Reduction of the mortgage interest deduction:
What is the risk for a household of ending up
underwater?
Peter Ronald de Graaf
Master student International Economics & Business
University of Groningen, Faculty of Economics and Business Student number: 1684930
Email: p.r.de.graaf.1@student.rug.nl
2 Abstract
This paper investigates the relationship between house prices and the mortgage debt of households after the reduction of the mortgage interest deduction. House prices as well as mortgage debt go down when the mortgage interest deduction is reduced. This would result in complications if the mortgage debt of households goes down later than the house prices. So far no research has been conducted on this order yet. This paper fills this gap, using a sample of European countries where a reduction in the mortgage interest deduction has taken place already to see what goes down first, house prices or mortgage debt.
3 I. Introduction
Due to the discussion at the moment in the Netherlands about the mortgage interest deduction and what to do with it, it is interesting what the expected effects of a limitation of the mortgage interest deduction on the market are. In this discussion it is often said that other countries do not have a mortgage interest deduction any more. As is show in table 1, this is not true. Most countries still have mortgage interest deduction, it is only less extensive than it is in the Netherlands because it has been reduced already in the past. In only some countries the policy has really been eliminated completely.
This paper investigates what has happened after the reduction of the mortgage interest deduction in the countries which have reduced or eliminated the mortgage interest deduction. Using these results from the past it can be seen what is to be expected for the Netherlands. Evidence from research on the countries in table 1 shows that house prices go down, or at least grow at a slower pace, after the reduction of mortgage interest deduction. (see Ball 2002; Englund, Hendershott and Turner 1995; Agel, Berg and Edin 1995) As Capozza, Green and Hendershott (1996) find debt is sensitive to the mortgage interest deduction as well. Less debt is used for buying a house when the deductibility is reduced. The question that will be discussed in this paper is which of the two
variables goes down first and which variable lags behind. The hypothesis used to investigate this is: If the mortgage interest deduction is reduced, then the decline in house prices lag behind the decline in mortgage debt as a percentage of GDP. When the hypothesis is not right and the opposite is true then many households would end up with a larger mortgage debt than their house is worth. In other words, they are underwater.
4 Table 1 Mortgage interest deduction in selected countries, year 2003
Country Mortgage interest Deduction? Yes/No Maximum tax rate Remark
Austria Yes 50% Tax deductible as special expense up to
a limit that goes to 0 as annual income increases
Belgium Yes 50% Tax deductible up to income from
immovable property, buts some additional deduction possible.
Deduction decreases over time, for at most 12 years
Denmark Yes 15% Deductible from capital income that is
subject to income tax, unlimited
Finland Yes 29% (flat
rate)
Normally deductible from capital income that is subject to income tax
France No - Gradually abolished over the period
1991-2000. Only tax credit in very special cases for loans before 1998
Germany No - Abolished in 1986 with the
introduction of a subsidy scheme (Eigenheimzulage)
Greece Yes 40% Since 2003, tax credit of 15% for
annual mortgage interest, subject to limits
Ireland Yes 42% “Tax credit” at 20% standard income
tax rate, at source, with (low) limits
Italy Yes 45% Tax credit of 19% for annual mortgage
interest, with limit
Luxembourg Yes 38% Tax deductible, with (low) limits Netherlands Yes 52% Tax deductible without limit, for at
most 30 years
Portugal Yes 40% Tax credit of 30% of mortgage interest
and amortization, with a limit
Spain Yes 45% Tax credit of 25% of amounts paid
(principal and interest) in first two years, 20% thereafter up to a threshold, and 15% for payments above the threshold
Sweden Yes 30% Interest not attributable to any source
of income is deductible from capital income
UK No - Abolished in a number of steps (1983,
5 II. The discussion about the mortgage interest deduction
Efficiency gains
Many believe that the mortgage interest deduction should be discarded, or at least reduced, in the Netherlands. For example van Ewijk, Jacobs and de Mooij (2006), van Wijnbergen et al. (2012), the political parties like the PVDA, D66, Groenlinks and Christenunie believe so, however not everyone agrees. Other political parties in the Netherlands, like the PVV, CDA and the VVD believe that the mortgage interest deduction should stay untouched. They are supported by a
research conducted by Follain (1998, 196) who says that in the USA “The additional tax revenue to be gained by eliminating the MID is far less than normally estimated” According to Follain (1998) it is not a good idea to eliminate the deduction in the USA. An important issue that he brings forward is portfolio reshuffling. Many, mainly rich, people can choose between debt financing and financing through their savings. With the mortgage interest deduction still in place, it is more profitable for them to use debt financing. This is for the reason that the after tax cost of this way of financing is low. As soon as the interest deduction is reduced or eliminated, it is no longer the best way for them to finance their house with debt. What these households do is lower the amount of debt they have and use more of their savings to finance their house. This is known as portfolio reshuffling and results in the fact that the tax revenue of the government only goes up by, in the case of the USA, ten billion. This is much lower than the forty billion that the government now spends on the mortgage interest deduction. For the government therefore, it means that the tax revenue will go up, but not by the amount that they spend on the policy now. The result is a relatively small gain for the government. The basics of these arguments are applicable to the Netherlands as well.
lump-6 sum taxes. These lump-sum taxes do not distort the decisions that the economic agents make. The disadvantage is that it is very difficult to find a lump sum tax that taxes everyone, without altering the decisions that the agents make. (Auerbach 2001). The concept of deadweight loss is not new, in 1938 Harold Hoteling already displays figure 1 in his paper. Figure 1 depicts the deadweight loss which is a result of the tax. After Hoteling many other economists have investigated this
phenomena, under which Harberger (1954, 1964a, 1964b, 1971). For the reason that Harberger contributed heavily to the understanding of the deadweight loss, the triangle LBM is often called the Harberger triangle these days. Assuming that point B is the efficient allocation in an efficient market with a tax of BL imposed on the good this decrease the price that the seller gets. At the same time it increases the price that the buyer has to pay. Due to this, the sellers are not willing to supply the same amount of goods as before, while the buyers do not want to buy as much as before any longer. For this reason, consumer surplus becomes smaller, as well as the producer surplus. This decrease in consumer and producer surplus is not fully recouped by increased tax revenues. The tax, which moves the market from the efficient level to an inefficient level results in the loss of the triangle LBM.
