competition facilitates room for entrants, the increase in income will therefore be dissipated fully by entry. Of course their can also be an intermediate level of price competition in that case part of the increased income is dissipated by entry and part is achieved through lowering the commission rates. Based on this example one can see that the level of price competition can be inferred from the reaction of the market on exogenous shocks. A reaction mainly through the commission rate indicates a high level of price competition a reaction mainly through entry or exit indicates imperfect competition. However in both cases the income of a broker is similar. For now it is assumed that the commission rate is constant and exogenous, further elaboration on this will follow in section 4.3.

The assumption that the commission rate is constant and exogenous implies that there is no price competition. If this would be the case this would imply a social loss, hence prices are supra competitive while at the same time incumbent brokers do not benefit from increased fees. However there is another source of social waste. To show this consider the productivity of a broker as it is formulated by Hsieh and Moretti (2003). j j j j

P

c

w

b

S

ty

productivi

⋅

=

(2)

Hsieh and Moretti describe productivity as the number of transactions per broker. Equation (3) shows that ceteris paribus the productivity of a broker is lower in regions where the housing price is higher. The lack of price competition not only results in high commission fees it also results in an inefficient number of brokers. The same amount of work could be produced by fewer brokers. Therefore entry of brokers due to higher housing prices is excessive, this is an externality created by the combination of inflexible commission rates and free entry. A Pareto efficient outcome in this market is therefore a market with a high level of price competition. In the next section an econometric model, based on the theoretical framework in equation (1), will be developed. This econometric model can be used to estimate the level of price competition and inefficiency in the Dutch estate brokerage market.

4.2

4.2 Econometric model

Econometric model

The theoretical framework in this thesis depends heavily on the framework suggested by Hsieh and Moretti (2003). Based on this framework the authors tested the validity of their model empirically for the US brokerage market between 1980 and 1990. They do this by postulating three hypotheses and subsequently testing these hypotheses separately. The three tested hypotheses are: i) The number of realtors increases

(18)

with the selling price of houses, ii) The productivity of realtors decreases with an increase of the housing prices, iii) There is no effect of the average housing price on the income of brokers. The exact empirical results are presented in chapter 6.7 however the outcomes of the analyses are in line with the proposed theoretical framework. To compare the results for the US market with the Dutch market I also replicated these analyses, further discussion on the empirical analysis can therefore be found in section 6.7. However in my view the empirical approach used by Hsieh and Moretti can be improved, in the next sections I discuss two alternative econometric models that also will be the basis for empirical analyses in chapter 6.

4.2.1 Housing price model

Hsieh and Moretti test their economic model by deriving three hypotheses and subsequently testing these hypotheses separately. However the theoretical framework also offers room to formulate an econometric model that incorporates these different models into one comprehensive econometric model. In corporation with my intern supervisor two new econometric models were developed. To show how this econometric model is formulated, I first rearrange equation (1) to the following equation:

j j j j

w

S

P

c

b

=

⋅

(3)

Looking at equation (3) one can see that the number of brokers is postulated as the dependent variable and the right hand side variables are postulated as the independent variables. As was discussed in section 4.1, is the number of brokers the mechanism through which the market reacts on exogenous shock. This was under the assumption that the commission rate is constant. By rearranging the model to equation (3) it can be tested whether this holds. The next step is formulating equation (3) in a way that it is empirically testable. In doing this the exogenous variables are transformed into natural logarithms. This has the advantage that percentage impacts of the variables on the number of brokers can be estimated. Considering that the commission rate is assumed to be constant it is dropped from the equation. The econometric specification than looks as follows:

ε

α

+

=

ln(

)

ln(

)

ln(

)

ln(

b

β

+

=

ln(

)

ln(

)

ln(

)

ln(

b

_j ₀ ₁

C

_j ₂

S

_j ₃

w

_j (5)

Where

C

jis the average commission fee in region

j

. It is clear that the advantage of commission fee model

compared to the housing price model is that the income per transaction is estimated in a more appropriate way. However both the econometric models are based on the same theoretical framework.

4.3

4.3 Assumptions of the model

Assumptions of the model

The theoretical model described in the previous section depends on several assumptions. In this section an overview is given of the key assumptions and it is discussed how these assumptions and possible violations of these assumptions can influence the model.

