market.
By
Patrick Merkx (6150667)
MSc. Thesis
MSc. Business Economics
Specialization: Finance
University of Amsterdam (UvA)
Supervisor: prof. dr. Marc Francke
Abstract
Several international studies have pointed out at the worsened position of First Time Buyers (henceforward FTB) in housing markets, while Van de Minne and Francke (2013) show an improved position of FTB in the Dutch housing market since the crisis. This thesis examines the position of FTB and Non First Time Buyers (henceforward NFTB) in the Dutch housing market. Firstly, the house price trend of both groups was examined. The average house price of NFTB increased more than the average house price of FTB during 1999-2008. This trend reversed after the crisis (2008-2011). Secondly, we examined factors that influence the transaction price. There are almost no differences in the constant quality indices. The difference in quality of houses bought was the largest explanatory factor for the different house prices. The difference in quality increased before the start of the crisis, indicating that the position of FTB worsened in the pre-crisis period. Subsequently, the impact of difference in quality decreased after the start of the crisis, indicating an improved position of FTB in the Dutch housing market. Finally, we showed that more FTB bought houses in better location since the crisis, while FTB did not buy better types of houses or houses with larger physical size characteristics.
Table of contents
Page
1. Introduction 4
2. Literature review 8
2.1 Position of FTB in European housing markets………. 8
2.2 Position of FTB in the Dutch housing market ………. 9
2.3 Review of hedonic price models……….……….. 10
2.4 Conclusion of literature review……….………... 12
3. Methodology 13 3.1 Factors affecting the transaction price………. 13
3.2 Model for estimating beta and lambda….……… 14
4. Data 17
4.1 Data design and measurement………. 17
4.2 Descriptive statistics……… 19
5. Results 22 5.1 Differences in constant quality indices……… 22
5.2 Marginal contribution of housing characteristics on transaction price……… 23
5.3 Impact effects on transaction prices……… 26
5.4 The relative change of housing characteristics……….... 28
6. Conclusion 34
7. References 36
Appendix A 38
Appendix B 46
1. Introduction
Due to the economic growth and historically low interest rates during the last two decades, financial institutions were able to advance higher levels of credit to consumers (Van de Minne and Francke, 2013). This resulted in an outstanding mortgage debt to GDP ratio that has increased from 61 percent in 1998 to 99 percent in 2008, one of the highest in Europe (European Mortgage Federation, 2008). As a result of this increased mortgage extension, combined with an increased demand and low interest rates, house prices more than tripled in the Netherlands between 1985-2008 (Neuteboom and Brounen, 2011).
Nowadays the Dutch housing market is entirely different. The credit crunch hit the Netherlands in 2008 (Boelhouwer and Priemus, 2014). House prices started to decline, and the number of houses for sale rose from 158.000 in 2008 to 207.000 in 2013. The Dutch government reacted to the financial crisis by lowering the transfer tax from 6 to 2 percent, and the maximum loan that was ensured by the Dutch Mortgage Guarantee (NHG)1 was raised from €265,000 to €350,000 in July 2009. Furthermore, to reduce the total mortgage debt of the Netherlands, only annuity mortgages keep their interest expense deduction and the loan to value2 is lowered by 1 percent every year (Schilder and Conijn, 2013).
The Dutch fiscal system stimulated large loan to value mortgage by their interest expense deductibility. Eventually, the loan to value will decrease since they profit from an increase in house prices while their debt amount remain fairly constant. However, in times when house prices remain the same or even decrease, as in the Netherlands after 2008, homeowners are confronted with negative home equity (van Hoek and Koning, 2012). When homeowners face negative home equity they are ’under water3’. In the worst scenario, homeowners will default on their mortgage payments. Many people in the Netherlands who bought a house between 2004 and 2008 are facing this negative home equity, causing major problems in the Dutch housing market (Schilder and Conijn, 2012).
1
The National Mortgage Guarantee assures Dutch households of a responsible and affordable mortgage. When changes in personal circumstances occur, for example a divorce or unemployment, households are offered a safety net. Households may therefore keep their homes and prevent a loss.
2 The loan to value is used to express the ratio of a loan to the value of an purchased house. The loan to
value was 106% in 2013 and will decrease to 100% in 2018. The original LTV gave FTB the opportunity to renovate the house or to buy some furniture.
3 The value of the mortgage is higher than the value of the house. This was the case for 1.4 million Dutch
Contrary to homeowners, there is a different group; young households who want to start in the Dutch owner occupied housing market, also referred as potential First Time Buyers (henceforward FTB) who by definition cannot have negative home equity. This group did not have a chance to build up wealth, making them completely reliant on mortgage debt when purchasing a home (van Hoek and Koning, 2012). Meen (2013) state that the inflow of FTB plays an important role for the overall health of a housing market and the actions of FTB may be crucial in housing market depressions. According to Neuteboom and Brounen (2011), FTB create essential liquidity at the bottom of the housing market, which eventually flows through the entire housing market, thus helping it out of a recession. Therefore, it is of great importance to understand the characteristics and the position of FTB in the Dutch housing market.
Before the financial crisis, FTB were facing high competition of Non First Time Buyers (henceforward NFTB) (Neuteboom and Brounen, 2006). Nowadays, many NFTB are not able to move to another house because they cannot finance their negative home equity with them, resulting in less competition. Van de Minne and Francke (2013) suppose that the position of Dutch FTB has improved. They state that FTB can buy houses they could not afford before the crisis, due to the decrease in demand of NFTB, low interest rates and declined house prices. See Figure 1 for some stylized facts.
Figure 1: Average transaction price index (Van de Minne and Francke, 2013) 100 120 140 160 180 200 220 240 260 280 300 1995Q 1 1995Q 4 1996Q 3 1997Q 2 1998Q 1 1998Q 4 1999Q 3 2000Q 2 2001Q 1 2001Q 4 2002Q 3 2003Q 2 2004Q 1 2004Q 4 2005Q 3 2006Q 2 2007Q 1 2007Q 4 2008Q 3 2009Q 2 2010Q 1 2010Q 4 2011Q 3 2012Q 2 Ind ex FTB All
Figure 1 shows that the average transaction price of Dutch households dropped significantly since the beginning of the crisis, while the average house price of FTB remained approximately constant or even increased a little bit.
Unlike Van de Minne and Francke (2013), various international studies have suggested that the position of FTB in the owner-occupied housing market has worsened, due to poorly performing economies and employment insecurity since the financial crisis (Andrew, 2012; Fisher and Gervais, 2011; Forrest and Yip, 2012). Mckee (2012) has labeled FTB as the ‘generation of rent’. In the authors’ opinion, young people in the UK have difficulties entering the owner occupied housing market and even when they leave their parental home, many face significant challenges accessing enough mortgage finance. Why is the international opinion about FTB so different from the opinion of Van de Minne and Francke (2013)? The key research question in this thesis is: Has the
position of FTB improved in the Dutch housing market since the beginning of the financial crisis in 2008? To answer this question, this thesis empirically examines if FTB
bought houses of better4 quality after the financial crisis. The position of FTB is compared to NFTB before and after the financial crisis in the Netherlands. One hypothesis is formulated to answer the research question.