Figure 1 Deadweight loss
7 The mortgage interest deduction is a subsidy. The idea behind this is however exactly the same as with a tax. As Sørensen (2010, 13) says if “ housing is subsidized in one way or another, the effective tax rate τh is negative” A subsidy therefore does have the same implications for the deadweight loss. Meaning that the subsidy results in higher prices for the seller, while for the buyer this results in lower prices. The difference is the subsidy, depending on the size of this subsidy the deadweight loss of taxation can be estimated. It has to be noted that when estimating the
deadweight loss, one should not be worried about to whom the costs and benefits go to. It is all about social welfare. (Harberger 1971) Furthermore Harberger (1971) already knows that for estimating the deadweight loss the competitive supply and competitive price is needed. Knowing what the competitive price and supply is, we know the starting point, point B in figure 2. Now depending on the shape and elasticities of the supply and demand curve, as well as the size of the subsidy the deadweight loss can be calculated. Van Ewijk, Jacobs and de Mooij (2007, 332) try to estimate the deadweight loss, resulting from the taxes and subsidies in the Netherlands and the result is that “assuming perfectly inelastic housing supply, eliminating the fiscal subsidy on housing and using the receipts to lower labour income taxes, increases welfare by 6 billion euro in 2006” Here it has to be said that the deadweight loss resulting from the mortgage interest deduction will be smaller. This is for the reason that it is highly unlikely that the housing supply is perfectly inelastic. Furthermore, as mentioned before van Ewijk et al. (2007) speaks about all subsidies to the housing market. The mortgage interest deduction is only one, although a significant one, subsidy and not all six billion Euro can therefore be gained by only eliminating the interest deduction. However, all in all, the gains from the deadweight loss that disappears will still be quite large.
International organizations, like the International Monetary Fund (IMF) and the
8 home ownership. This goal has not been achieved according to the OECD and this is supported by Atterhog (2005). He shows in his paper that the Netherlands have one of the lowest rate of home ownership in the selected countries. This would mean that the very generous interest deductibility that the Netherlands have has not been very effective compared to these countries where the mortgage interest deduction was less generous.
High debts
The other international organization, the IMF (2010, 5) says that to lower the loan to value ratio (LTV ratio) of the Netherlands “limitations on mortgage interest deductibility are desirable”. The IMF believes that the LTV ratio in the Netherlands is too high. The fact that the Dutch have a too large mortgage debt has been recognized by the Dutch policy makers. This results in the fact that the maximum mortgage for households in the Netherlands nowadays is 110% of the value of the home. Next to this the mortgage cannot be larger than three and a halve to four and a halve times the gross salary. Before your mortgage could be as large as six times your gross salary. (Brosens 2009). The result is that households are not able to borrow as much as they could before. Following the reasoning of Campbel and Cocco (2011) reducing the maximum amount that can be borrowed is positive. The reason for this is that they believe that the past financial crisis stems from the fact that households were able to borrow multiple times their income for a mortgage.
Giving a too large mortgage to persons who cannot afford it, is dangerous . As we know the crisis in which the world economy is in now, is also frequently called the debt-crisis. Not only the governments of Greece, Portugal, Spain, Ireland and Italy feel the consequences of too much debt also consumers do. In fact the crisis started off with the sub-prime crisis. This is due to the reason that mortgages were collected in CDO's in order to conceal the that these mortgages were bad mortgages. This means that the households who received these mortgages were very likely to default on their payments. To get them sold, the bad mortgages were packed together and in this way the packages received a higher credit rating. When house prices in the USA started to decline, the borrowers started to default massively on their mortgage. With as a result that many banks that owned these packages of loans did not receive their money, a worldwide crisis was started.
(Longstaff 2010).
9 Furthermore, besides defaulting when you can simple not pay the mortgage any longer, theory shows us that other reasons can arise. “The “option model” of mortgage default is traditionally interpreted as implying that borrowers should default if and only if they have negative equity in their home.” (Elul et al. 2010, 2) In common terms, to have negative equity, is often referred to as being “under water”. In the case of the housing market it means that your mortgage is larger than the value of your house. This however, does not mean that a borrower defaults immediately when the mortgage is only a bit larger than the value of the house. Research by Bhutta et al. (2010, 27) shows “that the median borrower does not walk away until equity has fallen to -62 percent of the house value” In the model of Campbel and Cocco (2011) default is triggered by negative home equity. Not everyone with a negative home equity defaults, but it is more likely that they will
default. According to Deng et al.(2000), who use an option approach, borrowers tend to wait before defaulting until they are almost sure that the value of their house will not exceed their mortgage debt ever again. This is because when they default they lose the option that the house prices will go up again and that they can benefit from this. So, when the LTV ratio is very high then the borrowers will default, not when the LTV ratio is only a bit larger than one. Evidence that the LTV is lowered after the removal of (part of) the mortgage interest deduction has been found. In 1990 Follain is not yet sure that this is the case, however in later publications it is shown that the LTV ratio is adjusted after the elimination of the mortgage interest deduction. “it is hard to be dogmatic without more evidence as to how households might adjust their LTVs” (Follain 1990, 140) Ten years later, Dunsky and Follain find that that the LTV ratio is negatively and strongly related to the after-tax cost of mortgage debt. They conclude that the borrowers who can rearrange the financing of their house will do so. This means that after the after tax costs of mortgage debt has gone up, the borrowers will decrease their loan. Capozza et al. (1997) assume as well that the LTV ratio goes down after the removal of the mortgage interest deduction. As well as Jackson (2005, 26) does: “Eliminating the mortgage interest deduction separate from tax reform would have several effects. It would tend to reduce debt relative to equity by ending the neutrality between equity- and debt-financing”
10 Negative sides to elimination
So when the mortgage interest deduction is limited the market starts moving to a new equilibrium with a lower LTV ratio and a smaller deadweight loss. The new equilibrium has a higher social welfare than the old equilibrium. However the path towards the new equilibrium is what most people are worried about. During this transition it is unknown what happens to the LTV ratio, which can have important negative consequences.
Theory and practice shows that after the removal of the mortgage interest deduction, house prices go down. The estimates differ substantially, but it is shown that house prices go down by at least 10%. Capozza et al. (1997) have estimated that the removal of the mortgage interest deduction in the United States would result in a price decline between 7% and 23%. They assume that the loan-to-value ratio of the households is 0,25, with a higher loan-to-value ratio, they expect that the decline in house prices would be bigger. The LTV ratio in the USA is normally around 80%, so the decline in house prices would be larger than the estimation by Capozza et al. (1997). An estimate by Boelhouwer and de Vries (2005) for the Dutch housing market, says that they expect that the price will go down by 23 to 30%. In this estimation they consider that the owner occupied housing market will be completely defiscalised. It has to be noted that in this case not only the mortgage interest deduction is eliminated, but for example also the tax on imputed rent is abolished. For this reason the estimation will not be exact, but what is clear is the large decline in house prices. For Sweden, a country where the interest mortgage deduction has been reduced in 1991 already, it is said that “the observed 30 per cent plunge in real house prices and virtual disappearance of construction are of concern, they are not mostly attributable to the tax/subsidy reform. By our estimates only a third to a half of the price decline can be so attributed.” (Englund et al. 1995, 353) This estimate therefore suggests that 10-15 percent of the decline in house prices is due to the reduction of the mortgage interest deduction.. This seems not to be that much, but the interest deduction has not been eliminated completely, it has only been reduced. With a complete
11 1980s, real house prices then fell by almost a fifth in the first half of the 1990s” At first sight the abolishment of the mortgage interest deduction did not result in a declining price. Yet in 1991, when further measures were taken to abolish the deduction the house prices declined a lot.
