1
Quantifying the effect of a change in the nominal and real
mortgage rates on the average housing prices in the
Netherlands.
University of Amsterdam
Ian IJnzen, 10659358, Economics and Business Specialization: Economics and Finance
Supervisor: Dr. Erasmo Giambona
ECTS: 12 Date: 14-06-2016
Abstract
The aim of this thesis was to quantify the effect of a change in real and nominal mortgage rates on the average housing prices in the Netherlands. To quantify this effect multiple OLS regression is used. The samples consisted of 84 and 76 quarterly observations (1995-I; 2015-IV) on multiple variables including both nominal and real mortgage rates. A negative relation between both nominal and real mortgage rates and housing prices has been found. This suggests that the increase in costs relating to increases in mortgage rates outweigh the effect of the tax deductibility of mortgage expenses. Thereby it is found that housing prices tend to be more sensitive to changes in mortgage rates in an environment of low interest rates.
2
Statement of originality
This document is written by Student Ian IJnzen 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
Table of Contents
1. Introduction 4
2. Literature review 5
2.1 Other studies using multiple OLS regression 6
2.2 Variables 6 3. Methodology 9 4. Data 10 4.1 Data construction 11 4.2 Summary statistics 11 5. Analysis 12 5.1 Hypothesis 12
5.2 Regression and results 13
5.3 Robustness check 15
6. Conclusion and discussion 15
6.1 Conclusion 16
6.2 Discussion 16
4
1.Introduction
Since the outbreak of the financial crisis in 2007 central banks around the world started to cut interest rates (Eichengreen et al. 2012). The European Central Bank first cut its main interest rates on November 12th 2008 and consecutively did so until the current rates of -0.40% ‘deposit facility’, 0.00% ‘fixed rate’ and 0.25% ‘marginal lending rate’ (ECB, 2016). This policy to encourage inflation led very low market rates. According to Baffoe-Bonnie (1998) the mortgage rate is directly linked to the regular interest rates. Historically the mortgage rates in the Netherlands are at unprecedented lows as well.
Furthermore, housing prices in the main cities in the Netherlands started to soar. The housing prices of the 4 biggest cities in the Ne
therlands (Amsterdam, Rotterdam, Utrecht and The Hague) rose on an average of 6,19% in 2015. Way above the national average since 1985 of 4,38% (CBS, 2016). The popular press writes about the possibility of a new housing bubble in Amsterdam in the Netherlands, some of the titles: “Fear of new housing bubble Amsterdam (AD, 2016)”, “Housing market explodes...” (Das Kapital, 2016), “Amsterdam in top global housing bubbles” (NOS, 2015). A reason given for the increasing prices in the housing market is the current low interest rates in the EU and the Netherlands. Lending money is cheap and saving rates are perceived to be the lowest in the recent past.
Connection between inflation, interest rates and housing prices have been found. Mishkin (1992) indicates a strong correlation between inflation and interest rates in the long-run. In periods of high inflation, housing and other tangible assets are viewed as more favorable (Harris, 1989). High inflation tends to increases the return from housing while decreasing that of stocks and bonds (Summers, 1981). According to Follain and Ling (1986) higher interest rates increases demand for housing, partly because of the tax deductibility of interest expenses.
On the other hand, Sommer, Sullivan and Verbrugge (2011) found in their research a significant negative effect of interest rates on housing prices. A decrease in mortgage rates from 6 to 1 percent leads to an increase in housing prices of 33 percent. This finding is supported by Baffoe-Bonnie (1998). He states that if the interest rates fall, everything else being equal, the real cost of housing decreases and therefore demand increases.
Currently, inflation and interest rates in the Netherlands are very low and some economists even fear a deflationary period. In an environment of low inflation and low
5
interest rates stocks are, as opposed to housing, seen as a more favorable investment (Summers, 1981). Nonetheless housing prices in the Netherlands rose rapidly the recent year. This study therefore aims at quantifying the relationship between mortgage interest rates and housing prices in the Netherlands using quarterly data from 1995-I to 2015-IV.
