Prices at real estate auctions : the role of market knowledge and the reason for sale

(1)

Prices at Real Estate Auctions: The Role of Market

Knowledge and the Reason for Sale

Dennis Kroon - 10852093

Master Thesis

MSc Business Economics

Master track: Real Estate Finance

Thesis supervisor: Dr. M.I. Dröes.

Second supervisor: Prof. Dr. M.K. Francke.

(2)

2

Statement of Originality

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

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

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

(3)

3

Preface

This thesis is written at the end of the Master’s phase of Business Economics: Real Estate Finance. For this thesis, data was collected from the weekly auction books of the auction house ‘de Eerste Amsterdamse Onroerend Goed Veiling’ and data was received from the Dutch Association of Realtors. I would like to thank the organizations that have made this data available, as well as the people who have helped establish this data.

I would also like to thank my thesis supervisor Dr. Dröes for his good guidance while writing this thesis, his monitoring of progress, as well as his critical comments. With his help I am able to present this research report in front of you.

Finally, I would like to thank all my friends, family, acquaintances and colleagues for the psychological support needed to bring this thesis process to a successful end.

(4)

4

Abstract

This thesis aims to determine whether hiring a real estate agent to buy property at auction has a significant effect on the discount of properties thus sold. The discount is the difference between predicted market value and auctioned property price. A real estate agent has experience and expert knowledge in the field, and this may influence auction discounts and result in higher discounts for property purchases at auction. Furthermore, this thesis investigates the effect of the reasons for sale on auctioned property discounts. The following reasons for sale are investigated in this thesis: ‘voluntary sales’, ‘forced sales’, ‘sales due to bereavement’ and ‘sales under court order’.

The overdue charges, all related costs that are unpaid, are not always clarified in the auction books, which results in bidders underestimating outstanding amounts. The expectation, then, is that the discount for ‘forced sale’ houses is lower than the discount for other reasons for sale.

Data is collected from the auction house ‘de Eerste Amsterdamse Onroerend Goed Veiling, Dutch Association of Realtors, and the land registry. It focuses on properties sold at auction in the Amsterdam municipality. The period investigated is from 2005 until 2014, and 747 house price transactions are included. The data received from the Dutch Association of Realtors is used to predict the market value of the auctioned properties. A hedonic regression model is used to estimate the market value. The average discount of the auctioned properties in Amsterdam is 34.5%. The effect of expert knowledge from real estate agents is regressed based on the discount. The results of this thesis show that hiring a real estate agent has no significant effect on the discount, which means that there is insufficient evidence that real estate agents outperform private buyers or vice versa. Furthermore, expert knowledge in the field does not seem to result in higher discounts.

Nevertheless, a real estate agent oversees the buying process, which could be valuable for other reasons, such as a reduction in risk.

In addition, the reason for sale is regressed, based on the auctioned property discount. It has been found that a ‘forced sale’ and ‘selling under a court’ order have significant effects on the auction discount. Furthermore, ‘forced sale’ property discounts are on average 7.9% lower than ‘voluntary sale’ property discounts. Properties sold under court order, however, are on average 16.3% lower than that for voluntary sales. The ‘sale due to bereavement’ reason also negatively influenced the discount and was on average 8.1% lower than that for voluntarily sales, although the coefficient ‘sale due to bereavement’ was not statistically significant.

(5)

5 The regression results on the reason for sale on the discount are as expected, although the discount for ‘sale under court order’ is lower than for ‘forced sales’. The difference in auction discount could be explained by the overdue charges. Overdue charges vary between the reasons for sale and are not clarified in the auction books, resulting in bidders not having a clear idea of overall costs. The results of this thesis show that it is wise to investigate the overdue charges and extra costs of the property that is being sold to better predict the total costs. This could result in a lower difference in discount between different reasons for sale.

(6)

6

1. Introduction

As a result of the financial crisis of 2008, around 1.5 million properties in the Netherlands are “underwater” (Centraal Bureau voor de Statistiek, 2014). This means that the value of the property is lower than the mortgage value. Several factors can result in having a mortgage that can no longer be paid, such as a loss of income or dissolution of marriage. This can lead to the forced sale of a property. In this situation, the bank offers a specific time frame to the owner to sell the property and pay off the mortgage. If this time frame ends and the property is still not sold, the property will be sold at a real estate auction. The percentage of ‘forced sales’ of the total number of auction sales was around 55% in 2007. The crisis of 2008 increased the percentage of ‘forced sales’ to 95% in 2009, which shows the effect of the financial crisis.

It is a fact that auctioned properties are sold with a discount. According to the research of Brounen and Rijk (2009), there is an average of a 37% discount – which is compared at market value – for properties that are forced sold at auction in the Netherlands. This discount holds the reason why it is interesting to buy a property at auction. This thesis investigates whether two new determinants can be added that have an influence on the discount at real estate auctions. These two new determinants are the value of expert market knowledge of real estate agents and the effect of the reason for selling.

First, a real estate agent has experience as well as expert knowledge in the field, and their expertise may subsequently have an impact on the auction discount and result in a higher discount for people who want to buy a property at auction. This is because real estate agents can have an information advantage over private buyers. Real estate agents have knowledge about the market conditions and easements of the property. Next to that, the real estate agent can advise his clients about the maintenance level of the property, if the maintenance level of the property is not mentioned in the auction forms. Besides that, the real estate agent has experience with the methods of buying and selling a property at auction. This expert knowledge is important to determine the market value of a property and to know how to optimize the clients’ wishes, including the price of the property. Another advantage of hiring a real estate agent is that it could prevent the buyer from buying a property for too much money (NVM, 2015). This study tries to measure the effect and the value of expert market knowledge on auction prices.

Related literature investigated the effect of bidder knowledge on buying a property. The results of Milgrom and Weber (1982) show that bidders with private information outperform bidders with public information at sealed-bid auctions and that bidders’ profit increased when more information was received. Related literature also investigated the

(8)

8 effect of hiring a real estate agent to sell a house. Levitt and Syverson (2008) conducted research on the effect of real estate agents on the sale of a property. They found that homes owned by real estate agents sell at a price of 3.7% more than other properties. However, research has not investigated the effect of broker knowledge on buying a property at auction.

Secondly, a property sold at an auction is not necessarily a ‘forced sale’; it can also be a ‘voluntary sale’, a ‘sale due to bereavement’ or a ‘sale under a court order’. The people who sell their property voluntarily at an auction accept a lower price because they want to sell their house faster than on the regular market. If a property is sold due to a bereavement the person that inherits the property could not pay the costs of the legacy, or there is nobody that inherits the property. The reason to sell a house under court order is decided upon by the court. Selling a house under court order has the result that there is no timeframe to sell the house at the regular market and that it must be sold at auction immediately.