It is shown that after the elimination of the deduction the market will reach a new
equilibrium, with lower house prices and a lower LTV ratio. However, in the literature it is only said that both will go down, not whether they go down instantly or with a lag. Thus in the path towards this lower house prices and lower LTV it might be true that the LTV first goes up, where after it goes down. This is possible for example when house prices decrease first and that two years later the mortgage that the household has is altered. In the first year the LTV ratio goes up and after two years the LTV ratio goes down. On the other hand, it might also be that the house prices decline faster than the mortgage debt does and that mortgage debt declines longer. Immediately after the reduction this results in a higher LTV while after a few year the LTV goes down. Many other dynamics can occur in this ratio, and without knowing what happens exactly to the nominator and denominator of this ratio, one cannot make sure why the ratio gets larger or smaller.
In the short run therefore, households who already had a very high ratio could end up with an even higher ratio. This could result in the fact that it becomes profitable for them to default on their mortgage, as discussed in the literature above. When only few households default due to a very high LTV it might be a personal drama for these households. However when only few default, no bank or country will get in trouble. It will be different when the aggregated LTV of a country becomes too high. In this case many households have a very high LTV, resulting in many defaults possible on mortgages, which might result in instabilities.
What is of interest is the LTV ratio of the whole country. As de Haan and van den Berg (2011, 0, abstract) “The Loan-to-value ratio for a group of households can be defined as the total amount of outstanding mortgage loans divided by the total value of the housing stock”. Hence, for a country all the values houses that are bought by households and all mortgage debt has to be
summed up in order to calculate the LTV ratio. However, as discussed, within the LTV ratio it is not clear what happens to the house prices and the mortgage debt, separate from each other. It is more useful to see what happens to the house prices and mortgage debt separately. For this reason in this research the focus will be on house prices and total mortgage debt of a country as percentage of the gross domestic product (GDP). In this way the changes in the country level LTV ratio can be estimated.
12 LTV ratio for the reason that Hess and Holzhausen (2008, 8) say: “Countries with higher loan-to value ratios usually also feature higher mortgage debt relative to GDP”. This is confirmed by the IMF (2008, 25) “higher LTV ratios and thus, presumably, higher stocks of mortgage debt”.
Meaning that an increase in mortgage debt as percentage of GDP leads to an higher LTV ratio and a decrease in a lower LTV ratio. The two variables are interlinked with each other, making it sensible to use these variables when trying to make the step from micro-level analysis to macro-level analysis.
The Netherlands, have on average a very high LTV ratio, in 2008 this was 110% on average, with an mortgage debt as percentage of GDP of 99.2%. These are very high levels, compared to the rest of the world. For this reason the short movement of the house prices and the mortgage debt is very important for the Netherlands. When in the path towards the new equilibrium the decline in house prices is very large directly after the elimination of the policy, the LTV ratio can move quickly to critical levels of defaulting where Bhutta et al. (2010) write about. With the risk of large instabilities occurring in a country. However, these instabilities will only occur when the house prices decline first and the mortgage debt declines later than the house prices. When this is the other way around, so that the mortgage debt declines first, where after the house prices decline, then these problems will not occur, resulting in the fact the argument of declining house prices is not as good an argument as it seems to be now to keep the mortgage interest deduction in place.
III. Expectations
The results that are expected, is that after the elimination of the mortgage interest deduction the mortgage debt as a percentage of GDP goes down first. After the decline in mortgage debt the house prices will go down. The reason for this is that when households want to buy a house a constraint on this is the ability for them to pay for this. The mortgage interest deduction results in a lower after tax cost to borrow money, which results in the fact that they can borrow more for the same amount of money. When the mortgage interest deduction is eliminated, the after tax cost for households becomes higher. The result is that they cannot borrow as much as before. They cannot afford to buy expensive houses any longer. Demand for these houses therefore goes down.
13 interest deduction is reduced, then the decline in house prices lag behind the decline in mortgage debt as a percentage of GDP.
My personal expectation is that this is at least going to be a lag of one year. This is for the reason that the housing market is a sticky market and houses are usually not sold in a few days. More often it takes a few months or more to get a house sold. At the moment this is even longer in the case of the Netherlands, according to an article on the website woningmarktcijfers.nl (2012) the time it took to sell a house in 2011 was on average 23 months, which shows that it is not the most flexible market at the moment. Furthermore, when someone is selling their house, the prices they ask is not adjusted daily. Therefore it could take a while before they actually ask a lower price for their house when the value of the house has been reduced.
IV. Data
The countries that are chosen for my research are chosen for the reason that these are
European countries that are a member of the OECD. These countries have a lot in common with the Netherlands, although they are not the same country. This gives a good ground to use the results that follow from this research for the case of the Netherlands. The information about the different
policies of countries regarding the mortgage interest deduction are taken from different researches on the countries. This resulted in the following information.
In the United Kingdom the mortgage interest deduction has been eliminated step-wise. After its introduction in 1969 the policy has been limited several times. The first time was in 1974, when a ceiling of 25000 pounds was put in place. This remained 25000 until 1983 when it was increased to 30000 pound. However, this “ was not enough to account for general price inflation and much too little to account for house price inflation.” (Adam, Browne and Heady 2010, 20) This ceiling was not increased any more. From 1991 onwards the maximum tax rate at which the home owners could claim relief of their paid interest was steadily reduced. From 25% in 1991, 20% in 1994, 15% in 1995 to 10% in 1998. In 2000 the policy was abolished completely.