H1: Dutch First Time Buyers have bought houses of better quality since the beginning of
the financial crisis in 2008.
Many households who have bought a house before the financial crisis are ‘under water’. Due to the drop in demand of NFTB, low interest rates and declined house prices it is expected that FTB bought houses of better quality after the financial crisis.
Just as in the paper of Van de Minne and Francke (2013), we examine the different house price trends of NFTB and FTB. We use hedonic price models to derive constant quality indices in order to explain these different trends. The advantage of this method, is that it indicates how much of a total price increase is due to house inflation and how much is due to an increase in quality (Bourassa et all, 2008). In addition to the hedonic price models, simple analyses show how the individual housing characteristics have developed during the sample period, indicating more precisely which poorer and better characteristics NFTB and FTB bought. This sub-Section will give answers to questions like; did FTB
4Van de Minne and Francke (2013) referred to better economic quality. So improvements like larger
buy houses with larger physical size characteristics or did they buy houses in better locations?
Using WoON5 data, the results of the hedonic price models show that there are almost no differences in the constant quality indices of NFTB and FTB. Changes in the quality of houses bought are the largest explanatory factor for the differences in house prices. The differences in quality increases before the crisis, indicating that FTB bought houses of poorer quality compared to NFTB, and were forced to buy houses at the bottom of the market. The impact of difference in quality decreased after the start of the crisis, indicating an improved position in the Dutch housing market. More FTB bought houses in better locations, while FTB did not buy better types of houses or houses with larger physical size characteristics since the crisis.
In total, this thesis contribution is threefold. It provides additional empirical proof to existing literature on the position of NFTB and FTB in the Dutch housing market. It also presents two constant quality indices , which can be used for further research into the Dutch housing market. Finally, it provides some interesting insights how NFTB and FTB value individual housing characteristics. The difference in marginal contribution for each characteristic on the house price is useful information for Dutch real estate agents and investors.
This thesis is structured as follows. The next Section provides a literature review about the position of FTB in Europe and the Netherlands. Also some background information about hedonic analysis is discussed here. The third Section contains the methodology used for this thesis. The data are described in the fourth Section. Section five presents the results, focusing on the differences between NFTB and FTB. Finally, this thesis concludes in Section six.
5 The WoON data sets are from the Dutch housing demand survey. The survey contains information
regarding household characteristics, the current housing situation of households and the housing desire for short-term and migration behavior. The ministry of Housing, communities and integration and Statistics Netherlands jointly carry out the WoON survey every three to four years.
2. Literature review.
The aim of this Section is to provide a clear understanding of the position of FTB in housing markets. First, the position of FTB in Europe is discussed before and after the crisis. Subsequently, the position in the Dutch housing market is discussed and some background information about hedonic analysis is reviewed in order to understand the method used for answering the hypothesis.
2.1 Position of FTB in European housing markets
One of the most significant social changes of the 20th century was the growth of the homeownership in advanced countries (Ronald, 2008). Promoting homeownership has played a huge role in the housing policy of governments across these countries. The expansion in homeownership rate was achieved by easy credit supply and low-cost homeownership products were introduced to attract the low-end of the population (Ebner, 2013). Yet, despite the increase in homeownership and their ability to secure their future wealth, FTB have difficulties to enter the owner-occupied housing market, which leaves them to move to rental accommodations or stay longer in parental home (Forrest and Yip, 2012). Mckee (2012) state that even when FTB want to leave their parental home, accessing homeownership is a challenge. Especially due to the more stringent lending criteria of financial institutions that followed from the financial crisis. The author concludes that people aged between 18 and 30 are in danger of becoming a lost generation and emphasizes that these hazards have international resonance, despite that the empirical evidence is largely based on the economy of the UK.
Studies about Southern Europe countries showed that FTB had limited access to mortgages before the crisis, which made it hard for FTB to buy their first house without substantial family assistance (Castles and Ferrara, 1996). On the other hand, countries like Denmark and the Netherlands are marked by relatively high homeownership among FTB, because they can (almost) fully borrow the value of their future property (Mulder and Billari, 2010). Altogether, there were some deviations in the homeownership rates of FTB within Europe.
Lennartz, Anrundel and Donald (2014) compared the absolute percentage homeowners, aged between 18 and 34, in fifteen advanced European economies before
(2007) and after the crisis (2011). Their results show that there was no significant growth in FTB homeownership rate in any country, with more or less small declines being the norm instead. Belgium, France, Finland and Germany had relatively stable homeownership rates among FTB in this post-crisis period. All Southern European nations, Luxemburg and Austria showed a substantial decline in homeownership rates. The highest absolute decrease of 10.2 percent was in the UK. A possible explanation for this decline is that the UK government increased the required down payment, which disproportionally affects younger UK households who have less time to save (Stephens, 2011). In the Netherlands, the percentage of homeowners, aged between 18 and 34, declined from 40.05 percent in 2007 to 34.67 percent in 2011; an absolute decline of 5.38 percent. Opposite to these findings, Van de Minne & Francke (2013) found that FTB did not buy less houses since 2008 while the total overall number of house transaction declined, which resulted in a relatively larger share for FTB in the Dutch housing market (see Figure 2 in sub-Section 4.1).
2.2 Position of FTB in the Dutch housing market
Neuteboom and Brounen (2006) analyzed FTB in the Dutch housing market between 1985 and 2002. Their research was done with the same house survey that is used in this thesis. They state that the borrowing capacity of Dutch households rose and the mortgage interest rate declined between 1985 and 2002. However, these factors are outweighed by the triple of house prices and therefor the affordability of FTB declined during 1985-2002. The ratio of net cost of housing to net income of the household increased from 14 percent in 1985 to over 23 percent in 2002. The accessibility of FTB has also weakened between 1985 and 2002. According to Neuteboom and Brounen (2006) the main reason for this, is the increased competition. For each affordable house for sale in 1985, a FTB had to compete with 177 other potential buyers while this number more than doubled by 2002. The overall conclusion of their research is that the position of FTB in the Dutch housing market worsened during 1985-2002. FTB had to switch to the cheapest types of houses since the price level of the remaining types became too high. Another research of Neuteboom and Brounen (2011) about households who had entered the market in 2005 or 2006 showed that the position of FTB was indeed relatively weak just before the crisis. The results highlight the fact that FTB needed to compete with richer, older buyers who had already built up some housing wealth during their housing career.