An important factor that plays a role when properties are forced sales is that there are overdue charges that have to be added to the total auction costs. The overdue charges are all related costs that are unpaid. This is also the case when properties are sold ‘due to bereavement’ and ‘under court order’. The average overdue charges differ based on the reasons for sale. The total number of these overdue charges, however, is not always mentioned clearly in the auction books, which leads to most people underestimating these overdue charges. This results in the expectation that the discount for properties that are sold with the reasons ‘forced sale’ and ‘sale under court order’ have lower discounts than those with other reasons for the sale. If a property is sold under court order the potential buyer could, for example, be responsible for the costs of expulsion from housing of the residents.

Researchers have investigated the effect of forced sales at auctions. The research of Brounen and Rijk (2008) is in line with the research of Mayer (1998) and the research of Allen and Swisher (2000). Mayer (1998) shows that properties that are sold in Los Angeles during the 1980s are sold with a discount of 0% to 9%, and that properties that are sold in Dallas are sold with discounts of 9% to 21%. Allan and Swisher (2000) show that property prices are on average 17.45% less when a property is sold at auction when compared to a regular sale. Now, although the discount of properties that are forced sales at auction is investigated by the existing literature, it has not investigated the other reasons for sale at auction, namely sale due to bereavement, sale under court order and voluntary sales.

Instead, previous literature has studied the average of all auctioned sales, which could lead to an underestimation of the effect of the reasons for sale. Therefore, it is worthwhile to investigate the reasons for sale; whether or not the discount for other reasons

(9)

9 for sale is as high as the discount for a property that is forced sold. In order to study this matter, the following research question has been formulated:

“How are the prices at residential real estate auctions in Amsterdam affected by the expertise of real estate agents and the reasons for sale of a property?”

To find an answer to this question, data from auction house ‘de Eerste Amsterdamse Onroerend Goed Veiling’ (EAOGV), data from the land registry and data from the Dutch Association of Realtors (NVM) is used. The dataset that is used for this thesis covers residential properties in Amsterdam which were sold at auction from January 2005 until December 2014. The dataset contains 747 properties.

Brounen and Rijk (2009) used a hedonic model to identify characteristics that determine the market value of the auctioned properties in the Netherlands. Property characteristics and transaction prices from the data of the Dutch Association of Realtors (NVM) database are used to estimate the market value accurately. The following characteristics are taken into account: the age of the property, the location of the property, the property type, parking availability and if the property has a garden. This model allowed them to provide a market value for their sample of 703 properties sold at auction.

The method that Brounen and Rijk (2009) used to estimate the discount of auctioned properties in the Netherlands is also used in this thesis to determine the market value of the auctioned properties in Amsterdam. To determine the market value even more accurately, two additional variables are added. The first addition to the model is whether the property is rented or not at the time of selling. The second is the condition of the ground lease; whether it must be paid annually or whether the ground belongs to the property owner or is paid off.

The results of this thesis show that hiring a real estate agent has no significant effect on the discount, which means that there is not sufficient evidence that real estate agents outperform private buyers or vice versa. It also means that expert knowledge does not seem to result in a higher discount.

Overall, it can be said that it is advisable to hire a real estate agent to buy a house at auction because the real estate agent takes care of the buying process, which could be valuable for other reasons, such as a reduction in risk.

The discount on auctioned properties can mainly be explained by the reason for the sale of the property and the property characteristics. The properties that are forced sold have a negative significant effect on the discount, at 95% confidence level. This means that properties that are forced sold have a discount that is on average 7.9% lower than properties that are sold voluntarily. Also, the reason sale under court order has a negative significant

(10)

10 effect on the discount of auctioned properties. If a property is sold under court order, the discount is on average 16.3% lower than when a property is sold voluntarily. The reason ‘sale due to bereavement’ also has a negative effect on the discount. Properties that are sold with this reason are sold with a discount that is on average 8.1% lower than voluntary sales. However, the coefficient ‘sale due to bereavement’ is not statistically significant. The results of this thesis on the reason for sale on the auction discount show that the reason for sale has an influence on the auction discount and that it adds a new determinant to the factors that determine the auction discount.

The reason for the difference in auction discount for the reasons for sale could be explained by the overdue charges. The auction books do not mention the overdue charges clearly and this could result in the fact that bidders do not have a clear idea of the overall costs. It would be useful for bidders if the auction house makes an overview of the overall costs for a property to solve this problem.

The remainder of this thesis is structured as follows. Chapter 2 discusses the institutional setup. Chapter 3 includes a literature review, where different determinants of the discount for properties sold at the auction are discussed. An explanation of how the discount and proxy for the market value of the auctioned properties are determined is given next in chapter 4. Here, it is also explained how the determinants of the discount are tested. Chapter 5 discusses the data that are collected from ‘de Eerste Amsterdamse Onroerend Goed Veiling’, the land registry and the Dutch Association of Realtors (NVM). Chapter 6 reviews the results of the regression output. Chapter 7 elaborates on the limitations and gives advice with regard to future research. In chapter 8 a conclusion is presented and the results are discussed.

(11)

11

2. Institutional setup

The Dutch real estate market is a specific market and is comparable with other countries. Around 60% of the properties in the Netherlands are owned by the people who live there (CBS, 2014). In the Netherlands it is possible to buy a property with a loan-to-value ratio above 100%, therefore it is not compulsory to make a down payment (Rijksoverheid, 2015). This is also one of the main reasons why 1.5 million properties in the Netherlands went “underwater” (Centraal Bureau voor de Statistiek, 2014). After the sale of a property at auction as ‘a forced sale’, the person will be jointly and severally liable to the bank, hence there is no foreclosure. This is different from the United States, where it is possible to hand in the keys of the property to the bank, and then the mortgage is paid off. In the Netherlands joint and several liability has the result that the borrower has to pay the residual debt back to the bank, no matter what.

Selling and buying at ‘de Eerste Amsterdamse Onroerend Goed Veiling’ (EAOGV) can be executed by brokers, members of the brokers association Amsterdam, private persons or money supplying institutions (EAOGV, 2015). At ‘de Eerste Amsterdamse Onroerend Goed Veiling’ two auction methods are combined. When the auction of a property begins, the bidding starts. The bid will increase in steps and the highest bidder is the provisional buyer of the property. This is called the English auction model. The provisional buyer signs the auction contract and the auctioned property will be auctioned again. In this round, the price starts above the price of the highest bid and the auctioneer will decrease the price in steps toward the bid of the provisional buyer. This is called the Dutch auction method. If someone wants to pay the amount that is called by the auctioneer which is above the provisional buyer’s price, then the bidder can call ‘mine’ and the property is sold to the bidder in the second round.