In France the mortgage interest deduction has been abolished and has been introduced again. In 1997 the deduction was abolished for new houses, while in 1998 it was also abolished for
14 For Spain the mortgage interest deduction for second homes has been eliminated in 1990. In 1998 the deduction has changed from a limit that depended on income to a deduction that is fixed for everyone. This resulted in the fact that the deduction became less profitable for high income households and more profitable for the lower incomes. In 2011 the deduction has been eliminated for households with incomes that are higher than 24000 Euro. (see Sanchez Martinez 2005; Vacher et al. 2011)
In 1987 Germany abolished the mortgage interest deduction, as well as the taxation of imputed rent. However, it extended other deductions for the owner-occupied dwelling. In 1996 the home was taken out of the tax on income. In 1996 on the other hand a new subsidy was introduced, the “Eigenheimzulage”. This is a subsidy for first time house buyers, which lasts for eight years. This subsidy can be seen as a form of mortgage interest deduction, however it is not exactly the same. This subsidy has been eliminated again in 2006 (see Scanlon and Whitehead 2004; Commissie Sociaal-Economische Deskundigen (CSED) 2010)
In Finland the mortgage interest deduction has been in place for quite a while already. In 1992 the deduction rate was still progressive, with an average deduction rate of 51.7% and the ceiling of the interest paid that could be deducted was 75%. In the following year, 1993, the ceiling was gone, so all the interest could be deducted while the deduction rate became a flat rate of 25%. This was in fact therefore a limitation of the policy. In 1996 the flat rate was increased to 28% and in 2000 to 29%, which was a relaxation of the policy. (Sarima, 2009)
In Italy the mortgage interest deduction has been subject to some changes already. Before 1993 the tax rate at which the interest could be deducted was the marginal tax rate at which the tax payer has to pay his/her income taxes. The limit that could be deducted was 3500 Euro for each taxpayer, so not per mortgage but per taxpayer. In 1993 this is changed to the fact that the maximum that can be deducted is 3500 Euro per mortgage, and the tax rate at which this can be deducted is made a flat rate for everyone of 27%. In 1995 this flat rate is reduced to 22% and in 1998 this is reduced again to 19%. (Jappelli and Pistaferri 2007)
Before 2000 interest was fully deductible in Greece. In 2000 this is limited, now interest is only deductible if the house is not larger than 120 square metres. From 2003 this became a tax credit equal to 20% of the interest on the principal home and with a home value of less than 200 000 Euro.(Hilbers et al. 2008)
15 maximum amount. Since the first of May 2009, the duration that a buyer can deduct his interest is restricted to seven years each time he buys a new house (Deeter 2009)
In 1999 the mortgage interest deduction was limited in Austria “tax relief for interest payments scarcely exists any longer in Austria and the United Kingdom (it has been further limited in Austria since 1999.)” (Neuteboom 2004, 181)
In the Netherlands the mortgage interest deduction has been in place for quite a while already. For housing it has been in place unlimited until 2001. In this year the maximum amount of time that households can subtract interest payments is 30 years for a mortgage furthermore the deduction for second homes is abolished. (see Boelhouwer et al. 2004; Scanlon and Whitehead 2004)
In Denmark the mortgage tax relief has been reduced for the first time in 1987. The
maximum amount of mortgage interest that can be deducted is reduced from 70% of paid interest to 50% in this year. (van Osch 2006). After this the effects were visible on the housing market and it took some time before the government reduced the deductibility again. In 1999 the maximum was reduced to 39% (Munnik, 2010) and in 2002 it was reduced again to 33%. (Erlandsen, Lundsgaard and Hüfner 2006)
Before 1992 the highest rate at which the mortgage interest could be deducted was 40% in Norway, this changed to a flat rate of 28% in 1992. After this no changes relating to the mortgage tax relief have occurred. (see Boelhouwer et al. 2004; CSED 2010)
Sweden is a well known example where the mortgage interest deduction has been limited rigorously. Before 1985 the highest rate at which the interest paid could be deducted was 80%, which was reduced in 1985 to 50%. This remained so until the rate was lowered to 47% in 1989. A few years later, in 1991, a large reduction was made again, this time the rate was lowered to 30%. The last reduction coincided with a recession and this resulted in a a large decline in the house prices in the years following 1991. Unfortunately, good data that is comparable with the rest of the dataset could not be found, therefore Sweden is not included. (see Boelhouwer et al. 2004; CSED 2010)
16 the loan could not exceed 90% of the purchase price of the house and, with some exceptions related to job mobility, a person who was granted a subsidized loan was only allowed to sell the house five years after the purchase.” (Martins and Villanueva 2005, 1603) The same authors claim that there are two other kind of mortgage interest subsidies, however they ignore these subsidies for they are particularly small. In the last quarter of 1998 the subsidy is reformed, a ceiling was placed on the the maximum price of the house that was bought. If the house was more expensive, then the households were not eligible for the subsidy any longer.
The information from the description above is used to construct a the database with the country names and years of reduction of mortgage interest deduction. The five years before and after the reduction are included as well. A time variable is created and the years after the reduction get a value of one to five, depending on how many years that year is after the reduction. The year that the deduction has been reduced receives the value zero. In the case of the Netherlands for example, the year 2001 is now year zero, the year 2002 is now year one and this goes on until year five. For another country year zero can be another year, for Norway year zero is 1992 and year one is 1993, also continuing until year five.
For all eleven years the observed values of house prices and mortgage debt as a percentage of GDP is added. For the mortgage debt as a percentage of GDP, data from the European Mortgage Foundation (EMF) is taken for all the countries under research. Mortgage debt as percentage of GDP represents the stock of mortgage debt of households during a given year as a percentage of GDP. The house prices of the countries are taken from the database of the Bank for International Settlements (BIS).
Boelhouwer et al. (2004) do not find a direct observable relation between the change in house prices and the elimination or limitation of the mortgage interest deduction. For this reason this paper uses the trends of the house prices and mortgage debt. What is to be expected is that after the reduction of the mortgage interest deduction, the trend in house prices and mortgage debt will change, most likely ending up on a lower growth path.
17 house prices and mortgage debt as percentage of GDP results the fact that the year zero until five now have an estimate for house prices and mortgage debt as percentage of GDP. In order to observe any deviation from the growth path, the real observed values are subtracted from the estimated values for both variables. We are now left with the estimated difference mortgage debt and with the estimated difference house prices for year zero until five.
Not all reductions in the mortgage interest deduction are included in the research. For example in the UK reductions were made in years that were very close to each other, 1991, 1994 and 1995. To include all of these would mean that either no full series of five years could be included, or that a year was treated as different years after the year zero. In my view this would create a bias and would result in non comparable data. Furthermore for countries with multiple changes, only the last change is included in the model, in order to give each country the same weight. Unfortunately, due to data limitations, reductions that have been made before 1990 have to be ignored. In addition, due to a change in measurement of mortgage debt as percentage of GDP by the EMF, the years three, four and five for Austria were not comparable to the other observations. Therefore only year zero, one and two of this country are kept in the dataset. The mortgage debt as percentage of GDP after 2010 is not available from the EMF yet. For this reason year five does not exist in the case of Germany.
It is noted by the researcher that only eleven observations for mortgage debt leading with five years are available, which is a quite small sample. However, most of the research uses the mortgage debt leading with four years.
18 Table 2 Summary table of the variables used
Variable Reflects Source
∆ House prices The change in the difference between the real and
estimated house prices.
Own calculations, using the house price data from the BIS.
∆ Mortgage debt The change in the difference between the real and
estimated mortgage debt as percentage of GDP
Own calculation, using the e-mailed data from the EMF
Inflation Yearly inflation rate IMF world economic outlook database, April 2012
GDP growth Yearly GDP growth, constant prices, measured in the national currency
IMF world economic outlook database, April 2012
Unemployment The rate of unemployment as percentage of the labor force
IMF world economic outlook database, April 2012
Population growth Growth of number of persons living in the country
IMF world economic outlook database, April 2012
Long term interest rate Long term (10 year) interest rate, per cent per year
OECD, monthly monetary and financial statistics, June 2012
Financial reform How much a country is financially reformed, ranging from 0 (not reformed) to 1 (fully reformed)
A new database of financial reforms created by Abiad et al. (2008)
Female labor participation Percentage of women participation in the labor force
OECD, Database on labor force statistics, 2010
Eurozone Country part of Eurozone
(1 = yes)
European Central Bank
Western Europe Country part of Western Europe (1=yes)
United Nations,
Geographical region and composition, 2011 Southern Europe Country part of Southern
Europe (1=yes)
United Nations,
Geographical region and composition, 2011 Northern Europe Country part of Northern
Europe (1=yes)
United Nations,
Geographical region and composition, 2011 Global recession Reduction made while world
was in global recession (1=yes)
IMF (2009)
19 Table 3 Descriptive statistics
Variable Type Obs Mean Std. Dev.