Free entry and exit

A key assumption in the theoretical model is free entry and exit. The importance of this assumption is that it ensures that excessive profits are dissipated through entry. If there were barriers to entry, incumbents would have certain power to keep the profits above competitive levels. The model therefore depends on the validity of this assumption. Several arguments support the validity of this assumption. Firstly, since 2001 every person is allowed to call him or herself a real estate broker, so there are no requirements to being a broker. Secondly, the fixed costs to enter the market are not substantial. Brokerage is a service, therefore a broker’s main capital is human capital. No large investments have to be done, such as investments in production equipment and inventories. This is also confirmed by a survey held among

(20)

brokers, results of this survey show that brokers do not experience substantial barriers to entry. Some brokers even stated that entry is to easy (Blauw Research, 2011). Considering these arguments it seems plausible to assume that the Dutch brokerage market is characterized by free entry and exit.

There are no differences in costs in between regions

The left-hand side of equation (1) is basically the average income per broker. However, this average income does not account for costs that brokers may incur in obtaining this income. Assuming that costs are constant across regions it can be excluded from the equation. But suppose that costs do differ per region. Fixed costs might be higher for example because of higher office costs. Or maybe more expensive houses are harder to sell and therefore brokers have to incur more costs to sell a house compared to cheaper regions. If this would be the case, ceteris paribus, the actual income per transaction is lower in high housing price regions than is assumed in the theoretical model. To understand the consequences of a violation of this assumption consider the following extreme example. Suppose that the differences in commission fees between regions can be explained completely through the difference in costs in this region. In that case the net income per transaction is similar for every region. Following the model this would imply that for this extreme example there is no relationship between the housing price/commission fee and the number of brokers in a region. Thus in general a violation of the cost assumption expresses itself in a smaller effect of the average commission fee on the number of brokers. Hsieh and Moretti also address this possible issue and show mathematically that in case costs are higher in high housing price regions this would imply a relatively lower number of brokers in this region. For this they assume that the costs made by a broker are

j

P

k

⋅

. Including this cost factor in equation (1) leads to the following equation:

j j j j j

w

P

k

b

S

P

c

=

⋅

−

⋅

(6)

This equation can be rearranged in a way that shows how the different factors influence the number of brokers:

)

/

(

1 )

/

(

j j j j j j

w

P

k

w

P

S

c

b

⋅

+

⋅

=

(7)

Equation (7) shows that when costs, k, in region j are higher that the number of brokers in this region is lower. For the empirical analysis this implies that if the assumption is violated the effect of the housing price/commission is biased downwards, meaning that the coefficients are closer to zero.

The number of transactions, the housing price and the reservation wage are exogenous and independent from each other

The model assumes that the right hand side variables of the model, number of transactions, average housing price and reservation wage, are exogenous and independent. This implies that a change in these variables is not caused by the behavior of brokers on the market or other factors in the model. If this would not be the case reverse causality could be an issue. If for example the number of brokers would influence the average selling price the assumed causality would be wrong and interpreting the results would not be appropriate. In this paragraph it is discussed whether these assumptions are valid.

(21)

First I consider the number of transactions (

S

_j), it seems appropriate to assume that the prices or services of a broker have little impact on the decisions of consumers to sell their houses. It seems more likely that matters such as divorces, employment and aging, and other macro and micro economic factors influence supply and demand on the housing market. The brokerage market is therefore a secondary market of the housing market. However it is also possible that sellers do not hire a broker. A study from EIM (2011) shows that about 13% of the consumers does not hire a broker. 8 percent point of these consumers already sold their house or did not want a broker. 3 percent point of the consumers found the broker to expensive. It is this group that possibly could be persuaded to hire a broker. However this marginal group of possible customers is so small that it seems safe to assume that this will not impact the analysis.

Considering the secondary market argument, the same rationale holds for the average housing price (

P

j). It seems unlikely that brokers can significantly influence the average housing price in the Netherlands.

Considering the relatively small number of brokers compared to the total labor force it also seems unlikely that the average income (

w

_j) in a region can substantially be influenced by developments on the brokerage market.

Based on the above arguments it seems appropriate to assume that

S

j,

P

jand

w

j from equation

(1) are exogenous factors in brokerage market, and therefore reverse causality issues can be ruled out. However it is also assumed that the exogenous factors are independent from each other. A violation of this assumption could lead multicollinearity issues. In the case of multicollinearity the exogenous variables are correlated with each other, this could bias the estimates and reduce the reliability of the estimates. Using statistical tools it can be verified whether this is an issue.