To bid at auction people have to sign a contract with a bank guarantee worth 10% of the price of the auctioned property, which states that the people are able to pay the amount that they bid at the auction. Likewise, people have to prove their identity. If people bid by themselves and are not able to pay the bid, they are responsible for the costs the seller made at auction. According to the general terms and conditions of ‘de Eerste Amsterdamse Onroerend Goed Veiling’, the real estate agent who buys on behalf of his client is jointly and severally liable for the auction buyer’s obligations. Hence, if the real estate agent’s client is not able to buy the property, then the broker is regarded as the purchaser. This is a risk for the real estate agent and as a result, real estate agents need to check the financial situation of their clients.

One of the advantages of the ‘de Eerste Amsterdamse Onroerend Goed Veiling’ is that auction books appear weekly. All relevant information about upcoming auctions and

(12)

12 the sold properties are included. This makes ‘de Eerste Amsterdamse Onroerend Goed Veiling’ a transparent and accessible auction house for real estate agents and private buyers, despite that the overdue charges are not always mentioned clearly in the auction books.

(13)

13

3. Literature review

In this chapter the related literature is discussed. First, the effect of expert knowledge of real estate agents on the price of properties is discussed. The second paragraph discusses the price of a property with regard to the time it is on the market. After that, the auction prices versus regular sale prices are discussed. In the fourth paragraph research in the Netherlands on regular sale prices and auctioned prices is studied. The fifth paragraph comprises a conclusion of the literature review.

3.1 The effect of knowledge on the price

When people buy a property in a regular sale, hidden defects have to be repaired by the party that is selling the property (Vereniging Eigen Huis, 2015). During an auction, buyers have to rely on the information that is given from the party that is selling, and hidden defects are not repaired by the seller (Vereniging Eigen Huis, 2015). Therefore, when it comes to buying a property at auction, it is advisable to hire a real estate agent with expert knowledge about the neighbourhood and possibly about the maintenance levels of the properties there. The real estate broker can have an information advantage over private buyers, especially when in most cases the level of maintenance is not mentioned in the auction forms (NVM, 2015). Moreover, it could also be possible that the property is rented to a third party. If this is not mentioned in the auction forms, it could result in a bid that is too high, because the costs to remove the renters from the property has to be paid by the buying party.

Milgrom and Weber (1982) investigated the value of information in a sealed-bid auction. One bidder had public information and another bidder had private information. They found that the bidder with only public information did not make a profit at equilibrium, while the bidder with the private information made positive profits. In addition, the informed bidder’s profit increased when more information was received. However, the authors did not investigate the value of information in an open-bid auction.

Levitt and Syverson (2008) conducted research on the effect of real estate brokers on the sale of a property. Levitt and Syverson (2008) found strong evidence that properties are sold faster when a real estate agent was hired and that homes owned by real estate agents sell at a price of 3.7% more than other properties. In the study, the authors focused only on the selling party but did not investigate the effect of real estate agents on the price for the purchase of a property nor the impact of broker expertise at auctions.

3.2 Price versus time on market trade-off

The research of Anglin, Rutherford and Springer (2001) investigated the trade-off between the price and the time that a residential property is on the market. When a property is to be

(14)

14 sold, the owner of the property must choose the initial offering price. The findings of Anglin et al. (2001) show that the price level of the property has influence on its marketability. The seller of the property faces a trade-off between a higher listing price and the time that the property is on the market. In the case of a higher listing price, the property will be on the market longer; and if the price is lower than the initial value, the property will be sold in less time. The listing price is one of the critical factors when it comes to buying a property. If the listing price is in the buyer’s target range, then the buyer can decide to make a bid.

Jud, Seaks and Winkler (1996) investigated the impact of brokers and marketing strategy on the time that a residential property is on the market. They found that marketing strategies that are related to the pricing of a property significantly influence the time on the market. This is in accordance with the results of Anglin et al. (2003). The study of Jud et al. (1996) finds no evidence that particular real estate brokers are able to sell a property within a shorter time. Also, the research of Ong and Koh (2000) shows that the time on the market is associated with greater capital gains.

Because properties sold at auction have to be sold directly, the time on the market is very short and the expected price will be lower. Next to that, the marketing strategies that are used for properties that are sold at auction are limited. The amount of available information that is given to potential buyers is often limited and not always complete. Existing literature looks at the length of time on the market as a variable to explain the discount on properties sold, which is not possible for properties that are sold at auction.

0

Expected time on market

Figure 1: Expected sale price vs. expected time on market

(15)

15

3.3 Auction Prices versus Regular Sale Prices

Extensive research has been done on the differences in prices between auctions and regular sales. Milgrom and Weber (1982) found that if buyers bid under competition, the transaction price will end up higher than when people negotiate in a closed circuit. The auction is a popular method for selling property because auctions lead to stable and efficient allocation outcomes. This is because the buyer is in a weaker negotiation position than buyers that negotiate in a closed circuit (Milgrom, 1987).

This method, the auction, is also used in the Dutch real estate market. Potential buyers have to give a written letter with their bid to the real estate agent before the deadline expires. After the expiration date, the real estate agent compares the bids, and the person with the highest offer will become the new property owner. By using competition and time pressure the transaction price will be maximized.

However, empirical studies that compare property returns between auctioned sales and regular sales, show that there is a discount on auctioned sales. William and Marvin (1996) investigated the relationship between foreclosure status and apartment prices, with a hedonic regression model. “An analysis of apartment sales in Phoenix, Arizona shows that foreclosure-status apartments sell at a discount of 22% compared to regular sales of apartments. This is done with the intention to control for differences in quality between foreclosure and non-foreclosure apartments.”

Mayer (1998) also investigated the difference between auctioned property prices and properties sold in regular sales in Los Angeles and Dallas during the 1980s. His results show that there is a discount of 0% to 9% for properties sold in Los Angeles, while properties sold in Dallas had discounts of 9% to 21%.

Allan and Swisher (2000) investigated price differences for Miami-Dade, Palm Beach and the West Coast. The results show that property prices are on average 17.45% less when a property is sold at auction when compared with a regular sale.

The discussed studies did not examine the properties in the Netherlands, and they only focused on foreclosure-status apartments. The following paragraph elaborates on the price difference between auction prices and regular sale prices in the Netherlands.