Dependent
∆ House prices Interval 73 -0.0441 0.1778
Independent
∆ Mortgage debt Interval 70 0.2754 0.5679
∆ Mortgage debt 1 year lead Interval 58 0.3207 0.6192 ∆ Mortgage debt 2 year lead Interval 45 0.3655 0.6860 ∆ Mortgage debt 3 year lead Interval 33 0.4195 0.7542 ∆ Mortgage debt 4 year lead Interval 22 0.4576 0.8166 ∆ Mortgage debt 5 year lead Interval 11 0.5489 0.9907 ∆ Mortgage debt 1 year lag Interval 58 0.2361 0.4828 ∆ Mortgage debt 2 year lag Interval 46 0.2069 0.4285 ∆ Mortgage debt 3 year lag Interval 35 0.1602 0.3283 ∆ Mortgage debt 4 year lag Interval 23 0.1246 0.2348 ∆ Mortgage debt 5 year lag Interval 11 0.0877 0.1519
Population growth Interval 72 0.0033 0.0023
GDP growth Interval 73 0.0232 0.0222
Longterm interest rate Interval 73 0.0560 0.0230
Unemployment Interval 73 0.0838 0.0492
Inflation Interval 73 0.0227 0.0152
Female labor participation Interval 71 0.5723 0.1171
Financial reform Latent 60 0.8982 0.0879
Eurozone Categorical 73
Western Europe Categorical 73
Southern Europe Categorical 73 Northern Europe Categorical 73 Global recession Categorical 73 Twentyfirst Century Categorical 73
20 Diagnostic tests
The model that will be set out in the next section is a multiple linear regression. Before going to this section, the results of the diagnostic tests that were done on the data are presented here. In order for a linear regression to deliver unbiased results, the model should not suffer from heteroscedasticity and multicollinearity of the independent variables. Besides this, outliers can have a large effect on the coefficients, therefore a check for outliers is advisable as well. Normality of the residuals is often claimed to be necessary, however this is only needed for the hypothesis tests to be valid and not for an unbiased estimation of the model coefficients. Furthermore, in order for the linear regression to be the best method, the independent variables should have a linear relationship with the dependent variable.
To look whether the data suffers from heteroskedasticity the post-estimation test, White test is conducted. No heteroskedasticity was found for any of the models.
Multicollinearity was checked by creating VIF tables after the regression model is estimated. It does not seem to be a problem either, until the dummy variables for dividing the sample into different geographical regions are included. In these cases the variables financial reform, female labour participation and long term interest rate seem to have multicollinearity problems with the dummy variable for either Eurozone, Northern Europe, Western Europe and Southern Europe. An explanation for this is that when countries are considered to be more alike, then these indicators do have the incentive to move in the same direction. For example, interest rates in Southern Europe that governments have to pay on their debt are much larger than they are in Northern Europe. However it should be taken into account when drawing conclusions that the multicollinearity exist in the model when these dummy variables are included.
Inspecting the data for outliers myself, some large estimated growth rates of Spain’s
estimated mortgage debt caught my attention. The values were checked whether they were typed in correctly and if the estimation was done properly, no problems were found. Testing for outliers using studentized residuals was done as well. When the residual is larger than two, two and a halve, it is advisable to pay some special attention to the observation. Two values of both Norway and Finland were detected as being modest outliers using this method. A closer look was taken at these values, however they do not seem to be a mistake, typing error or wrong estimation. Therefore the dataset is not altered and all observations are kept.
21 the variables a clear deviation from linearity could be found. Therefore, linear regression seems to be a suitable method for this research.
Using the Shapiro–Wilk test to test for normality of the residuals, the hypothesis that the residuals are normally distributed is in most models rejected. However, as we can see in figure S1 it does not seem that too much deviation from normality exists in the model, except for one
observation, which is one of the modest outliers that were detected before. When testing what happened when this observation was removed the Shapiro–Wilk test could not any longer reject that the residuals are normally distributed. However, as said before, for unbiased estimates of regression coefficients it is not needed that the residuals are normally distributed, therefore the observation that led to the deviations from normality is kept in the sample.
V. Model
What is to be researched, as mentioned before is the lag between the mortgage debt and the house prices. For the reason that the reduction in mortgage interest deduction causes both variables to decline, causality between mortgage debt and house prices is not examined. As said in the previous section, multiple linear regression analysis is used for this.
The first set of models (A1-A11) is very simplistic and just looks at the fact whether the change in house prices can be explained by the change in mortgage debt. The basic model therefore will be:
HOUSE_PRICES = β0 +β1MORTGAGE_DEBTt + ε (1)
Where HOUSE_PRICES is the difference between the estimated house price in a given year minus the actual measured house price. Mortgage debt is the estimated mortgage debt as percentage of GDP minus the actual measured mortgage debt as percentage of GDP. The subscript t changes in this estimation going from -5, a five year lag, to +5, a 5 five year lead. Here ε is the error term.
When we have a look at the results of the basic model without any lead or lag for mortgage debt, model A1, (table S1) it is seen that the change in mortgage debt does not explain the variation in house prices very well when no lags and leads are incorporated, with even a negative adjusted R². Besides this the model is not significant either.
22 to a lead of five years. Besides this lead a lag will be investigated as well, for the reason that it might be that my hypothesis is wrong and the house prices lead instead lag over mortgage debt after the reduction of mortgage interest deduction.. When the lags and leads are incorporated, model A2-A11, the model with a lead of four years for the mortgage debt gives a significant model (table 4; table S1; table S2). The coefficient is positive, meaning that an increase in mortgage debt results in higher housing prices. This positive coefficient is what is expected from the theory. This result means that when the mortgage interest deduction is eliminated, the mortgage debt goes down first, followed by the house prices four years later. The rest of the models are not significant.
Table 4 Basic model results of relation between mortgage debt as percentage of GDP and house prices
Variables Model A5
Intercept -0.0187
(0.0213) ∆ Mortgage debt 4 year
lead 0.0528** (0.0232) Observations 20 Adj. R² 0.1809 P-value model 0.0351
Note: Standard errors in parentheses. *,** and *** denotes significance at the 0.1, 0.05 and 0.01 level respectively.
Now the macro economic and social variables inflation, GDP growth, population growth, unemployment and long term interest rate are added as a control variable. This results in the next regression equation (2) and the next set of broadened models (B1-B11).