The remainder of this paper is organized as follows: First existing literature on housing prices will be discussed as well as different variables affecting housing prices. In chapter 3 the methodology used is explained and hypothesis are formed based on existing literature. The data used and accompanying summary statistics are described in chapter 4. Thereafter the results are presented in chapter 5. Finally, the most important findings and recommendations about future studies are discussed in chapter 6.
2. Literature review
Studies on housing prices can broadly be placed into two categories. On the one hand studies using a more finance based approach in which ratios likes rent or price-to-income are used to compare the returns of investing in housing relative to some other asset, or the costs of benefits of renting relative to buying are compared (McQuinn & O’Really, 2008). Case and Shiller (2003) state that fundamentals indeed are the real actors behind housing prices. But the problem with indicators is that they measure only one fundamental factor at a time (Case and Shiller, 2003).
The second approach is the econometric approach in which housing prices are regressed on different potential determinants of housing prices. In this manner more than one determinant of housing pricing is accounted for. However, McQuinn & O’Reilly (2008) argue that it is sometimes found that determinants which are believed to have a significant effect on housing prices appear to be non-significant in the regression model1). Despite this potential
caveat this study uses multiple regression since it is believed that this approach is most suitable to account for the multiple fundamental variables affecting housing prices. In the literature review first an overview will be given of previous studies which use multiple OLS regression. Then based on the literature, variables affecting housing prices will be described.
1 Other studies on the other hand found the expected significances of their variables; Harris (1989), Reichert
6
2.1 Other studies using multiple OLS regression
Harris (1989) used a similar approach to quantify the effect of interest rates on housing prices in the United States in the late 1970’s. He mainly focused his research on the effect of expected interest rates in combination with nominal interest rates, leading to the use of different regression parameters and a different time and sample period. In this thesis a more recent sample period is used and it will mainly focus on the effect of real and nominal interest rates on housing prices to leave out the possible estimation error of estimating future expectations.
K. Reichert (1990) assesses the role of mortgage rates on housing prices with a focus on regional differences in the United States in the period 1970-1980. Although Reichert comes to conclusions about the effect of mortgage rates on housing prices the sample period used is half the size of the period used in this thesis and also less recent.
In a more recent study McQuinn and O’Reilly (2008) assessed the effect of interest rates on housing prices in the Irish housing market. Although the paper is related to what will be discussed in this thesis, it uses some different regression parameters, a different sample market and a different sample period.
2.2 Variables affecting housing prices
In this paragraph different variables affecting housing prices are described. First the possible effects of interest rates on housing prices are elaborated. Secondly other possible determinants of housing prices are indicated.
Some contradictions about the effect of interest rates on housing prices are found in existing literature. On the one hand Follain and Ling (1986) indicate that higher interest rates tend to increase the demand for housing due to the tax deductibility of interest payments. Also Harris (1989) and Summers (1981) state that periods of high inflation generally increases the returns on housing and other tangible assets. Since a strong correlation between inflation and interest rates is found in the long-run (Mishkin, 1992), it could be that higher interest rates tend to increase the return on housing.
On the other hand, interest rates do increase the cost of owning a house. Since house purchases are often combined with mortgages and interest payment have to be paid over these mortgages, an increase in interest rates increases the costs for homeowners. This view
7
is supported by Sommer, Sullivan and Verbrugge (2011) and Baffoe-Bonnie (1998), they found a significant negative effect between mortgage rates and housing prices.
To better examine the above contradiction, first a difference between nominal interest rates and real interest rates has to be made. Harris (1989) states that the nominal mortgage rate is not always a good predictor of housing prices. In the late 1970’s housing prices rose although the nominal mortgage rates rose from 8 to 15 percent. Harris indicates that although the nominal mortgage rates rose, the real mortgage rates declined due to price appreciation and therefore the real costs of financing declined. Therefor in this study next to the nominal interest rates also real interest rates are used. According to the Fisher equation the ex post real interest rate equals nominal interest rates minus inflation (Mankiw, 2013).