(16)

16

3.4 Auction Prices versus Regular Prices in the Netherlands

Brounen and Rijk (2009) have compared auction prices to the estimated market value of properties in the Netherlands. After examining 290 ‘forced sale’ transaction prices of their sample, they found a discount of 37% compared to the predicted market price. The market price was predicted by a hedonic regression model. The findings of Brounen and Rijk (2009) agree with the results of William and Marvin (1996), Mayer (1998), and Allan and Swisher (2000), but the discount is higher than that in previous studies. The discount is highest for properties built before 1945 and for upstairs apartments that are located in less affluent areas. This is because people expect that older properties and properties in poorer locations need more maintenance than newer properties or those built in more affluent locations. The reason for the discount can be explained by the asymmetric information about the auction costs and the level of maintenance. The auction transaction costs differ among auction properties and are often poorly documented. Brounen and Rijk (2009) found that the costs range between 0,5% and 3,0%.

However, Brounen and Rijk (2009) did not investigate other reasons for sale at auctions, nor did they examine whether or not the expertise from a real estate broker has an impact on the auction price. Furthermore, Brounen and Rijk (2009) investigated the entire country and did not focus on one city. By focusing on ‘de Eerste Amsterdamse Onroerend Goed Veiling’ in Amsterdam, which documented the auction transaction costs of the auction well, we can exclude the asymmetric information about auction costs.

3.5 Conclusion literature review

Previous findings about the differences between auctions and regular sales show the presence of a discount. However previous research did not look further into the effect of expert knowledge and experience of real estate agents on the purchase of auctioned properties. In addition, previous findings do not yet explain the effect of the reason for sale. The study by Brounen and Rijk (2009) did not specify the reason for selling at auction. The reasons for selling that are investigated in this thesis are: forced auctions, auctions due to bereavement, voluntary auctions and public auctions under court order. These reasons are investigated in this thesis with the intention of providing new insight into auctions in Amsterdam.

(17)

17

4. Methodology

In this chapter the methodology that is used to determine the discount is discussed. Also, the method of testing the value of expert knowledge and the reason for sale is explained. The prices of the properties that are sold at auction need to have a market value to calculate the difference between the auctioned price and the market value. The market value will be predicted using the data from the Dutch Association of Realtors (NVM).

4.1 Predicting the market value of auctioned properties

To predict the market value for the auctioned properties, a hedonic regression model is used. The hedonic regression model captures the relation between the transaction price and the characteristics of a specific property. By making use of real market transactions the implicit price of the properties that are sold at the auction can be priced. The characteristics function as independent explanatory variables and may decrease or increase the property value. The hedonic regression model values each component separately and estimate a price for each determinant. In this case, the value of the property depends on property characteristics, location characteristics, and time characteristics. The more reference transactions are collected, the better the implicit property price can be estimated, meaning the transaction noise is minimized, because there could be transactions that are above or below the real market value.

By filling in the characteristics of the property in the estimated regression model, the predicted market value is calculated. Since the price and the numerical value of area in square metres are not linear functions, log (price) and log (square metres) are used.

Other characteristics that are being used to predict the market value of the property are: location, year of sale, property type, parking availability, garden availability, ground lease conditions and if the property is rented or not. For the characteristics: location, property type, parking, garden, rental status and ground lease, dummy variables are assigned. This is quite a standard hedonic setup.

The effect on the price for an extra square metre would be different for each location, because some locations are more popular than other locations. This effect is called ‘interaction effect’. By including the interaction term log (square metres) x location, the effect would be taken into account.

Another interaction effect that has to be taken into account, is the value of parking places. Parking is very expensive in the centre of Amsterdam, and is cheaper in other parts of the city. There are also less parking spaces in the centre than in other parts of Amsterdam.

(18)

18 Hence, a parking place in the centre of Amsterdam is more valuable than a parking place located outside the centre. The interaction term location x parking is added to the regression model to take this effect into account.

The final hedonic regression model that is estimated is:

1) Log (priceit) = β0 + βi1Building periodi + βi2Locationi + βi3log (square metresi) +

βi4transaction yearit+ βi5Property typei + β6Parkingi + β7Gardeni + β8Rentedi + β9 No Ground

leasei + βi10(Log (Square metresi) x Locationi)) + βi11(Locationi x Parkingi) +αi

+τ

t + εit

Here, Log (priceit) is the estimated market value of a property. Building period is an

indicator in which building period the property was built. Location is an indicator in which part of Amsterdam the property is located. Log (square metres) is an indicator of the logarithm of the number of square metres of a property. Transaction year is an indicator of the transaction date of the property. Property type indicates which type of property is sold. Parking is an indicator whether there is a parking place, and garden indicates whether there is a garden. Rented indicates if the property is rented to a third party. No ground lease indicates whether no ground lease has to be paid. Log (Square metres x Location) is the interaction effect between the number of square metres and location. Location x Parking is the interaction effect between the location of a property and the availability of a parking place.

α

i and

τ

t are included to take account of fixed effects for property specific

characteristics and time characteristics. εit is indicated as the error term.

4.2 Making a proxy for the discount

After making the hedonic regression model, the variables of the properties which are sold at auction is entered into the model to estimate the predicted market value. After using the hedonic regression model to predict the market value, the difference between the proxy for market value and auction value will be generated. The difference between the market value and the auction value is the discount that is expected for properties sold at auction.

Once the discount is predicted, there will be a test for the determinants of the difference. These determinants are the reason for sale and the expertise of the real estate agent. For the characteristics ‘sale under court order’, ‘forced sale’, ‘sale due to bereavement’, and purchased by a real estate agent, dummy variables are assigned. Also, the variables building period, location, log (square metres), transaction year, property type, parking availability, garden and the rental status of the property, are included as control variables. In principle, if auction prices and market prices are affected similarly, this will not be necessary. However, because there are different properties and different discounts

(19)

19 for properties, the control variables are taken into account. This results in the following regression model:

2) Log (differenceit) = β0 + βi1Real estate agenti + βi2Forced salei + βi3Sale due to

bereavementi + βi4Sale under court orderi + χit‘β + αi+ τt+ εit.

Where Log (difference) is the predicted auction discount of properties that are sold at the auction. Real estate agent is an indicator whether a property is bough by a real estate agent. Forced sale is an indicator whether a property is sold with the reason ‘forced sale’. Sale due to bereavement is an indicator whether a property is sold with the reason ‘sale due to bereavement’. Sale under court order is an indicator whether a property is sold with the reason ‘sale under court order’.

χ

it is an indicator if the control variables are taken into

account.

α

i and

τ

t are included to take account of fixed effects for property specific

characteristics and time characteristics. εit is an indicator for the error term.