HOUSE_PRICES = β0 +β1MORTGAGE_DEBTt + β3INFLATION + β4GDP +
β5POPULATION +β6 UNEMPLOYMENT + ε (2)
23 which it should have according to Tsatsaronis and Zhu (2004) who state in their paper that inflation should have a positive influence on house prices. GDP growth, population growth and
unemployment have the wrong sign and are not significant. Population growth, is expected to have a positive influence on house prices, however in this model it results in a large negative number. As will be shown in the next few models it will be significant there. This was not expected. A possible explanation for this might be because Sá (2011,1 abstract) believes that “immigration has a negative effect on house prices”. The sample of countries that is chosen has known substantial immigration flows, especially the countries that are in Northern Europe and Western Europe. As we see in the extended model C4 for Northern Europe the coefficient of population growth is very large and significant. (table S5). Leith (2012, no page) has an explanation for the fact that unemployment has the unexpected sign “ is because when levels of debt and asset values are rising, households feel richer (the ‘wealth effect’), spurring consumer confidence, spending and employment growth. However, once asset prices stop rising (or fall in value), the process of debt feeding asset prices feeding confidence and consumer spending can shift into reverse, causing rising unemployment.”
This is only a suggestion and this is not verified in this research. It is not clear why GDP growth
24 Table 5 Broadened model results of
relation between mortgage debt as percentage of GDP and house prices
Variables Model B5
Intercept 0.1299*
(0.0687)
∆ Mortgage debt 4 year lead 0.0987*** (0.0288) Population growth -9.5356 (10.1952) GDP growth -0.4175 (0.8468) Longterm interest rate -2.7242**
(1.0718) Unemployment 0.0954 (0.5805) Inflation 1.4433 (1.2715) Observations 20 Adj. R² 0.4657 P-value model 0.0194
Note: Standard errors in parentheses. *,** and *** denotes significance at the 0.1, 0.05 and 0.01 level respectively.
Arrived at the extended model (C1-C7), it has to be noted that t does not vary any longer and is now set to be constant. Mortgage debt is therefore now leading by four years over house prices for the reason that this the other leads and lags were not significant in the past models.
HOUSE_PRICES = β0 +β1MORTGAGE_DEBT4 + β3INFLATION + β4GDP + β5POPULATION +β6 UNEMPLOYMENT +β7FEMALE_LABOUR_PARTICIPATION
+β8FINANCIAL_REFORM + ε (3)
25 the fact that households are able to pay more for a house. Financial reform is included for several reasons which all come down to the better or cheaper access of households to the financial markets. Demand for housing increases, the result should be higher prices. We see that this does improve the model, the P-value of the model goes down a bit, and stays significant at the 0.05 level (table S5). However we see that the two newly added variables are not significant, neither are GDP growth and unemployment. Female labour participation has the expected sign as well as GDP growth. Financial reform and unemployment have the non expected sign. Mortgage debt is still significant at the 0.05 level and has the expected sign.
A reason why some control variables have the wrong sign and are insignificant can be due to the fact that the research aggregates different kind of countries at the time, the Southern, Northern and Western European countries. The three models C2-C4 separating the sample into the three different parts of Europe by using a dummy variable for each of them. Northern Europe and Southern Europe are significant at the 0.01 and 0.05 level respectively. The control variables that are left do not all behave as they are supposed to behave. Unemployment, population growth and financial reform have the wrong sign. For population growth and unemployment a possible explanation has been given already. It is not clear why financial reform has the wrong sign.
With the division between Southern, Western and Northern Europe the division was made on geographical grounds only. However, a more sensible division between countries would be whether the country is a member of the Eurozone or not. Being in the Eurozone can have a large impact on a country. More than just being in the same area as other countries. The countries in the Eurozone have been cooperation for a long time already, starting with the European Coal and Steel
Community (ECSC) which later became the European Economic Community (EEC) and ultimately the European Union. The countries in the Eurozone have intensified these cooperation and for this reason they are more interconnected than countries who have more autonomy left, like Norway and the UK. Therefore the dummy variable Eurozone is added, so whether a country was in the
Eurozone when it reduced the mortgage interest deduction, or that it was going to enter the Eurozone.
26 significant. Inflation has flipped sign and is now significant with the unexpected sign, however the coefficient is very small. The reason why this sign has flipped might be because of the
multicollinearity that occurred while adding the dummy variables. Still however, it is expected that being in the the Eurozone has a positive and significant effect on housing prices.
Table 6 Extended model results of relation between mortgage debt as percentage of GDP and house prices
Variables Model C1 Model C5
Intercept 0.0350
(0.3093)
-0.4877 (0.2829) ∆ Mortgage debt 4 year
lead 0.1025** (0.0327) 0.1024*** (0.0240) Population growth -26.9579** (11.3114) -76.2124*** (17.9223) GDP growth 0.2633 (0.9651) 1.0789 (0.7556) Longterm interest rate -3.5137**
(1.3778) 0.0513 (1.5328) Unemployment 0.7003 (0.7714 -0.7655 (0.7381) Inflation 2.6362* (1.4201) 0.4897 (1.2518) Female labour participation 0.3930 (0.2753) 1.2276*** (0.3370) Financial reform -0.1110 (0.2544) -0.0652 (0.1873) Eurozone 0.2366** (0.0765) Observations 19 19 Adj. R² 0.6066 0.7882 P-value model 0.0153 0.0020
Note: Standard errors in parentheses. *,** and *** denotes significance at the 0.1, 0.05 and 0.01 level respectively.
27 growth and long-term interest rate. All the other variables are insignificant. Besides this the signs of the variables do all, except for unemployment, financial reform and population growth, have the expected sign. (table S6). Due to the low significance of the dummy variable it seems highly unlikely that it does make any difference whether the reduction of the mortgage interest deduction in the recent past than in the more distant past.
GDP growth of the country itself has been included in the analyses already, however economies are vulnerable for events that happen in the world around them as well. For example, if the world is in a recession, then this can have a large influence on the country. Therefore the dummy variable Global recession is included to see whether this has a significant effect on the house prices in a country. The dummy variable in model C6 is not significant. (table S6) The control variables behave the same as with the variable that controls for time, however only mortgage debt, population growth and long-term interest are significant.
VI Implications of the research
Several implications of this research, can be used by the Dutch policy makers. What they have to note first is that the housing market is inefficient. This is a result of the subsidies and taxes that are in place in this market. Removing them would result in a higher total welfare for the country. Secondly, an important reason for still having the mortgage interest deduction, is because it would promote home-ownership. It is shown that, in spite of the generous deduction, the Netherlands have a very low home-ownership rate. The policy is just not effective for what it is meant for. As a third implication, the model shows that after the limitation of the mortgage interest deduction, the decline in mortgage debt leads over the decline in house prices by four years. This results support the hypothesis: If the mortgage interest deduction is reduced, then the decline in house prices lag behind the decline in mortgage debt as a percentage of GDP. The risk therefore that many households will end up underwater due to the reduction in mortgage interest deduction is
28 The variable Eurozone however is more important for it is something which the Netherlands face nowadays and the linkage is strong. The countries really work together and start having more and more in common. This dummy variable is very significant, with mortgage debt leading by four years being highly significant as well. However, care is needed in this case, for the data suffered from multicollinearity with these dummy variables. Furthermore, the government does not have to wait for the global recession to be over. This is for the reason that the dummy variable global recession is far from significant, meaning that this does not make any difference. Finally, when controlling for time, no other implications can be made. Meaning that reductions in the policy made a long thirty years ago have the same effect as policies made just a few years ago.