Another possible determinant of housing prices is the amount that can be borrowed. McQuinn & O’reilly (2008) argue that the demand for housing is mainly a function of the amount that prospective home-buyers could lend from financial institutions. In the Netherlands the maximum value of a mortgage relative to the price of a house is only set by the government since 20122). In their study, McQuinn & O’Reilly (2008) made a model to
estimate the average amount a household can borrow from financial institutions. This is an annuity of the fraction of current disposable income kYt that is being used for mortgage
payments, discounted at the current mortgage interest rate for a horizon equal to the term of the mortgage τ McQuinn & O’Reilly (2008).
𝐵𝑡= 𝑘𝑌𝑡(
1−(1+𝑅𝑡)−𝜏
𝑅𝑡 )
Demographic changes could also be of influence on housing prices. In their study Mankiw and Weil (1989) found that demographic changes can affect the housing demand and therefore home prices. They found that the increase in US population during the baby boom after WWII, caused an increase in demand for housing when this baby boom generation entered the housing market. This led to a sharp rise in housing prices in the 1970s. In addition according to Himmelberg, Mayer and Sinai (2005), like most assets, the return on owning a house is twofold. The capital gain (in housing price appreciation) and dividend (in housing rents saved by owning a home). When rental prices increase the dividend gain of owning a home consequently increases as well. Therefore owning a home should
8
become relatively less expensive than renting a house. This in turn increases demand and should drive up prices of housing. Changes in rental prices should thus affect housing prices (Himmelberg, Mayer & Sinai, 2005).
Another key-variable affecting housing prices besides interest rates is income (Sommer, Sullivan & Verbrugge, 2011). Also, Davidoff (2006) shows that homeowners on an average are wealthier than renters and that the income of these homeowners positively correlate with housing prices. The common measurement of national income is the gross domestic product (GDP).
According to Reichert (1990) employment rates do affect housing prices as well. As stated earlier, changes in income are presumed to have an effect on housing prices. Therefore short-term variations in transitory income may cause housing prices to fluctuate (Reichert, 1990). The best way to grasp these short-term fluctuations in income according to Reichert is the employment rate. Baffoe-Bonnie (1998) complies and finds that employment is an up-to-date general economic indicator.
Also, Reichert (1990) argues that seasonality can affect housing prices as well. Harris (1989) found that housing prices tend to be highest in the second quarter and lowest in the fourth quarter.
Thereby Himmelberg, Mayer and Sinai (2005) state that sensitivity of housing prices to fundamental changes are higher in an environment of low long-term interest rates. Himmelberg et al. (2005) do not give a clear definition of when interest rates are presumed to be low. In this thesis interest rates are presumed to be low when the mortgage interest rates fall below the first quartile of the sample, in other words, the 25 percent lowest of data points.
A last comment about the Dutch housing market has to be made. In contrary to most other countries the Netherlands have a relatively generous tax system for home-owners (Boelhouwer et al., 2004). There is no taxation on imputed rents and mortgage rate payments are deductible from income taxes. Boelhouwer et al. (2004) indicate that this legislation decreases mortgage expenses which in case affect housing prices. When comparing results between the Netherlands and other countries if might therefore be taken into consideration that the overall price level in the Netherlands could be higher.
During the sample period two small changes have taken place in this legislation. In 2001 the terms on mortgages of which mortgage rates are tax deductible was set to maximally 30 years (Boelhouwer et al., 2004). Secondly since 2013 the mortgages rates payments are
9
only tax deductible when you also pay-off the mortgage itself (Belastingdienst, 2016). Since other macro-economic were present in both periods and the changes in legislation aren’t substantial, these changes are not accounted for in the regression.