Several versions of this model are tested. The first three models test the relation of the effect of hiring a real estate agent on the discount of auctioned properties. The second model includes clustered robust standard errors and the third model includes robust standard errors and time effects. The fourth model tests the reasons for sale on the discount of auctioned properties. The fifth model tests the relation between the reasons for sale and the effect of hiring a real estate agent on the discount. The sixth model includes all the control variables and the seventh model tests the same but takes into account random effects. All six regression models include the effect of a property that is rented to a third party, because a property that is rented to a third party has a much lower market value than an empty property. Additionally, more rented properties are sold in the data sample collected from ‘de Eerste Amsterdamse Onroerend Goed Veiling’ than in the data sample received from the Dutch Association of Realtors.

The variable ‘real estate agent’ is tested for statistical significance. If the output shows that the variable is positively significant, it means that real estate agents make better deals at auctions than private buyers.

The variables, ‘sale under court order’, ‘forced sale’ and ‘sale due to bereavement’ are tested on their statistical significance. If these variables are statistically negatively significant then it provides evidence that the discount is lower for these types of sale than for ‘voluntary sales’.

The variables ‘voluntary sale’ and ‘purchased by a private buyer’ are not included in the regression model, because these two variables function as reference categories for the variables that are included in model 2.

(20)

20

5. Descriptive statistics

This paragraph describes the data that is collected from auction house ‘de Eerste Amsterdamse Onroerend Goed Veiling’, the land registry, and from the Dutch Association of Realtors (NVM).

The data from ‘de Eerste Amsterdamse Onroerend Goed Veiling’ is not electronically available and has to be collected manually from the weekly auction books. The sample period is from January 2005 until December 2014. The variables that are collected from ‘de Eerste Amsterdamse Onroerend Goed Veiling’ are:

 address  transaction date  transaction price  auction cost  overdue charges  type of property  number of rooms  number of square metres  parking place availability  garden availability  ground rent status  rented or not rented  property buyer  the reason for selling

The age of the building, postal code and the number of square metres are available on the website of the land registry. This data can be found by typing in the address of the property.

The collected dataset that is being used contains 754 property price transactions. All properties that are sold are located in Amsterdam. Properties with commercial uses and industrial uses are not taken into account. The auctioned properties with a price above 1 million euros and above 400 square metres are eliminated. This excludes the outliers in the dataset. After eliminating the outliers in the dataset, the final dataset consists of 747 observations.

To predict the market value of the auctioned properties, the data from the Dutch Association of Realtors is used. Since 1985 all members of the Dutch Association of Realtors exchange information about property price transactions that is saved in the database of the Dutch Association of Realtors. Besides the transaction prices, the property

(21)

21 characteristics, neighbourhood characteristics and time characteristics are saved in the database. The dataset that is used for this thesis contains all property price transactions of the properties that are sold in Amsterdam in the period between 2005 and 2013.

The total number of property price transactions in Amsterdam is 131.000. After eliminating the types of properties that are not in the dataset (e.g. service apartments and entire properties), and the outliers, the dataset is reduced to 51.032 property price transactions. Unfortunately, the year 2014 is not available in the dataset of the Dutch Association of Realtors. To make use of the collected data for the year 2014 from ‘de Eerste Amsterdamse Onroerend Goed Veiling’, the average property price increase in Amsterdam (which is determined by the Dutch Association Realtors) is used. The average property price increase in 2014 was 1.6% (NVM, 2014).

5.1 Auction discount

The data on the costs of the auctioned properties from ‘de Eerste Amsterdamse Onroerend Goed Veling’ is summarized in Table 1. Table 1 shows that the average buying price of the properties sold at the auction is €183.187 with a standard deviation of €110.216. The lowest price in the sample is €50.000 and the highest price in the sample is €810.000. If people hired a real estate broker, they pay on average a 1.79% fee to the broker (Makelaarsland, 2015). The broker fee is on average € 3.279 with a standard deviation of €1.973. The maximum broker fee was €14.499. Besides the broker fee, it is possible that due payments have to be paid. The due payment is on average €1.417 with a standard deviation of €2.168. The highest due payment in the sample is €17.200. The auction costs for the ‘de Eerste Amsterdamse Onroerend Goed Veiling’ are €1.500 for properties with a price below €525.000, while above €525.000 the auction costs are €1.500 till 525.000 plus 0.4% for the remaining amount. The average auction cost is €1510 in the sample.

Table 1: Summary statistics of the costs of the auctioned properties.

Notes: Based on data from 2005 until 2014 for houses sold at auction in Amsterdam. The total amount of transaction is 747.

Variable Mean St. Dev. Minimum Maximum

Buying price € 183.187 € 110.216 € 50.000 € 810.000 Auction costs € 1.510 € 81 € 1.500 € 2.640 Due payments € 1.417 € 2.168 € 0 € 17.200 Broker fee € 3.279 €1.973 € 895 € 14.499 Total buying price € 189.393 €112.054 € 52.395 € 827.139

(22)

22 To predict the market value of the auctioned properties the coefficients of the properties are regressed based on the logarithm of the transaction price of the properties. Table A in the appendix shows the coefficients of transactions in the period 2005 until 2013 in Amsterdam. The R2_{shows that 88% of the variation of the property price is predicted by}

the variation of the coefficients that are used in the hedonic regression model. One can see in Table A of the appendices that the size of the dwelling and the type of property have a significant effect on the price of the properties. Larger properties are selling for a higher price on average. Detached dwellings are more valuable than other types of dwellings. Gallery apartments/block apartments have the lowest value in comparison with other types of properties. The transaction year has also a significant effect, as already expected, the years 2005 to 2008 have a more positive effect on the price then the years after the financial crisis of 2008. The year 2014 was not available among the data from the Dutch Association of Realtors. The property price increase of 1.6% for the year 2014, which is predicted by the Dutch Association of Realtors, is used to make a prediction for the market value of the auctioned properties in 2014.

Having a garden, having a parking place and not having to pay ground rent are also significantly positive influences on the transaction price. If a property is rented to a third party, this has a significantly negative effect on the property value.

Properties built in the period from 1960 until 1989 negatively influence the transaction price and properties in the period before 1945 and those built after 1990 are positively influencing property prices.

A property sold in the centre of Amsterdam or Amsterdam South has a more positive effect on the price than a property sold in Amsterdam South-East. After generating the interaction term size x location, the location coefficients change. The centre of Amsterdam then has a negative effect on the price. The interaction term location x size positively influences the price. The reason for the change of location coefficients is that properties built in the centre are built before 1945 and are smaller than properties that are built in other locations. The interaction terms of size with other locations is negative for North and North-West, while the other interaction terms positively influence the price. A parking place in the centre is more expensive than in the other locations. The only place where the interaction location x parking is higher is in Amsterdam West.