The commission rate is constant and exogenous; there is no reverse causality between the commission rate and the number of brokers

The most trivial and crucial assumption in the theoretical framework is the assumption that c is constant. Implicitly this assumes that there is no price competition. As was mentioned in previous section the model can react on exogenous shocks in two ways, through the commission rate or entry and exit of brokers. By assuming c is constant one assumes that exogenous shocks are fully corrected through entry and exit. This is a strong assumption, however it is easy to see whether this assumption is violated. In case that c is constant exogenous shocks will be corrected fully through the number of brokers on this market. A 1% increase in the housing prices will results in a 1% increase of the number of brokers in this region. If it is empirically shown that this effect is smaller than 1% this indicates that the assumption is violated and that there is some degree of price competition. In case of perfect price competition prices are based on market behavior of competitors and costs, therefore there should be no relationship between the fee of a broker and the housing price (assuming no differences in costs). From the model it follows that in that case there should be no effect of the housing price/commission fee on the number of brokers in this region. A violation of the assumption is therefore not a problem for the empirical analyzes, it is even an informative indicator on the level of price competition on the estate brokerage market.

(22)

4.4

4.4 Hypotheses

Hypotheses

This chapter started with an elaboration on the theoretical framework that is used in this thesis. The elaboration of this framework showed that under certain assumptions the brokerage market features excessive entry and imperfect price competition. Continuing on this theoretical framework, two new econometric models were formulated. This section discusses what empirical outcomes would support the validity of the theoretical framework. For this three hypotheses for the empirical analyses are formulated, these hypotheses are:

• Perfect competition (PC) hypothesis: Competition on prices is perfect; an exogenous change in housing prices/commission fees is fully corrected through commission rates on the brokerage market.

• No competition (NC) hypothesis: There is no price competition; an exogenous change is fully corrected through entry and exit on the brokerage market. The market facilitates excessive entry on the brokerage market.

• Imperfect competition (IC) hypothesis: There is imperfect price competition; an exogenous shock is partly corrected through prices and partly through entry and exit. The market facilitates excessive entry on the brokerage market. (From now on: Imperfect Competition Hypothesis)

In the empirical analysis the effect of exogenous variables on the number of brokers is studied. Empirical results that are in line with the Perfect Competition hypothesis show coefficients on the housing price/ commission fee that are not significantly different from zero. This implies that there is no measurable effect of the independent variables on the number of brokers, based on the theoretical framework it can be concluded that brokers compete on price. Empirical results that are in line with the No Competition hypothesis show significant coefficients equal to 1. This implies that an exogenous difference in between regions is fully compensated through entry or exit of brokers. This would also support the assumption that the commission rate is constant and exogenous. The imperfect competition hypothesis is supported in case the coefficients are in between 0 and 1 and are significantly different from 0 and 1. In that case part of the exogenous differences in between regions is corrected through prices and part of the differences is corrected through entry and exit. Coefficients close to 1 indicate a low level of price competition the largest part of exogenous differences is corrected through entry and exit. Coefficients close to zero indicate a high level of price competition, the largest part of exogenous differences is corrected through prices.

Outcomes that are in line with the NC and IC hypothesis do not only indicate imperfect price competition they also indicate inefficiencies. Entry is, as a result of imperfect price competition, excessive. This is because the productivity of brokers in regions with higher prices is lower. Empirically this should be shown in analyses that show a negative relationship between the average housing price/commission fee and the productivity. Replication of the Hsieh and Moretti analysis on the relationship between productivity and the housing prices can further substantiate this analysis.

(23)

In the following table the coefficients that are in line with the different outcomes are summarized:

Table 2 – Hypothesis and corresponding empirical outcomes

General outcome PC hypothesis NC hypothesis IC hypothesis Variable:

P

j Coefficient:

α

₁

0

1

=

α

1

=

1

0 <

α

1

<

1

Variable:

C

_j Coefficient:

4.5 Conclusion

Conclusion

In this chapter, I developed a framework for the empirical analysis. For this, I used the theoretical framework of Hsieh and Moretti (2003). In this model entry and exit of brokers is the mechanism through which the estate brokerage market reacts on exogenous changes. For example an exogenous 10% increase in the average selling prices of a house will result in a 10% increase in the number of brokers. The predictions from this model are that there are no excessive profits, the market is inefficient, prices are uncompetitive and the functioning of the market is socially wasteful. Although these are remarkable predictions it should be noted that they heavily depend on the assumption that brokers do not compete on prices. To test the theoretical model I proposed two new econometric models. Based on the econometric models I postulated three hypotheses: the Perfect Competition hypothesis, the No Competition hypothesis and the Imperfect Competition hypothesis. The NC and IC hypotheses are in line with the proposed theoretical framework and the predicted outcome of this framework. The PC hypothesis is confirmed when there is no effect of the commission fee/housing price on the number of brokers. Empirical analysis will show whether the proposed model can be confirmed in reality. The results of these regressions are shown in chapter 6.