What has to be mentioned is that in this research no distinction is made between extensive and relatively restricted possibilities to deduct the mortgage interest. Neither for large or small
29 the elimination of extensive and restricted possibilities for mortgage interest deduction would not alter the results. However, this non expected difference in the ratio has not been examined in this research and therefore it might be an option for future research to examine this.
VI. Conclusion
This research has given several reasons why the mortgage interest deduction should be reduced. The gains that can be achieved by this are large in the case of the Netherlands. This is due to the fact that the subsidy, mortgage interest deduction, is very large in the Netherlands. A large subsidy results in a large deadweight loss. Besides this, the mortgage debt of the Dutch households is very large, measured on a micro-level as LTV or on the macro-level as percentage of GDP. Lower debts result in a more stable environment in the country. It is recognized that during the transition from the old to the new equilibrium difficulties might arise. However, the results show us that the decline in house prices lags four years behind the decline in mortgage debt. The hypothesis is supported by the model. For this reason only a very small probability to end up under water is present. It is not likely that the reduction of the mortgage interest deduction will result in large complications.
A limitation of the research is that many control variables do not have the expected sign. This might be due to the fact that the sample is quite small, with as result that these variables in this small sample behave differently than expected. A suggestion for future research therefore is to enlarge the sample of countries and therefore enlarge the sample with more observations.
Furthermore, the collinearity of the variables when dummy variables for Eurozone, Western Europe, Northern Europe and Southern Europe are included is a limitation.
As this research already showed however, many different opinions about the mortgage interest deduction are present. Researchers do not agree on each other whether the policy should be eliminated, and if they agree that it should, they do not agree on how it should be eliminated.
Politicians in the Netherlands do, as shown, not agree on this topic either. Many different opinions about the policy are present, resulting in the fact that I do not expect that something substantial in the near future is going to happen with relation to the mortgage interest deduction in the
30 References
Abiad, Abdul, Enrica Detragiache, and Thierry Tressel. 2010. A New Database of Financial Reforms. IMF working paper no. 08/266, International Monetary Fund, Washington, D.C.
Adam, Stuart, James Browne, and Christopher Heady. 2010. Taxation in the UK. In Dimensions of Tax Design: the Mirrlees Review, edited by James Mirrlees, Stuart Adam, Timothy Besley, Richard Blundell, Stephen Bond, Robert Chote, Malcolm Gammie, Paul Johnson, Gareth Myles and James Poterba, Oxford: Oxford University Press
Agell, Jonas, Lennart Berg, and Per-Anders Edin. 1995. The Swedish Boom-to-Bust Cycle: Tax Reform, Consumption and Asset Structure. Swedish Economic Policy Review. 2 (Fall): 271-314. Atterhog, Mikael. 2005. Importance of Government Policies for Home Ownership Rates. Working Paper no. 54. Swedish Royal Institute of Technology, Stockholm
Auerbach, Allen J., and James R. Hines. 2001. Taxation and economic efficiency. Working paper no. 8181. National Bureau of Economic Research, Cambridge
Ball, Michael. 2002. RICS European housing review 2002. London: Royal Institution of Chartered Surveyors
Bhutta, Neil, Jane Dokko, and Hui Shan. 2010. The Depth of Negative Equity and Mortgage Default Decisions. Finance and Economics Discussion Series no. 35. The Federal Reserve Board, Washington, D.C.
Boelhouwer, Peter J., Marietta Haffner, Peter Neuteboom, and Paul de Vries. 2004 House Prices and Income Tax in the Netherlands: An International Perspective. Housing Studies 19 (May): 415–32.
31 Bracke, Philippe. 2011. How long do housing cycles last? A duration analysis for 19 OECD
countries. IMF working paper no. 11/231, International Monetary Fund, Washington, D.C.
Brosens, Teunis. 2009. Hypotheek risico's voor huizenbezitters. Economisch Statistische Berichten 4570 (October): 614-17
Campbell, John Y., and Joao F. Cocco. 2011. A model of mortgage default. Working paper no. 17516. National Bureau of Economic Research, Cambridge
Capozza, Dennis R., Richard K. Green, and Patric H. Hendershott. 1996. Taxes, mortgage borrowing and residential land prices. In Economic effects of fundamental tax reform, edited by Henry J. Aaron and William G. Gale. Brookings Institution Press, Washington, D.C.
Capozza, Dennis R., Richard K. Green, and Patric H. Hendershott. 1997. Taxes and house prices. Working paper series no. 97-16. Ohio State University, Ohio
Commissie Sociaal-Economische Deskundigen. 2010. Naar een integrale hervorming van de woningmarkt. Den Haag: Sociaal Economische Raad
“Mortgage Interest Relief (TRS) changes for May 2009” Karl Deeter, accessed April 28, 2012, www.mortgagebrokers.ie/blog/index.php/2009/04/29/mortgage-interest-relief-trs-changes-for-may-2009/
Deng, Yongheng, John M. Quigley, and Robert van Order. 2000. Mortgage terminations, heterogeneity and the exercise of mortgage options. Econometrica 68 (March): 275-307
Dunsky, Robert M., and James R. Follain. 2000. Tax-induced portfolio reshuffling: the case of mortgage interest deduction. Real Estate Economics 28 (4): 683-718
32 Elul, Ronel, Nicholas S. Souleles, Souphala Chomsisengphet, Dennis Glennon, and Robert Hunt 2010. What “triggers” mortgage default. Working paper no 10-13. Federal Reserve Bank of Philadelphia, Philadelphia
Englund, Peter, Patric H. Hendershott, and Bengt Turner. 1995. The tax reform and the housing market. Swedish Economic Policy Review 2 (Fall): 319-56
Erlandsen, Espen, Jens Lundsgaard, and Felix Huefner. 2006. The Danish housing market: Less subsidy and more flexibility. Economics department working paper no 513. Organisation for Economic Co-operation and Development, Paris
Ewijk, Casper, Bas Jacobs, and Ruud de Mooij. 2006. Doelmatigheidswinst van minder hypotheekrenteaftrek. Economisch Statistische Berichten 4489 (June): 292-95
Ewijk, Casper, Bas Jacobs, and Ruud de Mooij. 2007. Welfare effects of fiscal subsidies on home ownership in the Netherlands. De Economist 155 (3): 323-36
Follain, James R. 1990. Mortgage choice. Areuea journal 18 (2): 125-44
Follain, James R., and Lisa Sturman Melamed. 1998. The false messiah of tax policy: what
elimination of the home mortgage interest deduction promises and a careful look at what it delivers. Journal of Housing Research 9(2): 179-99
de Haan, Jan, André van den Berg. 2011. “Decomposing changes in aggregate loan-to-value ratios”. Paper prepared for the economic measurement group workshop, November 30 – December 2, 2011, Sydney
Harberger, Arnold C. 1954. Monopoly and resource allocation. The American economic review 44 (2): 77-87
33 Harberger, Arnold C. 1964b. Taxation, resource allocation, and welfare. In The Role of Direct and Indirect Taxes in the Federal Reserve System. Princeton, Princeton University Press
Harberger, Arnold C. 1971. Three basic postulates for applied welfare economics: an interpretive essay. Journal of economic literature 9(3): 785-97
Hess, Andres, and Arne Holzhausen. 2008. The structure of European mortgage markets. Special focus: Economy & Markets 01/2008
Hilbers, Paul, Alexander W. Hoffmaister, Angana Banerji, and Haiyan Shi. 2008. House price developments in Europe: a comparison. Working paper 08/211, International Monetary Fund, Washington, D.C.