After the coefficients of the data from the Dutch Association of Realtors were generated, the market value for properties sold on the auction is predicted. The average market value of the properties that are sold at auction is €289.089. The discount is calculated by subtracting the market value from the price of the auctioned properties. The discount for auctioned properties is on average 34,5% for the entire dataset. Table 2 shows

(23)

23 that the average discount for ‘forced sales’ is 34% and the discount for voluntary sales is on average 37%. This shows that the results of this research are in accordance with the results of the research of Brounen and Rijk (2009), who found that the average discount of properties sold at the auction was 37%.

The reason for this small difference in the discount for ‘forced sales’ and ‘voluntary sales’ could be that most people do not take the overdue charges into account. The properties that are sold voluntarily are on average more expensive than properties that are sold as ‘forced sales’. In addition, Table B in the appendix shows the differences in buying behaviour between real estate agents and private buyers. Private buyers are more likely to purchase properties sold as ‘forced sale’, properties that are not rented and properties located in South-East, while real estate agents are more likely to purchase properties that are sold as ‘voluntary sale’, properties that are rented and properties located in New-West. The difference in buying behaviour has an effect on the discount of auctioned properties. Therefore the discount between real estate agents and private buyers differ.

Semi-detached properties are sold with the largest discount if the property is a forced sale (43%) and corner houses are sold with the largest discount if the property is voluntarily sold (55%), followed by terraced houses and gallery/block apartments. The smallest discount is for upper-floor apartments, 32% discount for ‘forced sales’, and a 33% discount for ‘voluntary sales’. There were no voluntary sales of semi-detached houses, thus these are not taken into account.

Properties that are forced sold in the period after 2000 are sold with the highest discount and the forced sold properties in the period 1960 - 1970 are sold with the smallest discount. This is not the case for voluntary sales; properties sold in the period 1945 - 1959 were sold with a discount of 47%, while the lowest discount is for properties in the period 1971 - 1980, 35% respectively.

Larger properties are sold with a higher discount than smaller properties. Properties with less than 50 square metres are sold with a discount of 20% for ‘forced sales’ and 26% for ‘voluntary sales’. Properties sold as ‘forced sales’ with 100 – 125 m2_{and 126 – 150 m}2

have the highest discount (37%), and properties sold as voluntary sales with more than 150 square metres are sold with a discount of 47%.

As already discussed, the location in Amsterdam also influences the discount. Properties that are sold as ‘forced sales’ in the centre of Amsterdam are sold with an average discount of 31%, and properties that are voluntarily sold have a discount of 25%. The highest discount is for properties that are sold as ‘forced sales’ in Amsterdam West, with a discount of 39% and properties that are voluntarily sold in Amsterdam-South that have a discount of 48%.

(24)

24

Table 2: Market value in comparison with the average transaction prices of properties at auctions.

Notes: Based on data from 2005 until 2014 for houses that are sold at auction in Amsterdam. The total number of observations is 747. The number of forced sales is 591 and the number of voluntarily sales is 132.

Forced sales (n=591) Voluntarily sales (n=132) Market value Auction price Difference Market value Auction price Difference Type of property Gallery/Block apartment € 238.385 € 155.214 -35% € 284.817 € 178.275 -37% Corner house € 343.110 € 226.725 -34% € 537.268 € 242.912 -55% Detached house € 648.791 € 457.549 -29% € 367.989 € 181.414 -51% Ground-floor apartment € 329.395 € 226.318 -31% € 359.805 € 227.550 -37% Terraced house € 316.887 € 200.474 -37% € 555.284 € 343.307 -38% Semi-detached house € 684.197 € 420.549 -39% - - - Upper-floor apartment € 315.536 € 215.669 -32% € 292.465 € 195.773 -33% Building period Before 1945 € 338.455 € 230.010 -32% € 344.503 € 219.887 -36% 1945-1959 € 208.356 € 142.304 -32% € 221.234 € 138.967 -47% 1960-1970 € 214.617 € 149.197 -30% € 221.484 € 120.763 -45% 1971-1980 € 199.981 € 125.408 -37% € 185.335 € 119.576 -35% 1981-1990 € 228.766 € 149.742 -35% € 354.880 € 213.987 -40% 1991-2000 € 337.218 € 217.072 -36% € 310.642 € 189.812 -39% After 2000 € 359.690 € 227.340 -47% € 418.833 € 239.519 -43% Size of the property less than 50 m2 _{€ 149.672} _{€ 120.231} _-20% _{€ 188.683} _{€ 139.660} _-26% 50 - 75 m2 _{€ 212.460} _{€ 140.881} _-34% _{€ 249.080} _{€ 151.440} _-39% 76 - 100 m2 _{€ 270.243} _{€ 174.832} _-35% _{€ 298.896} _{€ 243.757} _-18% 101 - 125 m2 _{€ 334.066} _{€ 209.589} _-37% _{€ 391.360} _{€ 234.996} _-40% 126 - 150 m2 _{€ 438.570} _{€ 275.781} _-37% _{€ 523.761} _{€ 335.262} _-36% more than 150 m2 _{€ 660.015} _{€ 466.643} _-29% _{€ 789.793} _{€ 421.895} _-47% Location Centre € 364.748 € 251.385 -31% € 375.419 € 283.435 -25% New-West € 266.767 € 177.847 -33% € 353.517 € 198.769 -44% North € 267.642 € 167.254 -38% € 247.310 € 148.893 -40% East € 363.198 € 240.496 -34% € 388.912 € 251.731 -35% West € 285.452 € 173.751 -39% € 285.896 € 160.079 -44% South € 416.114 € 303.281 -27% € 411.769 € 261.277 -37% South-East € 214.159 € 135.514 -37% € 270.359 € 139.352 -48% Total average € 276.860 € 182.237 -34% € 340.203 € 214.908 -37%

(25)

25

5.2 The person who buys the property and the reason for sale

The data that is collected from ‘de Eerste Amsterdamse Onroerend Goed Veiling’ contains information on buyers of the property, who are real estate brokers and private buyers. Between the years 2005 and 2014 in the sample, there were 117 purchases by private buyers of auctioned property, which is around 16% of all collected transactions, and 630 times a real estate agent bought an auctioned property, which is around 84%. There are not enough transactions in the dataset on sales due to bereavement, or sales under court order to make a proper summary of these types of sales; therefore this is not taken into account in Table 2.

The financial crisis did not have an impact on the ratio between the properties bought by real estate agents and those bought by a private buyers. Around 85% of the properties were bought by real estate agents. Since the year 2007 there is a downward shift in the percentage of properties bought by real estate agents. During the crisis, more private buyers decided to purchase property independently, and since the year 2012 there is an upward shift in the percentage of properties bought by real estate agents.