Van de Minne and Francke (2013) agree with this weakened position of FTB before the crisis. They compared the position of FTB with the rest of the market before and after the crisis. Therefore they studied the price trend of FTB houses in comparison to the overall Dutch housing market. During the period of 1995-2001 the prices of FTB houses rose by 70%, while the average Dutch house price rose by 120%. Prices increased approximately the same (30%) between 2002 and 2008. However, prices of FTB houses rose 5%, whereas the prices of the overall Dutch housing markets declined 15% during 2009-2012. Van de Minne and Francke (2013) explained this difference, on a macro-level basis, by assuming that FTB bought houses with different quality characteristics over the years. During 1995-2008 house prices increased faster than the borrowing capacity of FTB, so they were forced to buy houses of poorer quality, meaning a decline in the affordability rate. Since 2008, FTB are buying houses of better quality, due to the worsened position of NFTB in combination with declined house prices and interest rates.
Schilder and Conijn (2013) share the opinion of Van de Minne and Francke (2013). They state that despite restrictions on borrowing, the position of FTB on the Dutch housing market has improved as a result of falling house prices since the beginning of the crisis. After all, house prices fall stronger than the borrowing capacity of FTB, causing a higher affordability rate for them. For NFTB, however, this fall in house prices gives a less rosy perspective. The height of their mortgage exceeds the value of their house, which led to less movement in the Dutch housing market and an even better position for FTB (Schilder and Conijn, 2013).
2.3 Review of hedonic price models
Statistical models assume that there is a relationship between prices and house characteristics, so that different prices can be explained by differences in house characteristics when valuing real estate properties. The economic foundation for these models lies in the hedonic price theory. Hedonic regressions are widely used for academic real estate papers for almost 40 years (Dunse and Jones, 1998). Rosen (1974) formalized the interpretations of the hedonic price model. Each characteristic contributes to the value of a good. This value cannot be traded individually. Therefore, the price paid for a particular house is the sum of the prices that the market gives to all the characteristics associated with the house. Regressing the transaction price on their various characteristics yield the marginal contribution of each housing characteristic. More generally, the
coefficients of the estimates represent the prices of these characteristics (Bourassa, Hoesli and Sun, 2006). By comparing the coefficients of NFTB and FTB, the difference in marginal contribution of each housing characteristics on the housing price is examined. The differences in marginal contribution create two different values for an identical house with the same characteristics. Although it seems illogical when a house with identical characteristics has two different values, there are papers that examined the effects of different subject characteristics on house prices.
Genesove and Mayer (2001) studied the effect of such subject characteristics on house prices. They examined a boost cycle in the Boston housing market during 1990-1997. They found that sellers who are subject to losses set higher asking prices of 25-35 percent of the difference between the expected selling price and the original purchase price and attain higher selling prices of 3-18 percent of that difference. In summary, Genesove and Mayer (2001) found that loss aversion of a seller helps explaining house prices in the real estate market. When house prices decrease after a boom, many households have a market value below the price they had paid for them. As a result, owners who are averse to losses will have an incentive to set a higher selling price that exceeds the level they would set when they did not face that loss. Based on these findings, Genesove and Mayer (2001) conclude that the real estate market is not a perfect asset market.
Besides the marginal contribution of housing characteristics on the house price, hedonic price models are widely used in the academic literature to derive constant quality indices (Goodman, 1978; Mark and Goldberg, 1984). The transaction prices are corrected for differences in housing characteristics, indicating how much of a total price increase is due to house inflation and how much is due to an increase in quality (Bourassa et all, 2008). There are two basic ways to construct this index using hedonic regressions. In the first approach, one overall hedonic regression is performed and time dummies are included. In the second approach, a separate regression is done for each period. The first approach is used for this thesis because it was more consistent with the design of the data. This method is, for example, used for a single-family house index in the UK or to construct various types of residential house indices in Switzerland (Bourassa et al. 2006).
2.4 Conclusion of literature review
Although extensive research has been carried out on the position of FTB, there have been no empirical papers that compared the changed position of NFTB and FTB in the Netherlands since the crisis. The international literature is more pessimistic about the position of FTB, while Van de Minne and Francke (2013) provided one the first papers highlighting the improved position of FTB in the Dutch housing market. This thesis examines the assumptions of the paper of Van de Minne and Francke (2013) and thereby provides an empirically contribution to the Dutch housing literature.
3. Methodology:
The paper of Van de Minne and Francke (2013) showed that the price development of FTB houses followed a different course compared to the average house prices of all Dutch households. In this thesis we look explicitly at the differences in house price appreciation between NFTB and FTB. Firstly, we present a yearly price index for both groups, based on average transaction prices, for our sample period (1995-2011) using micro data. Next we construct a constant quality index for NFTB and FTB using a hedonic price model. Comparing these sets of indices and examining the marginal contribution of the housing characteristics on house prices helped answering the hypothesis: Dutch First Time Buyers
have bought houses of better quality since the beginning of the financial crisis in 2008.
3.1 Factors affecting the transaction price.
How much of the difference in price is caused by difference in quality? Potential factors that influence the transaction price are examined to answer this question. These factors are shown below.
!.!!"# = !.!!"# !!"# + !!"#! + !.! !"# (1) !.!!"#$ = !.!!"#$ !!"#$+ !!"#$! + !.! !"#$ (2)
Where ! is the average transaction price. The natural logarithm of the transaction price is taken6 instead of the transaction price itself. Subscript t stands for time. The average housing characteristics like garden area, numbers of rooms, but also geographic location are presented in row vector ! . ! is the vector of parameter of corresponding characteristics. This parameter shows how much value is given to this characteristic. The log price level per year is denoted by λ. Equation (1) and (2) show that there are several possible causes for the different transaction prices. First of all, it could be a composition effect (! !), this effect represent the difference in quality in this thesis. Secondly, the time effect could cause different transaction prices (λ). This effect reflects the differences in house inflation. The intuition behind the time effect is that the market segment where FTB buy their houses has a different price development compared to the segment of
6
NFTB, due to different demand and supply dynamics. Finally, (ϵ) indicates the average unobserved factors.
To examine the precise impact of the said effects on the transaction prices, equation (2) is subtracted from (1) into equation (3). The results of equation (3) show what part of the difference in the transaction prices between NFTB and FTB is caused by the composition (differences in quality) and time effect (difference in house inflation), for each year in the sample period.
!.!!"#$− !.!!"# = !.!!"#$− !.!!"# !!"#+ (!!!"#$− !!!"#) + (3) !!"#$ − !!"# !
.!!"#$ + (!.! !"#$− !.! !"#)
(!.!!"#$ − !.!!"#) is the difference in the natural logarithm of the average transaction price for NFTB and FTB. The difference in house inflation is presented by (!!!"#$− !
!