To conclude, a clear consensus about the role of nominal interest rates on housing is not found in existing literature. Real interest rates on the other hand are mainly presumed to have a negative effect on housing prices due to increase of real housing costs. Other possible determinants of housing prices are the amount that can be borrowed from financial institutions, demographic factors, rental prices, income and employment rates. According to Reichert (1990) and Harris (1989) also seasonality can influence housing prices. Lastly it could be that the overall price level in the Netherlands is higher in comparison to other countries because of the tax deductibility of mortgage interest expenses.
3.Methodology
To quantify the effect of a change in real and nominal mortgage rates on the average housing prices in the Netherlands, multiple OLS-regression is used. Multiple regression provides a mathematical way to quantify how a change in one variable affects another variable, holding other factors constant (Stock, Watson 2015). Following from the above literature a first model can be constructed.
𝑃𝑅𝐼𝐶𝐸 = 𝛽0− 𝛽1𝑀𝑂𝑅𝑇𝐺𝐴𝐺𝐸 𝑅𝐴𝑇𝐸 + 𝛽2𝐺𝐷𝑃 + 𝛽3𝑃𝑂𝑃 + 𝛽4𝑉𝐴𝐿𝑈𝐸. 𝑇𝑂. 𝐿𝐸𝑁𝐷
+ 𝛽5𝑅𝐸𝑁𝑇 + 𝛽6−8𝑆𝐸𝐴𝑆𝑂𝑁𝐴𝐿𝐼𝑇𝑌 + 𝛽9𝐸𝑀𝑃𝐿𝑂𝑌𝑀𝐸𝑁𝑇 + 𝑒𝑖
In this model simply all relevant variables are added as independent variables which are then regressed on the average housing prices in the Netherlands. Some problems arise doing this. First a high correlation between population, GDP and the rent index lead to multicolinearity within the regression model. Second autocorrelation of the error term occurs (Durbin-Watson score of 0,209). Third problem is the interpretation of the results. An increase in population of one citizen does not compare well with a one percentage point increase in mortgage rates.
To deal with multicollinearity between GDP and Population the two variables are divided on each other. Since for this research it is more to research income of a household than the income of one person, it is multiplied by the average number of persons per household in the Netherlands in time t. This variable is added to the regression model as
10
GDPperHOUSEHOLD. Secondly the rent Index variable is dropped from the regression because of the high correlation with the GDPperHOUSEHOLD variable.
Furthermore, natural logarithms are taken. The relationship between the independent variables and the dependent variable is now one of elasticities. This means that a 1% change in the independent variable, accounts for a β% change in housing prices. This makes it easier to interpreted the results. Finally, a one period lagged autoregressive error term (AR1) is added to control for the autocorrelation in the error term. The complete regression model looks as follows, model 1:
𝐿𝑛(𝑃𝑅𝐼𝐶𝐸) = 𝛽0− 𝛽1𝐿𝑛(𝑀𝑂𝑅𝑇𝐺𝐴𝐺𝐸𝑅𝐴𝑇𝐸) + 𝛽2𝐿𝑛(𝐺𝐷𝑃𝑝𝑒𝑟𝐻𝑂𝑈𝑆𝐸𝐻𝑂𝐿𝐷) + 𝛽3𝐿𝑛(𝑈𝑁𝐸𝑀𝑃𝐿𝑂𝑌𝑀𝐸𝑁𝑇) + 𝛽4−6𝑆𝐸𝐴𝑆𝑂𝑁𝐴𝐿𝐼𝑇𝑌 + 𝛽7𝐴𝑅1 + 𝑒𝑖
The above model will be used to estimate the effect of both real and nominal interest rates, therefore two separate regressions will be done.