Figure 3 shows that the reason for sale has changed over the years. In the years from 2005 until 2007, there is a decline in the percentage of ‘forced sales’ and an increase in the percentage of ‘voluntary sales’. The ‘forced sales’ were around 55% in 2007, and the ‘voluntary were around sales’ 40% in 2007. The crisis of 2008 increased the percentage of ‘forced sales’ to 95%, and the other 5% were ‘voluntary sales’. In the years from 2010 until 2014, the number of forced sales was stable, at around 70%.

Figure 2: Percentages of transactions per year for the reasons for sale, 2005-2014.

0% 20% 40% 60% 80% 100% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Forced sale Sale due to bereavement Sale under court order

(26)

26

5.3 Control variables

Control variables are added to the auction regression model to control for exogeneity. The sample period that is investigated is the period between January 2005 and December 2014. As already mentioned, the financial crisis had a big impact on the value of properties, which can be seen in the graph from Centraal Bureau voor de Statistiek (2015). From 2001 until 2007 the property prices increased continuously. After the crisis, the property prices decreased and became more stable in 2013 and 2014. Because the year of the auction influences the price, time fixed effects will be taken into account in the regression model.

Figure 3: Price Index of existing properties, 2010 = 100, 2005-2014. (CBS, 2015).

Notes: Based on data from the Centraal Bureau voor de Statistiek (2005) for the period 2005 until 2014. The index is based on the value of properties in the Netherlands.

The location of the properties is an important determinant of the value, therefore this will be included in the hedonic regression model. This will be done by using the postal codes of all 747 properties in the sample. Amsterdam is the largest city of the Netherlands and has 810.937 citizens (CBS, 2014) with a total surface of 219 square kilometres. The city can be split into eight parts: Amsterdam Centre, Amsterdam East, Amsterdam North, Amsterdam New-West, Amsterdam West, Amsterdam Westpoort, Amsterdam South and Amsterdam South-East. The centre of Amsterdam is the oldest part of Amsterdam and dates back to the late Middle Ages and the Golden Age. The newest part of Amsterdam is Amsterdam IJburg. Most of the properties sold are located in Amsterdam South-East. There are no auctioned sales in Amsterdam Westpoort. The number of sales per location ranges between 53 and 218 over the sample period.

The type of the property is also an important determinant of the property value. Detached houses have a higher value than semi-detached houses and apartments have a higher value than upstairs apartments. Different types of properties are used as determinants to calculate the expected market value of the auctioned properties.

0 20 40 60 80 100 120

(27)

27 The types that are investigated are apartments, upstairs apartments, downstairs apartments, corner houses, detached houses, row houses and semi-detached houses. Table 3 shows that most of the properties that are sold in the sample are apartments (41.60%). The semi-detached houses are the least frequently sold, only sold four times in the sample.

Table 3: Frequencies of type of properties and locations, 2005-2014.

The properties in the centre of Amsterdam are older than the properties in the North of Amsterdam. The age of the buildings in the sample ranges between 1602 and 2010. Newer buildings have lower maintenance costs than pre-war properties. However, some people prefer living in the centre of Amsterdam. This could influence the bidding by potential buyers. One can see in Table 3 that properties built in the period between 1940 and 1960 were only sold 35 times at auction and that most of the properties that were sold were built in the period between 1980 and 2000.

The number of square metres is a criterion from the Dutch Association of Realtors to measure the value of property. This is taken into account in the hedonic regression model used. The number of rooms is also available. However, using both determinants as control variables does not make sense, therefore the number of rooms is not taken into account – because bigger properties usually have more rooms than smaller properties.

The average property size in the sample is 114 square metres, with a standard deviation of 71 square metres. The lowest number of square metres in the sample is 26 and the highest number of square metres is 400.

Type of property Frequency Percentage Location Frequency Percentage

Apartments 384 51.5% Centre 85 11.4%

Corner houses 32 4.3%

New-West

145 19.4%

Detached houses 15 2.0% North 61 8.2%

Ground-floor apartments

37 5.0% East 53 7.1%

Terraced houses 81 10.7% West 106 14.2%

Semi-detached houses 4 0.5% South 79 10.6%

Upper-floor apartments 194 26.0% South-East

218 29.1%

(28)

28 Parking is expensive in the centre of Amsterdam with the price per hour at €5, while the price for parking in other places ranges between €1.40 and €4. The availability a parking place nearby a property is valuable and will have an effect on a property bid at auction. Only 11% of the properties in the sample have their own parking space nearby. This is taken into account in the hedonic regression model. This is also the case for gardens in Amsterdam. Amsterdam has a high population density which influences the number of gardens for properties. Not all properties have a garden in Amsterdam. 24% of the properties in the sample have gardens.

The municipality of Amsterdam owns around 80% of all the land in Amsterdam. A large part of this land is leased by individuals and companies. The municipality remains as landowner while the individuals and companies have the right to use the land and the users have to pay ground rent to the municipality in return. The canon is a business right and this means that the rights are connected to the land and not to the resident. The rights to use the land can be passed to another resident without asking permission from the municipality. In the sample, 34% of the land is owned by individuals, 32.5% of the land leases are paid off, and 33.5% of the ground leases are paid every year. If a property is sold while it is rented to a third party, then the price of the property will be lower than when the property is not rented. In the sample, 22.5% of the properties that are sold are rented to third parties.

In principle, the control variables used in the auction regression model should fit with the variables used to predict the market value of the auctioned properties. The expectation is that this is not the case because the samples differ, as demonstrated in Table C in the appendix. The properties that are sold at auction have on average a higher market value than the properties in the sample from the Dutch Association of Realtors. This could be the result of the lower number of observations that are collected from ‘de Eerste Amsterdamse Onroerend Goed Veiling’. Table C in the Appendix also shows that there are more properties that are rented to a third party in the data sample of ‘de Eerste Amsterdamse Onroerend Goed Veiling’. If a property is rented this has a significant effect on the property price. The regression results in Table A of the appendices show that the price of a property is on average 23,6% lower if a property is rented to a third party. Therefore, this will be taken into account in the auction regression models.

(29)

29

6. Results

In this chapter the regression results are presented. As already described in the chapter about descriptive statistics, there is an average discount of 34.5% on the properties that are sold at the auction. This discount made it possible to regress the value of expert knowledge in the field and the reason for selling a property at the auction on the discount of the auctioned properties.

The variable ‘private buyer’ is used as a reference category to estimate the value of expert knowledge, and the coefficient ‘real estate agent’ is regressed based on the logarithm of the discount. The results are summarized in Table 4. Each column reports a different regression and each row reports a coefficient estimate and the standard error.