(24)

5555

Data

The previous chapter went into detail on how the market for estate brokerage will be analyzed. In this chapter I will discuss the data used for the empirical analysis. In section 5.1 the best approach for the unit of analysis is determined. Subsequently I will address the different data sources in section 5.2. In section 5.3, I will present the descriptive statistics for the most important variables in the dataset. Lastly, in section 5.4 I will briefly conclude on the main findings in this chapter.

5.1

5.1 Unit of analysis

Unit of analysis

As became clear in the methodology chapter the empirical analyses will be based on cross-region analyzes. For this I will use OLS regression techniques. For the analyses it is necessary to determine what unit of analysis is most appropriate for the research. The possible different levels of regions are as follows: postal code, place of residence, municipality, COROP-region6_{and province. In this choice it is important to} minimize the interaction effect in between regions. In this there is a trade-off between reliability (i.e., the number of observations) and validity (i.e, no interaction effects between regions). For instance using postal code as unit of analysis will ensure a large number of observations what will make it easier to run reliable statistical analyses. However the validity of using postal code as unit of analysis is questionable. For example a broker may be located in postal code X while he or she may also be active in postal code Z and Y. This may be problematic because information in the dataset on this broker, such as commission fee, may also apply to region Z and Y. As a result of this the empirical analyses might be biased.

To overcome possible measurement errors the unit of analysis will have to be large enough to cancel out interaction effects between different regions. However it is not possible to totally cancel out this problem and as I wrote before it is a tradeoff between reliability and validity. For this purpose it is useful to define the geographical market of a real estate broker in the Netherlands. A market survey conducted by Blauw Research (2011) shows that for the selected observations in this study 56% of the brokers is active locally, 41% regionally and 4% at a supra regional or national level. Therefore it seems safe to assume that using a regional unit of analysis overcomes the largest part of possible interactions effects with other regions. This is also confirmed by results of the Elzinga-Hogarty test on a sample of municipalities in the Netherlands. Part of this test is the Little-Out-From-Inside (LOFI) test. The idea of the LOFI test is that for the selected market the producers do not significantly export their products or services to other geographical markets (Davis and Garces, 2009). Results of this test are shown in the scatter plot in figure 3. This figure shows the results of a randomly drawn sample of municipalities and regions in the Netherlands. On the vertical axis the LOFI values for the observations in the sample are shown. This LOFI value was estimated by looking at what percentage of the sales of brokers were sold in their region/place of residence. The horizontal axis shows the number of inhabitants per observation. In literature there is some discussion on what level the E-H test is significant, the prevailing opinion however is that this level is between 80-90% (Davis and Garces, 2009). From the scatter plot, in figure 3 we can roughly conclude that this level is reached for areas with a number of inhabitants between 100,000 and 200,000.

(25)

Based on the results of the survey and the LOFI ratios found it seems that using a regional level as unit of analysis is the most appropriate way. Regions fulfilling these criteria are COROP areas. Therefore in the empirical analysis the COROP area level is used as unit of analysis. The list of different regions can be found in appendix A1.

5.2

5.2 Data sources

Data sources

For the construction of the dataset four different data sources were used namely: The Dutch statistical office (CBS), membership lists of broker’s associations, the chamber of commerce (Kvk) and a survey among brokers conducted by Blauw Research. Below some more information on the data obtained from these sources.

5.2.1 Membership lists from the broker associations

One of the most important variables in the empirical analysis is of course the number of brokers. For this the member lists were obtained from the three brokers associations in the Netherlands: NVM, VBO and VastgoedPRO. These lists make it possible to accurately find the number of brokers per region. Disadvantage of this data source however is that it does not contain information on unorganized brokers. However about 96% of the transactions are carried out by organized brokers (Blauw Research, 2011). Therefore this does not have to be a big problem. However if the ratio of unorganized broker differs for different regions this might bias the results. To control for this problem there are two more sources for the number of brokers, these sources also contain unorganized brokers.

An empirical analysis on entry and competition in the Dutch Real Estate brokerage Market

University of Amsterdam

Faculty of Economics and Business