!"#). To examine the effect of differences in quality, the average housing characteristics of NFTB and FTB are multiplied with the regression coefficients of FTB !.!!"#$− !.!!"# !!"#. However, there is a possible drawback by calculating the composition effect with the beta of FTB and not with the beta of NFTB. When the beta’s of both groups are significantly different, a rest term needs to be taken into account !!"#$ − !!"# !
.!!"#$. The average difference of the residuals is presented by (!.! !"#$− !.! !"#) and is always zero. However before we can continue our analysis we need to find estimates for beta (!) and lambda (!).
3.2 Model for estimating beta and lambda.
We estimate a hedonic price model with ordinary least squares. Beta and lambda are chosen such that the difference between the sum of squared differences between the resulting model values and transaction prices is as small as possible. The error is expected to be normally distributed with mean 0 and variance !!. (Stock and Watson, 2003).
A pragmatic approach is used to identify which sets of characteristics produces the best fit, shown by the adjusted R squared. The presence of multi-colinearity in the hedonic price model is examined by eyeballing the magnitude of the standard errors of the regression coefficients to see if the values of these coefficients were plausible. NFTB and
FTB are modeled together into equation (4) and therefore we allow the coefficients and log price level to vary between both groups. The model now becomes;
!!" = !!"! + !!"!"#!
!"!"#!!"# + λ!+ !!"!"#!!!"#+ !!" (4)
!!" shows the log transaction price of house i at time t. The housing characteristics are presented in row vector !. ! is the vector of parameter estimates of corresponding characteristics. ! represent the house inflation and !!" is the error term.
Year dummies were entered to estimate the time effect (λ!). This effect is presented as constant quality index. All year dummies are multiplied with the First Time Buyer dummy (!!"!"#), creating two indices for NFTB and FTB. These constant quality indices show how much the transaction price is today for the same quality house as in the base year. Movements in these indices are caused by house inflation and not by differences in quality of houses bought. Furthermore, all housing variables are multiplied with the First Time Buyer dummy (!!"!"#), creating new interaction variables. These interaction variables show if FTB valued housing characteristics differently than NFTB. More specifically, λ! and ! captures the effect of the house inflation and housing characteristics for NFTB, and (λ!+ !!!"#) and (! + !!"#) captures these effect for FTB. The difference in house inflation between NFTB and FTB is therefore indicated by !!!"# and the differences in housing characteristics by !!"#. T-statistics are calculated, using the coefficients and standard errors of these variables, whether the differences between NFTB and FTB are statistically significant at a 5% level.
When the variable !!"# is significant it shows that the characteristic has two different marginal contributions to the house price. For example: !!"# of the garage variable is 0.20 and significant. This means that the transaction price of a FTB house with a garage increases 20% more than an identical NFTB house with a garage. In other words, the marginal contribution of a garage is much higher for FTB than for NFTB. Therefore, real estate agents should always focus on FTB when they sell a house with a garage in this example. From economic theory, it seems illogical when a garage has two different values. Therefore, we regressed an extra hedonic price model for robustness (7), were we omitted the interaction terms of the housing characteristic. This hedonic price model examines how the composition effect and constant quality indices behave, when there are no differences in marginal contribution of housing characteristics on the transaction price.
The exact specification and results are presented in appendix A. Furthermore, for extra robustness and justification of model (4), we regressed two separate models for NFTB and FTB. The results of these hedonic price models are presented in Appendix B.
Next to the hedonic price models, simple analyses are used to compare the means of the housing characteristics. The results of this analysis show how the difference in quality of houses bought between NFTB and FTB is expressed into individual housing characteristics, indicating more precisely which poorer and better characteristics NFTB and FTB bought.
4. Data
Data from the Dutch housing demand survey (WoON 2006, 2009 and 2012) have been used to answer the hypothesis. The survey contains information regarding housing characteristics, the current housing situation of households and the housing desire for short-term and migration behavior. The Communities and Integration, Ministry of Housing and Statistics Netherlands jointly carry out the WoON survey every three to four years.
4.1 Data design and measurement
After merging the WoON datasets of 2006, 2009 and 2012 the sample consisted of 215.011 respondents. Since this thesis focuses on homeowners, all households that were marked as tenants were excluded. Secondly, all respondents who bought their house before 1995 were deleted. People who bought their house in 20127 were also deleted. Furthermore, filters were applied to avoid outliers and inaccurate results. All criteria that were used for the data selection procedure are listed in Table 13 in Appendix C.
A general problem with the data of The WoON survey is that we cannot determine the previous type of accommodation of the respondent, implying we did not know whether a homeowner is actually a FTB or not. In previous studies the selection criterion for FTB was based on age. Just as in the paper of Van de Minne and Francke (2013), people aged between 18 and 35 are defined as FTB in this thesis. The following formula was used.
!"# = !"# !" !"#$%&'"&( − !"#$%& !"#$ − !"#$ !" ℎ!"# !"#$ℎ!"# ≤ 35
To illustrate: if the respondent was 45 in the sample of 2012 and bought the house in 2000, the respondent was 33 when the transaction took place, and therefore, is defined as FTB. A total of 28.157 respondents were labeled as FTB and 24.518 respondents as NFTB. Of course there are respondents who are under 35 and bought a house for the second time, and some respondents who bought their first house when they were over 35.
7 There were only 30 respondents who bought a house in 2012. This rate is too low to draw reliable
An assumption is made that these errors cancel each other out. Although, this disadvantage has to be taken into account for the reliability of the results.
Figure 2: Number of FTB transactions (Van de Minne and Francke, 2013)
Figure 2 shows the number of FTB transaction on the right axis and the relative percentage on the left axis. FTB did not buy fewer houses since the start of the crisis. However, because the total number of transactions has fallen dramatically since 2008, the relative share of FTB transaction increased. The possible explanation why FTB did not buy fewer houses after the crisis is that they do not face negative home equity like many NFTB (Van de Minne and Francke, 2013).
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 0,0% 10,0% 20,0% 30,0% 40,0% 50,0% 60,0% 70,0% 80,0% 1995Q 1 1995Q 4 1996Q 3 1997Q 2 1998Q 1 1998Q 4 1999Q 3 2000Q 2 2001Q 1 2001Q 4 2002Q 3 2003Q 2 2004Q 1 2004Q 4 2005Q 3 2006Q 2 2007Q 1 2007Q 4 2008Q 3 2009Q 2 2010Q 1 2010Q 4 2011Q 3 2012Q 2
4.2 Descriptive statistics
Table 1: Examined variables in hedonic price models
Variable name Form Categorized
Transaction price Log
House size !! Log
Bathroom Dummy
Bath (with shower), shower, separate bath/shower and other types of bath
House type Dummy
Detached house, semi-detached house, corner house, attached house, apartment and other private property
Building period Dummy
<1945, 1945 -1959, 1960-1969, 1970-1979, 1980-1989, 1990-1999 and >2000
Garden size !! Dummy 0, 1-49, 50-99, 100-249 and >249 !!