As is mentioned in the literature review Himmelberg et al. (2005) argue that the sensitivity of housing prices to fundamental changes are higher in an environment of low long-term interest rates. To test this concept, a third and fourth regression will be done. In these regressions only data were interest rates are in the first quartile of the sample will be incorporated. Since the dataset now is no longer a perfect time series, no autocorrelation in the error-term occurs. The AR1 variable is therefor dropped from the regression. The second model looks as follows, model 2:
𝐿𝑛(𝑃𝑅𝐼𝐶𝐸) = 𝛽0− 𝛽1𝐿𝑛(𝑀𝑂𝑅𝑇𝐺𝐴𝐺𝐸𝑅𝐴𝑇𝐸. 𝐿𝑂𝑊) + 𝛽2𝐿𝑛(𝐺𝐷𝑃𝑝𝑒𝑟𝐻𝑂𝑈𝑆𝐸𝐻𝑂𝐿𝐷) + 𝛽3𝐿𝑛(𝑈𝑁𝐸𝑀𝑃𝐿𝑂𝑌𝑀𝐸𝑁𝑇) + 𝛽4−6𝑆𝐸𝐴𝑆𝑂𝑁𝐴𝐿𝐼𝑇𝑌 + 𝑒𝑖
Again a separation will be made between nominal mortgage rates (3) and real mortgage rates (4).
4. Data
In this chapter the data sample is described. In paragraph 4.1 the origins of the data as well as transformations done to variables are specified. In 4.2 the descriptive statistics of the dataset are described.
11
4.1 Data collection
Data on housing prices in the Netherlands is, as are most data in this research, collected from the Central Bureau of Statistics of the Netherlands (CBS). The CBS is responsible for collecting and presenting reliable data based on scientific-, social-, macro-economic- and price- variables (CBS, 2016). The data on housing prices used is quarterly data from 1995-I to 2015-IV. The data on mortgage rates is collected from the CBS over the period 1995-I to 2003-IV and by ‘De Nederlandse Bank’ over the period 2003-I to 2015-2003-IV. Although the terms of the mortgages differ (5-year period C.B.S. to 1> to ≤ 5 D.N.B.) the data in the overlapping period maximally differs 0.0030 (0.30%). To transform nominal interest rates to real interest rates, the Fisher equation is used.
Unemployment measures are collected from the CBS. As well is data on rents, which are only available as year-on-year change in percentages. To fit this measure with the rest of the variables, first the geometric mean is taken to transform the data from yearly to quarterly data, thereafter an index is made based on these quarterly changes.
To compute the income of a household in a certain time period, the GDP per capita measure is multiplied by the average amount of persons per household in each period, ranging from 2,35 in 1995 to 2.18 in 2015 (CBS, 2016). Since no data is available on either the Value.to.Lend ratio nor on the fraction of household income used for mortgage payments and the R2 already is high (0.996), this variable is omitted from the regression.
4.2 Descriptive statistics
Table 1 gives an overview of the descriptive statistics of the dataset. Differences can be seen between the nominal and real interest rates. Both minimum and maximum values are logically lower for real interest rates. Also the standard deviation of real interest rates is lower. The highest relative standard deviation (Mean divided by std. dev.) is that of housing prices.
The average housing prices in the Netherlands were the lowest in the start of the sample in 1995. Prices gradually increased to the maximum value of 259.425 in the third quarter of 2008. Thereafter prices declined till the second quarter of 2013 to start rising again for the rest of the sample period.
12
Table 1: Summary statistics
Summary Statistics
Variables (N=84) Mean Std. Dev. Min. Max.
Price 196238 50437 89792 259425
Rates Nominal 4.84 1.07 2.64 7.77
Rates Real 4.54 0.95 2.59 6.65
GDP per Household 73518 12709 48133 90298
Unemployment 5.39 1.48 3.1 8.4
In the data, clear declining trends are visible in both nominal and real interest rates. Clear increasing trends are visible in housing prices and GDP per household. The trend in unemployment rates cannot easily be specified.
Furthermore, descriptive statistics of the first quartile of both real and nominal interest rates are given in Table 3. A substantial decrease in standard deviation in both nominal and real rates is seen.
Table 2: summary statistics low interest rates
low interest rates
Variables Mean Std. Dev. Min. Max.