Column 1 in Table 4 presents the results of the variable ‘real estate agent’ on the logarithm of the discount without location effects and time fixed effects. The coefficient on ‘real estate agent’ is negative (-0.041), which means that hiring a real estate agent results in a discount that is on average 4.1% lower than for private buyers. However, the coefficient is not statistically significant which means that there is not enough evidence that private buyers outperform real estate agents with expert knowledge. The regression in Column 2 shows the effect of adding clustered standard errors. This is taken into account to allow for heteroskedacity within an entity, even though the errors are treated as uncorrelated across entities.

The results in column 2 show that there were no changes when clustered standard errors were added. In column 3 the time fixed effects are added. The coefficient ‘real estate agent’ now less negatively influences the discount of auctioned properties (-0.013). Figure 4 shows the importance of including time fixed effects. The financial crisis had a positive significant effect on the discount of auctioned properties, which means that the discount increased during the crisis and it shows that bidders are taking fewer risks in times of financial crisis. In 2005 and 2006, the discount on auctioned properties was stable, but the discount increased during the period 2007 to 2012 and decreased in 2013 and 2014. The year 2012 had the largest effect on the discount; the discount was around 33% higher than in 2005. However, the coefficient ‘real estate agent’ is not significant in the regression of column 3.

(30)

30

Figure 4: The effect of time on the discount of auctioned properties.

Notes: Based on the estimated coefficients of the time fixed effects from 2005 until 2014 for houses that are sold at auction in Amsterdam. The total number of observations is 747.

Column 4 of Table 4 presents the results of the variables for the reason of selling on the discount of auctioned properties. The coefficient on ‘forced sales’ is negatively influencing the discount (-0.098), which means that the discount of properties that are sold with the reason ‘forced sale’ result in a lower discount than properties that are sold voluntarily, although the coefficient is not statistically significant. The other reasons that are regressed on the discount of auctioned properties are the reason ‘sale due to bereavement’ and ‘sale under court order’. The coefficient ‘sale due to bereavement’ is also negatively influencing the discount of auctioned properties (-0.224) and is significant at the 95% confidence interval. The reason ‘sale under court order’ is negatively influencing the discount of auctioned properties (-0.129) and is significant at the 90% confidence interval. This means that the discount on auctioned properties is higher for voluntary sales, which is in line with the expectations.

Column 5 presents the results on the coefficients reason for sale and real estate agent. The coefficient on ‘real estate agent’ is still negatively influencing the discount (-0.048) and is now significant at the 90% confidence interval. This significant effect is not in line with the expectation that expert knowledge would result in a higher discount. Also, the reasons for sale are still negatively influencing the discount; “forced sales” (-0.010), “sale due to bereavement” (-0.234), and “sale under court order” (-0.138). The coefficients ‘forced sale’ and ‘sale due to bereavement’ are significant at the 95% confidence interval and the coefficient ‘sale under court order’ is significant at the 90% confidence interval.

0,00% 5,00% 10,00% 15,00% 20,00% 25,00% 30,00% 35,00% 2004 2006 2008 2010 2012 2014 2016

(31)

31 Column 6 of the regression in Table 4 include additional potential determinants of the discount on auctioned properties along with location effects and time effects. The variables that are included are related to the characteristics of the property. The regression in column 6 shows interesting results, because including the additional variables reduces the effect of the variable ‘real estate agent’. Hiring a real estate agent results in a discount that is on average 1.3% lower. The coefficient is not statistically significant, which means that there is not enough evidence that private buyers outperform real estate agents with expert knowledge in the field or vice versa. The coefficients for the reason for sale also changed. The reason ‘forced sale’ decreased (-0.079) and is still significant at the 95% confidence interval. The reason ‘sale due to bereavement’ increased (-0,081) and is not statistically significant anymore. The coefficient of the reason ‘sale under court order’ decreased (-0.163) and is significant at the 95% confidence interval.

Column 6 shows that size of the property has a positively significant effect on the discount, at the 99% confidence interval. This means that larger properties are sold with a higher discount than smaller properties. Buildings that were constructed from1960 until 1970 are sold with a discount that is 9.3% lower than properties that are built in the period 1945-1959, which is the reference category. The coefficient is significant at the 95% confidence interval. Also the variable ‘no ground rent’ is significantly influencing the discount of auctioned properties (0.031).

The regression results that are presented in column 7 take into account random effects. The results do not differ from the results of column 6. The coefficient real estate agent is still negatively influencing the discount of auctioned properties (-0.013). The reasons ‘forced sale’ and ‘sale under court order’ have the same coefficients and are still significantly influencing the discount of auctioned properties.

The broker fee that the real estate receives is on average 1.79%. The extra costs for hiring a real estate agent after reducing the broker fee from the average price is around 1%. Thus, it is clear that the value of expert knowledge of real estate agents does not result in a higher discount in monetary terms. The difference in the discount can be explained. If the real estate agent and the client agree on a maximum price for a property, the real estate agent could say ‘mine’ earlier in the second round if the price is below the maximum amount that the client wants to spend. This guarantees the client that he or she will be the new owner of the property and the real estate agent receives his broker fee. A higher buying price would result in a higher fee for the real estate agents and could apply as an incentive for them.

(32)

32 Nevertheless, it could be useful to hire a real estate agent, because the real estate agent takes care of the buying process. As a result, the real estate agent clients can be certain about how to buy property and how the bidding process works. The real estate agent can also warn clients about risky properties that are being sold at auction. Besides that, a real estate agent has up-to-date information on financial terms and trends in the market, which can save the client much time. The real estate agent can also advise his clients on the ground lease conditions, the auction costs, the due payments and the notary fees. Therefore, the knowledge of the real estate agent can still be useful, while the costs for hiring one are minimal, around 1%, and while it eliminates much uncertainty for clients.

The most interesting reasons for selling are ‘forced sale’ and ‘sale under court order’. The properties that are forced sold have a negative significant effect on the discount, at 95% confidence level. This means that properties that are forced sold have a discount that is on average 7.9% lower than properties that are sold as ‘voluntary sales’. Also, the reason sale under court order has a negative significant effect on the discount of auctioned properties. If a property is sold under court order the discount is on average 16.3% lower than for properties that are sold as ‘voluntarily sales’. This is in line with the expectations, because the overdue charges are not always mentioned clearly in the auction books. This could result in people underestimating the overdue charges, which in turn results in a discount that is lower on average. The percentages are different for the reasons for sale, because the number of overdue charges differ among the reasons for sale. The reason ‘sale due to bereavement’ is not statistically significant in column 6 and 7. This means that the discount for sales due to bereavement do not statistically differ from voluntarily sales. The number of transactions is lower compared with properties that are voluntarily, forced or sold under court order. More transactions could increase the significance level.

Prices at real estate auctions : the role of market knowledge and the reason for sale