Balcony/roof terrace
size !! Dummy 0, 1-4, 5-9 and >9 !!
Garage/ carport Dummy Garage (and carport), only carport or no garage/carport
Location Dummy Forty COROP-regions
The variables that are used for the hedonic price models are listed in Table 1. Transaction price and house size are log transformed. The variable bathroom is divided into four dummies: houses with a bath (and shower), houses with a shower, houses with separate bath and shower, and houses with an other type of bath. The type of house is divided into six dummies: detached house, semi-detached house, corner house, attached house, apartment and other private property. The building period is also divided into dummies, just as the garden size and balcony/roof terrace8. Furthermore, the garage is divided into three dummies: if the respondent has a garage (and carport), only a carport, or none of those. Finally, the location of the house. The location is of great importance for the transaction price and is based on COROP-regions 9. A list of the COROP-regions can be found in Table 14 in Appendix C. Table 2 and 3 present descriptive statistics. However, due to the size, COROP-regions are left out.
8
We have classified respondents with no answer for garden size or balcony/roof terrace size into the 0 square meter class.
9
The zip code was not available in each WoON dataset. Therefore we examined the location of a house by means of COROP-regions. A COROP-region is a regional area within the Netherlands. COROP means ‘’Coordinatie Commissie Regionaal Onderzoeks Programma’’. These regions are more accurate than on a provincial level and are often used for analytical purposes. Each region consists of a number and has a central city. The Netherlands is divided into forty COROP-regions.
Table 2: Frequency table (House types, Garage, Bathroom and Building period) N FTB % N NFTB % House type Detached 2746 9.8% 4817 18.6% Semi-detached 4099 14.6% 3905 15.9% Corner house 4078 14.5% 2994 12.2% Attached 10478 37.2% 6093 24.9% Other property 298 1.1% 494 2.0% Apartment 6458 22.9% 6215 25.3% Total 28157 100% 24518 100% Garage
Garage (with carp)
8136 28.9% 11050 45.1% Carport 1134 4.0% 1293 5.3% No garage/carp 18887 67.1% 12175 49.7% Total 28157 100.0% 24518 100.0% Bathroom
Bath (with shower)
6387 22.7% 4478 18.3% Shower 11059 39.3% 9403 38.4% Separate bath/shower 10641 37.8% 10553 43.0%
Other type of bath 70 0.2%
65 0.3% Total 28157 100% 24518 100% Building period >1945 6133 21.8% 6289 25.7% 1945-1959 1919 6.8% 3420 13.9% 1960-1969 3127 11.1% 2480 10.1% 1970-1979 3714 13.2% 2078 8.5% 1980-1989 3521 12.5% 2455 10.0% 1990-1999 4572 16.2% 2987 12.2% >2000 5180 18.4% 3809 19.6% Total 28157 100,0% 24518 100,0%
Table 2 shows that a larger percentage of NFTB (18.6%) lived in detached houses compared with FTB (9.8%). 45.1 percent of NFTB had a garage, against 28.9 percent for FTB. The bathroom and building period variable do not show relevant differences. Houses that were build before 1945, and after 2000, were both the most popular for both groups.
Table 3: Descriptive Statistics (Transaction price and physical size characteristics)
FTB NFTB
Mean S. D. Min Max Mean S. D. Min Max Transaction price (000) 190.2 94.6 51.8 125 235 135.8 50.4 2000
House size m2 125.5 45.6 35 500 136.2 51.7 35 500
Number of rooms 4.4 1.3 1 15 4.66 1.43 1 15
Garden size m2 180.8 651.7 0 14000 312.6 953.9 0 14200
Balcony/Roof m2 3.6 7.3 0 94 5.4 9.7 0 96
Table 3 shows that the average house price was €45,000 higher for NFTB during the sample period. The house size was on average 11!! more for NFTB. NFTB had on average 0.16 more rooms. Furthermore, there was a big difference in garden size. NFTB had on average 140!! larger garden surfaces. Finally, NFTB had on average 1.8!! more surfaces on their balconies or roof terrace.
Altogether, Table 2 and 3 indicate that on average, NFTB bought houses with better housing characteristics during the sample period (1995-2011). However, the next Section focuses on the change of these housing characteristics during the crisis, thereby examining if FTB bought better quality houses.
5. Results
As explained in Section 3, we firstly present a yearly price index for both groups, based on average transaction prices, for our sample period. Hereby it comes clear which index outperforms the other one.
Figure 3: Transaction price indices based on average prices for NFTB and FTB
Figure 3 shows the development of the transaction price based on average prices for NFTB and FTB during 1995-2011. Generally speaking, this period can be divided into three phases. House prices increased approximately 22% for both groups during 1995-1999. Between 1999 and 2008, the price index for NFTB increased with 59%, while the price index for FTB rose with 38%. This trend changed when the financial crisis started. Both indices started moving together after 2008. The average house prices of NFTB decreased with 8% between 2008 and 2011, while the average house price of FTB decreased with 3%.
5.1 Differences in constant quality indices
In order to explain these different house price trends we inserted year dummies in the hedonic price model (4) to create two constant quality indices for NFTB and FTB. These indices are presented in Figure 4.
100 120 140 160 180 200 220 Ind ex NFTB FTB
Figure 4: Constant quality indices of NFTB and FTB.
Figure 4 shows that the index of FTB was slightly higher during the sample period (1995-2011). However, almost all the differences in the log price level are not significant, which can be seen from the year dummies in Table 8 in Appendix A. The insignificant differences in log price level indicate that the two indices follow (almost) the same path. Therefore, the segment where NFTB and FTB buy their houses does not has different demand and supply dynamics. Focusing on the hypothesis: differences in constant quality indices (almost) did not have an effect on the different house prices.
5.2 Marginal contribution of housing characteristics on transaction price
After the constant quality indices were compared, the differences in marginal contribution of the housing characteristics were examined. The marginal contributions of the housing characteristics are used to calculate the composition and rest effect.
Colum 2 in Table 4 shows the regression coefficients of all respondents. The difference between FTB and all respondents are presented in column 4. More specifically, column 2 shows how NFTB valued each housing characteristic and column 4 shows the difference in marginal contribution of the housing characteristics on transaction price between NFTB and FTB. FTB valued housing characteristic different when a variable in column 4 is significant at a 5% level.