Low Rates Nominal (N=21) 3,49 0,48 2,64 4,14
Low Rates Real (N=20) 3,29 0,42 2,59 3,83
5. Analysis
In this section, first hypotheses will be formed. Secondly the regression and results will be discussed. In paragraph 5.3 a robustness check is done to examine the differences between model 1 and model 2.
5.1 Hypothesis
Following from the research objective: Quantifying the effect of a change in the real and nominal mortgage rates on the average housing prices in the Netherlands. Hypotheses about β1 can be formed. Based on previous studies it is expected that real mortgage rates have a
13
Some studies suggest that there might be a positive relation between housing prices and nominal mortgage rates because of the tax deductibility of interest rates (Follain & Ling, 1988).However, in this sample the negative correlation between nominal rates and housing prices is stronger than that of real mortgage rates (-0.643; -0.522). Based on this negative correlation, a negative relation between nominal mortgage rates and housing prices is expected. Hypotheses for the first two regressions therefore are:
Regression 1: 𝐻0: 𝛽1 = 0 𝑣 𝐻1: 𝛽1 < 0
Regression 2: 𝐻0: 𝛽1 = 0 𝑣 𝐻1: 𝛽1 < 0
Secondly hypotheses about the third and fourth regression are formed. Based on existing literature it is expected that in an environment of low interest rates the sensitivity of housing prices to changes in mortgage rates is higher (Himmelberg et al. 2005). Hypotheses for the last two regressions are:
Regression 3: 𝐻0: 𝛽1𝑟𝑒𝑔3= 𝛽1𝑟𝑒𝑔1 𝑣 𝐻1: 𝛽1𝑟𝑒𝑔3> 𝛽1𝑟𝑒𝑔1 Regression 4: 𝐻0: 𝛽1𝑟𝑒𝑔4= 𝛽1𝑟𝑒𝑔2 𝑣 𝐻1: 𝛽1𝑟𝑒𝑔4> 𝛽1𝑟𝑒𝑔2
5.2 Regression and results
In the first regression model the nominal interest rates including all control variables are regressed on the average housing prices in the Netherlands. The second regression regresses the real interest rates instead of the nominal interest rates. All other variables are the same. To eliminate the autocorrelation of the error term, next to the added AR1 variable an iterative Cochrane-Orcut procedure was employed in all regressions. The results of the regression are summarized in Table 3.
In model (1) all variables except GDP per household and Q1 have significance levels of 5 percent or lower. This is the same for model (2). The R2 in both model (1) and (2) can be
considered high, R2=0.996 and 0.994 respectively. Indicating that almost all variance in
housing prices could be explained by the variables in the regression model.
The main focus of this thesis is the effect of nominal and real interest rates on housing prices. In this regression analysis a significant relation between nominal mortgage rates and housing prices is found. The beta indicates that a 1% relative increase in mortgage rates
14
decreases housing prices by 0.047%. So an increase in mortgage rates from 5 to 6% (a relative increase of 20%) decreases housing prices by 0.94% (-.047* 20 percent).
A significant relation between real interest rates and housing prices is indicated as well. The effect of -0.046% is slightly smaller than that of nominal interest rates. Other studies found somewhat higher levels of the coefficient of real mortgage rates. Reichert (1990) for example found Beta’s ranging from -0.05 to -0.43 in 9 regional regressions in the U.S. A possible explanation is the use of nominal instead of real housing prices in this study.
Table 3: results Netherlands (1995-I:2015-IV)
Variables (1) (2) (3) (4) Constant 0.833** 0.808** 11.6*** 7.25*** (0.023) (0.029) (0.000) (0.004) Rates -0.047** -0.046** -0.186*** -0.15* (LnRatesR/N) (0.048) (0.040) (0.000) (0.062) GDP household 0.038 0.042 0.122 0.510** (LnGDPHousehold) (0.572) (0.542) (0.282) (0.015) Unemployment -0.048*** -0.046*** -0.241*** -0.286*** (LnUnemplyment) (0.003) (0.007) (0.000) (0.003) Q1 0.008 0.000 -0.021 0.008 (0.220) (0.969) (0.129) (0.759) Q2 0.017*** 0.015** 0.002 -0.003 (0.008) (0.027) (0.826) (0.910) Q3 0.026*** 0.020*** -0.003 0.006 (0.000) (0.000) (0.805) (0.856) AR1 0.908*** 0.907*** - - (0.000) (0.000) - - DW 1.996 2.010 2.193 1.886 R2 0.996 0.994 0.872 0.667 n 84 76 21 19
In compliance with existing literature (Reichert, 1990; Harris, 1989) seasonality in the housing market is found. The third quarter seems to be the quarter with the highest prices
15
followed by Q2. No significant difference between the fourth quarter and the first quarter is found. Also, in contrary to existing literature no significant relation between GDP per household and housing prices is indicated in both models.
From regression (3) and (4) it is clear that the lower sample size limits the accuracy of the model. The R2 in both models is lower and the differences between coefficients other than
the mortgage rates are higher. Also the sample is too small to explain seasonality in the housing market. However, the main aim of these regressions was to see whether housing prices are more sensitive to mortgage rates changes in an environment of lower interest rates. Coefficients on both real and nominal interest are higher in these models. This Indicates that indeed there is a relation between lower interest rates and the sensitivity of housing prices to these rates.
To conclude, in compliance with existing literature the regression models imply a significant relation between both real and nominal mortgage rates and housing prices. Also significant coefficients are indicated for the seasonality effect and unemployment. However, the model failed to explain the effect of income on housing prices. Furthermore, a positive relation between lower interest rates and the sensitivity of housing prices to interest rates is found.
5.3 Robustness check
Since no autocorrelation occurs in the model for regression (3) and (4), the autoregressive error-term AR1 is omitted from the model. The absence of this variable may affect the outcomes of these regressions. As a robustness check, regression (3) and (4) are done again with the AR1 variable included in the model. By adding the variable, the effect of both nominal and real mortgages decreases slightly. The coefficient on nominal mortgage rates decreases from -0.186 to -0.15 and the coefficient on real mortgage rates decreases from -0.15 to -0.128, also the significance of real interest rates increased slightly from 0.062 to 0.043 respectively. However, the decrease in coefficients does not alter the overall indication that in an environment of low interest rates housing prices are more sensitive to changes in mortgage rates.
16
6.1 Conclusion
The aim of this thesis was to quantify the effect of a change in real and nominal mortgage rates on the average housing prices in the Netherlands. In existing literature, a relation between mortgage rates and housing prices has been found. On the one hand Follain and Ling (1988) indicate a possible positive relation between housing prices and interest rates due to the tax deductibility of mortgage rates expenses. Other studies (Harris 1989; McQuinn & O’reilly 2008; Reichert 1990) imply a negative relation between housing prices and interest rates due to increases in the real costs of owning a home.
To quantify the effect of a change in mortgage rates, multiple OLS regression was used. The sample included 84 quarterly observations (1995-I; 2015-IV) on multiple variables including both nominal and real mortgage rates. To account for autocorrelation of the error term a Cochrane-Orcut procedure was employed. A negative relation between both nominal and real mortgage rates and housing prices is found. This suggests that the increase in costs relating to increases in mortgage rates outweigh the effect of the tax deductibility of
mortgage expenses. Also, the outcome of this regression analysis supports the theory of Himmelberg et al. (2005) that housing prices tend to be more sensitive to changes in mortgage rates in a low interest rate environment.
6.2 Discussion
In this thesis nominal housing prices are used to estimate the effect of different variables on housing prices. Since inflation has a negative effect on housing prices, the overall price level in real terms is lower than housing prices used in this study. This could have an influence on the outcome of the regression and might explain the insignificance of income in three out of the four models.
Due to multicollineairity the rentindex variable is dropped from the regression. So is the variable on the value to lend ratio due to lack of information. Existing literature suggest a relation between both variables and housing prices. Therefore, the models used in this thesis do not fully grasp all variables affecting housing prices.