100 120 140 160 180 200 220 240 In d ex NFTB FTB
Table 4: Regression coefficients of housing characteristics
LN Transaction Price Coefficient ALL Prob Coefficient FTB Prob
Constant First Time Buyer
9.520 .281 .000 .000 COROP-regions10 Noord-Friesland -.513 .000 -.088 .002 Z-O Friesland -.445 .000 -.091 .008 Z-O Drenthe -.544 .000 -.075 .023 Z-W Overijssel -.280 .000 -.070 .045 Twente -.408 .000 -.055 .008 Groot Amsterdam -.015 .253 .068 .000 Zeeuwsch Vlaanderen -.546 .000 -.081 .001 Zeeland overig -.374 .000 -.059 .002 Garage/Carport Garage .148 .000 -.034 .000 Carport .086 .000 -.021 .098 House type Detached .274 .000 .088 .000 Semi-detached .142 .000 .074 .000 Corner house .066 .000 .082 .000 Attached .045 .000 .073 .000
Other house types .208 .000 .084 .001
Bathroom
Bath (shower) .030 .227 -.067 .063
Shower -.023 .353 -.070 .050
Separate bath and shower .109 .000 -.094 .009
Building Period Before 1945 -.078 .000 -.019 .030 1945-1959 -.173 .000 .026 .036 1960-1969 -.186 .000 .042 .000 1970-1979 -.158 .000 .036 .000 1980-1989 -.119 .000 .031 .002 1990-1999 -.032 .000 .012 .169 Physical size Garden 1-49 !! .018 .060 .054 .000 Garden 50-99 !! .016 .092 .065 .000 Garden 100-249 !! .054 .000 .060 .000 Garden > 250 !! .161 .000 .038 .010 Balcony/roof ter. 1-4 !! .067 .000 -.015 .130 Balcony/roof ter. 5-9 !! .075 .000 .000 .961 Balcony/roof ter. > 9 !! LN house size !! .127 .457 .000 .000 -.027 -.083 .001 .000 Method Observations OLS 52,675 S.E of regression 0.297 Adj. R2 0.636
10Only significant COROP-regions are presented in Table 4. Full regression results are presented in Table
The coefficients in Table 4 show that FTB who are living in Noord-Friesland, Zuid-Oost Friesland, Zuid-Oost Drenthe, Zuid-West Overijsel, Twente, Zeeuwsch Vlaanderen and Zeeland paid between 5.5%-9.1% less for a house than NFTB. Contrary, the transaction price of FTB increased 6.8% more to live in Groot-Amsterdam. The rest of the interactions COROP-regions were not statistically significant, indicating that FTB and NFTB value these regions the same.
FTB paid 3.4% less for a garage (with a carport). More surprisingly are the regression coefficients for types of houses. FTB paid 8.8% more for a detached house compared to NFTB. FTB paid also more for a semi-detached house (7.4%), corner house (8.2%) and attached house (7.3%). A house with a separate bath and shower is valued 9.4% less by FTB. This percentage is 7.0 for a house with a shower. Furthermore, FTB paid around 3.5% more when a house was built between 1945-1989. FTB paid 1.9% less for a house built before 1945. The impact of the building period is quite complex, because newer houses typically command a price premium. However, older houses may also command a premium. Therefore these results do not need to be interpreted as important because it is difficult to classify building periods in the poorer or better quality scale.
Physical size variables are the last characteristics being discussed. FTB, valued a house with a garden between 3.8%-6.5% more than NFTB. Furthermore, only the >9 !! balcony/roof terrace dummy was significant, meaning that FTB and NFTB value a balcony or roof terrace the same, unless the surface is >9!!.
From economic theory it seems illogical that there are differences in marginal contribution of the housing characteristics on the house price. However, the FTB dummy absorbs these differences. FTB paid 28.1% more for a house in the basis year (1995) compared to NFTB, controlling for all other effects. This can be mainly attributed to the different marginal contribution of the house size variable. To illustrate: the average house size is approximately 130!! (Table 3). The value of a FTB house should have the same value as an identical NFTB house with a house size of 130!!. Therefore, the coefficient of the FTB dummy + (0.457-0.083)*LN (130) = 0.457*LN (130). In other words 0.083*LN (130) = 28.1%. However, this percentage is 40.4%, indicating that an FTB house with 130!! has a 12.3% lower value compared to an identical NFTB house with a house size of 130!!. This percentage corresponds much more with the FTB dummy (-10.5%) from the robustness check in Table 9 in Appendix A.
There was a multi-colinearity problem when number of rooms and house size were entered into one model. The most common method to solve this problem is to leave one variable out of the model (Stock and Watson, 2003). The model with house size has the best fit and lowest error. Therefore, the variable number of rooms was left out. The diagnostics of the models show a high adjusted R-squared of 0.636, meaning that 63.6 % of the variance in LN transaction price is explained by the explanatory variables in the model. Figure 13 in Appendix C presents the residuals histogram of the main hedonic regression (4). The normal probability plot of the histogram supports the condition that the error terms are normally distributed (Stock and Watson, 2003).
A disadvantage of the hedonic price models is that it was not possible to observe all characteristics that influence the transaction price of a house. For example seller characteristics, like loss aversion and equity constraints, which were examined by Genesove and Mayer (2001). Despite that these data were not available in the WoON, they could lead to omitted variables bias. Therefore the results of the hedonic price models should be interpreted with caution.
5.3 Impact effects on transaction prices.
Now that lambda and beta are known, we examine how much of the differences in transactions prices are caused by differences in quality and by differences in constant quality indices. Therefore, as explained in sub-Section 3.1 equation (2) is subtracted from equation (1), resulting in equation (3). Column 2 in Table 5 shows the difference in the natural logarithm of the transaction price between FTB and NFTB for each year in the sample period (!.!!"#$− !.!!"#). Column 3 shows how much of column 2 can be addressed to the composition effect !.!!"#$ − !.!!"# !!"#. Column 4 shows how much of column 2 can be addressed to differences in constant quality indices (!!!"#$− !
! !"#). The rest term is presented in column 5 !!"#$− !!"# !
.!!"#$. However, this term is adjusted with the FTB dummy. The FTB dummy absorbs the different marginal contribution of the house size variable, as already mentioned in the previous sub-Section. These two variables caused the rest term to become very large11. Therefore we added an extra effect: ‘unobserved composition effect’, which is presented in column 6. The outcomes of the composition effect and constant quality indices strongly correspond with
11The rest term was around 38% each year. Without the difference in marginal contribution of the house
the results from Appendix B, were we omitted among other things the different marginal contribution of the house size variable. We assume that the unexplained composition effect is most likely caused by unobserved quality characteristics such as differences in quality within COROP-regions.
Table 5: Impact on transaction price.
Year Total Composition Constant Q. Rest Unobs. Comp.
96 13.89% 6.44% -2.35% -0.53% 10.32% 97 13.97% 5.35% -1.12% -0.58% 10.32% 98 11.65% 5.63% -3.92% -0.38% 10.32% 99 10.50% 5.34% -4.95% -0.21% 10.32% 00 16.25% 9.27% -3.74% 0.39% 10.32% 01 20.61% 11.81% -1.34% -0.19% 10.32% 02 21.89% 14.40% -2.93% 0.11% 10.32% 03 19.41% 10.58% -1.93% 0.43% 10.32% 04 21.11% 11.84% -1.30% 0.24% 10.32% 05 22.10% 13.76% -2.30% 0.31% 10.32% 06 21.73% 14.50% -3.37% 0.28% 10.32% 07 25.03% 16.64% -2.33% 0.39% 10.32% 08 27.60% 18.27% -1.11% 0.13% 10.32% 09 29.26% 19.16% -0.33% 0.11% 10.32% 10 26.69% 16.02% 0.57% -0.23% 10.32% 11 23.17% 13.28% -0.18% -0.26% 10.32%
It is apparent from Table 5 that the percentage of the composition effect more than tripled between 1999 and 2008, meaning that a larger part of the differences in transactions price between NFTB and FTB was caused by differences in quality. The differences in quality increases before the crisis, indicating that FTB bought houses of poorer quality compared to NFTB and were forced to buy houses at the bottom of the housing market. The impact of the composition effect is decreasing after the start of the crisis. This means that a smaller part of the differences in transaction prices was caused by differences in quality. Therefore, FTB bought houses of better quality compared to NFTB after the crisis, indicating an improved position on the Dutch housing market.
A detailed description about the impact of the constant quality indices is left out in this sub-Section, cause we already mentioned that the differences in the log price levels are not significant in most years.
5.4 The relative change of housing characteristics
This sub-Section focuses on how the differences in quality are expressed in individual housing characteristics, showing more precisely which poorer and better characteristics NFTB and FTB bought before and after the start of the financial crisis. Weights12 are given to respondents. Hereby the housing characteristics become representative for all Dutch households. Figure 3 showed that the transaction price indices of NFTB and FTB started to deviate after 1999 until 2008, when they started moving together. Therefore, the relative changes of the housing characteristics are examined between 1999-2008 (before the start of the crisis), and 2008-2011 (after the start of the crisis).
First of all, the physical size characteristics are compared. Figure 14 in Appendix C gives an overview of the development of the physical size characteristic for NFTB and FTB during the sample period.
Figure 5: Relative change of physical size characteristics before the crisis
Figure 5 shows that the average number of rooms13 decreased 11.61% for FTB before the crisis, while the average house size decreased 13.96%, and the average garden size decreased 55.73%. The average numbers of rooms decreased 3.44% for NFTB before the crisis. Average house size decreased 3.64%, and average garden size decreased 22.32%.
12
Weighting technique options are applied in the WoON datasets. By means of these techniques, weights are given to respondents, recovering the representative character as well as possible.
13Although, we left out the number of rooms in the hedonic price model due to multi-colinearity problems,
we think it is interesting to examine if FTB bought houses with more rooms after the crisis.
Numbers of rooms House size Garden area Roof terrace/ balcony NFTB -3,44% -3,64% -22,32% 6,78% FTB -11,61% -13,96% -55,73% 13,43% -60% -50% -40% -30% -20% -10% 0% 10% 20% 30% 40%
The average size of balcony/roof terrace increased 13.43% for FTB between 1999 and 2008 while for NFTB this was 6.78%. The decline of most physical size characteristics for both groups is interesting because that points out that the average Dutch household moved to smaller houses in 2008 compared with 1999. Furthermore it is interesting to see that the garden size of FTB more than halved in the pre-crisis period, meaning that FTB on average have bought houses with smaller or even no garden in 2008.
Figure 6: Relative change of physical size characteristics after the crisis
Figure 6 shows that the average number of rooms remained almost the same for both groups after the crisis. Average house size decreased with 2.5% for FTB in comparison with a decrease of 5.94% for NFTB. The average garden size decreased with 4.86% for FTB. This percentage was -15.19 for NFTB. The most surprising result is the difference between balcony/roof terrace sizes, a 28,50% increase for FTB, against a 33,96% decline for NFTB.
It can be concluded that the relative change of physical size characteristics was in favor for NFTB before the financial crisis. Oppositely, Figure 6 shows that the changes are in favor for FTB after the start of the financial crisis. So based on the physical size characteristics of houses bought, the position of FTB relatively improved compared with NFTB after the crisis. However, the negative signs in Figure 6 (except for roof terrace/balcony) indicate that FTB did not buy houses with larger physical size characteristics in 2011 compared with FTB in 2008.
Numbers of Rooms House size Garden Area Roof Terrace/Balcony NFTB -0,68% -5,94% -15,19% -33,96% FTB -1,16% -2,50% -4,86% 28,50% -60% -50% -40% -30% -20% -10% 0% 10% 20% 30% 40%
House types are a good indicator of quality. The variable is divided into six categories. Detached and semi-detached can be seen as better quality while apartments are usually of poorer quality. Figure 15 in Appendix C presents an overview of the types of houses bought by FTB and NFTB in 1999, 2008 and 2011.
Figure 7: Relative change of types of houses bought before the start of the crisis
Figure 7 shows the change of types of houses bought between 1999 and 2008. The most striking result to emerge from Figure 7 is that the number of FTB who bought an apartment nearly doubled. This percentage is 46.88 for NFTB. The number of FTB who bought a detached house decreased 56.14% between 1999 and 2008 while this percentage was -22.58% for NFTB. Furthermore, the number of FTB that bought a semi-detached house decreased 31.60% in comparison with an 8.92% decline for NFTB.
Figure 8: Relative change of types of house bought after the start of the crisis
Detached Detached Semi- Corner House Attached Apartment Other NFTB -22,58% -8,92% -0,22% -8,46% 46,88% 2,79% FTB -56,14% -31,60% -7,25% -2,79% 92,16% -53,54% -80% -60% -40% -20% 0% 20% 40% 60% 80% 100%
Detached Detached Semi- Corner House Attached Apartment Other NFTB -23,65% 9,44% 3,43% 26,42% -5,78% -62,54% FTB 2,72% -22,70% 28,32% -8,56% 16,06% 26,35% -80% -60% -40% -20% 0% 20% 40% 60% 80% 100%
Figure 8 represents the changes in types of houses bought between 2008 and 2011. Relatively more FTB bought detached houses, corner houses and apartments, while relatively more NFTB bought semi-detached and attached houses since the crisis.
Altogether, it is clear from Figure 7 that more FTB bought ‘poorer’ types of houses compared with NFTB in the period before the start of the crisis. The mixed results in Figure 8 show that there is no group that bought relatively better quality types of houses after the start of the crisis. Furthermore, the combination of positive and negative signs in Figure 8, makes it hard to conclude if FTB bought on average better types of houses in 2011 than in 2008.
The last variables that are examined are the COROP-regions. Figure 16 and 17 in Appendix C show the development of the geographic location of houses bought. The location of the house has a major impact on the price. Because there are no official guidelines for ‘better’ and ‘poorer’ COROP-regions, we qualified seven regions as ‘better’. First of all, we looked at the ten COROP-regions with the highest average housing price. Secondly, only COROP-regions that are located in the Randstad14 were taken into account. Seven locations fulfill these criteria: Groot-Amsterdam, Utrecht, Agglomeratie ‘s-Gravenhage, Gooi-Vechtstreek, Agglomeratie Leiden en Bollenstreek, Delft en Westland and Agglomeratie Haarlem.