Although the absence of the first-period lagged autoregressive error does not alter the indication of an increase of the effect of mortgage rates when in a low interest rate
17
environment, it does change the coefficients. Therefore direct comparison between regression models (1),(2) and (3),(4) may not be accurate.
Lastly, as discussed in the literature the tax deductibility of interest rates in
Netherlands may have increased the overall price level of housing in the Netherlands. The outcomes in all regression analysis are not corrected for this possible increase in price level. International comparison with countries that do not have the same taxation legislation could thus be imprecise.
18
Reference list
Baffoe-Bonnie, J., (1998); The dynamic impact of macroeconomic aggregates on housing prices and stock of houses: A national and regional analysis. Journal of Real Estate Finance
and Economics, 17, 179-197. Doi:10.1023/A:1007753421236
Boelhouwer, P., Haffner, M., Neuteboom, P., & De Vries, P., (2004) House prices and income tax in the Netherlands: an international perspective. Housing Studies, 19, 415-432, DOI: 10.1080/0267303042000204304
Case, K. E., & Shiller, R. J., (2003); Is there a bubble in the housing market? Brookings Papers
on Economic Activity, 3, 299-342. Doi: 10.1353/eca.2004.0004
Eichengreen, B., Mody, A., Nedeljkovic, M., & Sarno, L. (2012); How the subprime crisis went global: Evidence from bank credit default swap. Elsevier, 31, 1299-1318. Doi:
10.1016/j.jimonfin.2012.02.002
Follain, J. R., & Ling, D. C., (1988); Another look at tenure choice, inflation and taxes. Journal
of the American Real Estate & Urban Economics Association, 16, 207-229. Doi:-
Harris, J. C., (1989); The effect of real rates of interest on housing prices. The Journal of Real
Estate Finance and Economics, 2, 47-60. Doi: 10.1007/BF00161716
Himmelberg, C., Mayer, C. J., & Sinai, T. M. (2005); Assessing high housing prices: Bubbles, fundamentals and misconceptions. Journal of Economic Perspectives, 19, 67-92. Doi: 10.3386/w11643
Mankiw, N. G., & Weil, D. N., (1989); The baby boom, the baby burst, and the housing market. Elsevier, 19, 235-258. doi:10.1016/0166-0462(89)90005-7
Mankiw, N.G., (2013); Macroeconomics. Houndmills, England: Palgrave Macmillan McQuinn, K., & O’Reilly, G. (2008); Assessing the role of income and interest rates in determining house prices. Elsevier, 25, 377-390. Doi: 10.1016/j.econmod.2007.06.010. Mishkin, F, S., (1992); Is the Fischer effect for real?: A reexamination of the relationship between inflation and interest rates. Journal of monetary economics, 30, 195-215. Doi: 10.1016/0304-3932(92)90060-F
19
Reichert, A. K. (1990); The impact of interest rates income and employment upon regional housing prices. The Journal of Real Estate Finance and Economics, 3, 373-391. Doi:
10.1007/BF00178859
Sommer, K., Sullivan, P., & Verbrugge, R. (2011); Run-up in the house price-rent ratio: How much can be explained by fundamentals.
Stock, J. H., Watson, M. W. (2015); Introduction to econometrics. Harlow, England: Pearson. Belastingdienst., (2016); Woning, uw hypotheek of lening, wat is aftrekbaar?,
(hypotheek)rente. Retrieved from: link
Bloomberg. (2016, April 19). Quotes, Netherlands Government 10 Yr Bond NA. Retrieved from: http://www.bloomberg.com/quote/GNTH10YR:IND.
European Central Bank. (2016, April 19). Statistics, monetary operations, Key ECB Interest Rates. Retrieved from:
https://www.ecb.europa.eu/stats/monetary/rates/html/index.en.html. Central Bureau of Statistics. (2016, April 19-21). Statline database.