14
The Randstad is an industrial and metropolitan conurbation in the Netherlands. With a population of 7.100.100, one of the largest seaports (Rotterdam) and one of the largest European airports (Schiphol) it is one of the most important economic and densely populated areas in Europe.
Figure 9: Relative change of geographic location of houses bought before the start of the crisis
Figure 9 shows that the differences between the two groups are small in Utrecht, Groot-Amsterdam and Den-Haag before the crisis. For Haarlem, Gooi-Vechtstreek and Leiden Bollenstreek the differences run up to more than 35% in favor of NFTB, indicating that more NFTB bought houses in these regions. The number of houses bought by FTB in Delft Westland rose with 55.17 percent, against a decline of 22.80 percent of NFTB.
Figure 10: Relative change of geographic location of houses bought after the start of the crisis
Utrecht Haarlem Agg Amsterdam Groot Vechtstreek Gooi Bollenstreek Den-Haag Leiden Westland Delft
NFTB 8,06% 31,87% 19,95% 98,06% 14,13% 4,23% -28,80% FTB 3,68% -6,73% 15,83% 33,53% -21,96% 0,89% 55,17% -80% -60% -40% -20% 0% 20% 40% 60% 80% 100%
Utrecht Haarlem Agg Amsterdam Groot Vechtstreek Gooi Bollenstreek Den-Haag Leiden Westland Delft NFTB -2,65% -33,47% 31,95% -5,13% 50,02% -5,14% -28,90% FTB 2,63% 37,63% 53,17% -53,33% 63,35% 7,27% 12,63% -80% -60% -40% -20% 0% 20% 40% 60% 80% 100%
Figure 10 shows that the differences are small in Utrecht after the start of the crisis. The differences in Groot Amsterdam are more than 20% in favor for FTB. In Leiden Bollenstreek and Den-Haag this percentage lays around 13 percent. In Haarlem and Delft Westland the difference run up to more than 40%. The number of houses bought by NFTB in Gooi Vechtstreek decreased with 5,13 percent, against a decline of 55,33 percent of FTB.
Altogether, NFTB bought houses in better locations compared to FTB before the crisis, and FTB bought more houses in better locations compared to NFTB after the crisis. Thereby are the positive percentages presented in Figure 10. These numbers indicate that more FTB bought houses in better COROP-regions in 2011 compared with FTB in 2008.
In addition to sub-Section 5.3, were we presented the differences in valuation of the housing characteristics between NFTB and FTB, this sub-Section focused on the development of these characteristics during the sample period. Based on the physical size characteristics, the type of house and location, it can be concluded that the position of FTB worsened compared to NFTB in the Dutch housing market before the crisis. Contrary, looking at the physical size characteristics and location, the position of FTB started to improve after the crisis in comparison with NFTB. However, when comparing FTB after the crisis with FTB before the crisis, this analysis showed that more FTB bought houses in better locations, while FTB did not buy better types of houses or houses with larger physical size characteristics.
6. Conclusion:
The economic recession, decreased real incomes, and precarious employment have caused some turbulent times in the Dutch housing market since the start of the financial crisis. The international literature writes about the weakened position of FTB, while Van de Minne and Francke (2013) state that the position of FTB in the Dutch housing market has improved since the crisis. The purpose of this thesis was to examine if the position of FTB has improved in the Dutch housing market since 2008 and if FTB have bought better quality houses since then. Therefore we compared the position of NFTB with FTB using hedonic price models. The empirical implications are tested with data from the WoON (2006, 2009, 2012).
Firstly, the house price trend of both groups was examined. The average house price of NFTB increased more than the average house price of FTB during 1999-2008. However, this trend reversed after the start of the financial crisis. Secondly, we examined factors that influence these price trends. The results show that there are almost no differences in the constant quality indices of NFTB and FTB. The impact of the composition effect is the largest explanatory factor for the differences in house prices. Differences in quality increased before the start of the crisis, indicating that the position of FTB worsened in the pre-crisis period. While the impact of differences in quality decreased after the start of the financial crisis, indicating an improved position for FTB in the Dutch housing market.
Returning to the hypothesis it can now be said that the position of FTB has improved, which is in line with the paper of Van de Minne and Francke (2013). However, FTB did not buy better types of houses or houses with larger physical size characteristics. Only the location of houses bought by FTB improved since the start of the crisis. The main conclusion from this thesis is that the discourse on the housing possibilities of FTB that has been prominent in the international literature (Forest and Yip, 2012; Mckee, 2012) does not find reverberation in the Netherlands, where FTB nowadays have enough chances to succeed.
Although this thesis has successfully provided a better insight in the position of FTB in the Dutch housing market, some limitations need to be acknowledged. The available sample period was until 2011. Therefore, the results do not cover a large post-crisis period. Another limitation of this thesis is that is unlikely to observe all
characteristics that influence the transaction price of a house. Unobserved seller characteristics, like loss aversion and equity constraints that influence the transaction price, which were examined by Genesove and Mayer (2001) could lead to omitted variable bias. Therefore the results should be interpreted with caution. Furthermore, for the reliability of the results we have to take into account that we did not know what the previous accommodation of a respondent was and therefore categorized all respondents between 18-35 as FTB.
Altogether, the findings do support strong recommendations to potential FTB who are planning to buy their first house. Especially after taken into account the drop in demand of NFTB, historically low interest rates and declined house prices. A key policy priority should therefore not to focus on FTB, but on NFTB, who are under water instead. For example: raising the tax deductibility of the negative home equity of NFTB. Another recommendation to the Dutch government is to focus on households who are living in poorer COROP-regions. Figure 10 in sub-Section 5.4 showed that more FTB bought houses in better regions after the crisis. A disadvantage of this trend is that less liquidity floats into the poorer regions, which causes a delay in the recovery of the housing market in these regions.
Where Meen (2013) stated that the action of FTB might be crucial in housing market depressions. Neuteboom and Brounen (2011) stated that the liquidity that FTB create at the bottom of a housing market will eventually flows through the entire sector, thus helping it out of a recession. The ability of FTB to restore the problems on the housing market was behind the scope of this thesis. Therefore, further research should focus on the effect of the improved position of FTB on the housing circulation in the Netherlands, and especially the effect on NFTB who are standing under water. Another possible area for future research should concentrate on the relationship between the position of FTB and the economic conditions? Is this relationship counter-cyclical? Answers to this question could be helpful for governments to prevent problems for FTB in future economic turbulent times. Finally, for more accurate results about NFTB and FTB in the Dutch housing market in the post-crisis period, further research should be done with the WoON (2015).