Hoteling, Harold. 1938. The general welfare in relation to problems of taxation and of railway and utility rates. Econometrica 6(3): 242-69
International Bureau of Fiscal Documentation. 2003. European Tax Handbook 2003. Amsterdam: International Bureau of Fiscal Documentation
International Monetary Fund. 2008. Chapter 3. The changing housing cycle and the implications for monetary policy. In World Economic Outlook 2008. Washington D.C.: International Monetary Fund
International Monetary Fund. 2009. Global economic slump challenges policies. World Economic outlook update, January 28, 2009. Washington D.C: International Monetary Fund
International Monetary Fund. 2010. Kingdom of the Netherlands – Netherlands staff report for the 2009 article IV consultation. Washington D.C: International Monetary Fund
34 Longstaff, Francis A., 2010. The subprime credit crisis and contagion in financial markets. Journal of Financial Economics 97 (3): 436-50
Madden, Janice F. 1980. Urban land use and the growth in two-earner households. The American Economic Review 70 (May): 191-97
Martins, Nuno C., and Ernesto Villanueva. 2005. The impact of mortgage interest-rate subsidies on household borrowing. Journal of Public Economics 90: 1601-23
Munnik, Thomas. “Beperking van de hypotheekrenteaftrek. Wat kan worden geleerd van het buitenland?” Thesis, Erasmus university Rotterdam, 2010
Neuteboom, Peter. 2004. A comparative analysis of the net cost of a mortgage for homeowners in Europe. Journal of housing and the built environment 19 (2): 169-89
“Unemployment and house prices” Leith van Onselen, accessed July 4, 2012, http://www.macrobusiness.com.au/2011/05/unemployment-and-house-prices
Organisation for Economic Co-operation and Development. 2004. Housing policies, excerpt. In OECD economic survey of the Netherlands 2004. Paris: Organisation for Economic Co-operation and Development
van Osch, Brenda. 2006. “Eigen huis: Morrelen aan de aftrek” Elsevier, July 27, 2006. Accessed May 11, 2012.
http://www.elsevier.nl/web/1085132/Dossiers/Huizenmarkt-1/De-hypotheekrenteaftrek/Eigen-huis-Morrelen-aan-de-aftrek.htm
Sanchez Martinez, Maria T. 2005. Distributive Aspects of Public Expenditure on Housing in Spain. International journal of housing policy 5(3): 237-53
35 Sørensen, Peter. B. 2010 Calculating the Deadweight Loss from Taxation in a Small Open Economy. Report to the Expert Group on Public Economics (ESO) 2010:4 Supplement. Stockholm:
Regeringskansliet, Finansdepartementet.
Tsatsaronis, Kostas, and Haibin Zhu.2004. What drives housing price dynamics: cross-country evidence. BIS Quarterly Review March: 65-78
Vacher, Jerome, Keiko Honjo, Florence Jaumotte, Rafael Romeu, Alessandro Giustiniani, and Edouard Vidon. 2011. Spain: Selected issues. IMF country report 11/216. Washington D.C: International Monetary Fund
Woningmarktcijfers.nl “Verkooptijd koopwoning opgelopen tot 23 maanden.”Accessed May 13, 2012.
http://www.woningmarktcijfers.nl/nieuws/verkooptijd_koopwoning_opgelopen_tot_23_maanden.ht m
van Wijnbergen, Sweder, Herman Wijffels, Casper de Vries, Leo Stevens, Elmer Sterken, Dirk Schoenmaker, Hugo Priemus, Rick van der Ploeg, Kees Koedijk, Bas Jacobs, Kees Goudswaard, Harry Garretsen, Jean Frijns, Piet Eichholtz, Johan Conijn, Dirk Brounen, Lans Bovenberg, Arnoud Boot, Peter Boelhouwer, Harald Benink, Roel Beetsma and Eric Bartelsman. 2012. “Naar een duurzame financiering van de woningmarkt.” Accessed June 2, 2012,
36 Appendices
Table S1 Basic model results of relation between mortgage debt as percentage of GDP and house prices Variables Model A1 Model A2 Model A3 Model A4 Model A5 Model A6
Intercept -0.0306 (0.0181)* -0.0291* (0.0168) -0.0262 (0.0169) -0.0221 (0.0181) -0.0187 (0.0213) -0.00001 (0.0307) ∆ Mortgage debt 0.0072 (0.0289) ∆ Mortgage debt 1 year
lead
0.0175 (0.0243)
∆ Mortgage debt 2 year lead
0.0313 (0.0219)
∆ Mortgage debt 3 year lead
0.0327 (0.0213)
∆ Mortgage debt 4 year lead
0.0528** (0.0232)
∆ Mortgage debt 5 year lead 0.0030 (0.0283) Observations 69 56 43 31 20 9 Adj. R² -0.0140 -0.0088 0.0241 0.0436 0.1809 -0.1411 P-value model 0.8044 0.4728 0.1611 0.1347 0.0351 0.9192
37 Table S2 Basic model results of relation between mortgage debt as percentage of GDP and house prices Variables Model A7 Model A8 Model A9 Model A10 Model A11
Intercept -0.0337 (0.0247) -0.0451 (0.0305) -0.0640 (0.0434) -0.0553 (0.0614) -0.0594 (0.1104) ∆ Mortgage debt 1 year lag 0.0154 (0.0462) ∆ Mortgage debt 2 year lag 0.0122 (0.0647) ∆ Mortgage debt 3 year lag 0.0478 (0.1203) ∆ Mortgage debt 4 year lag 0.0206 (0.2348) ∆ Mortgage debt 5 year lag 0.1411 (0.6518) Observations 58 46 35 23 11 Adj. R² -0.0158 -0.0219 -0.0254 -0.0472 -0.1054 P-value model 0.7396 0.8512 0.6936 0.9308 0.8334
38 Table S3 Broadened model results of relation between mortgage debt as percentage of GDP and house prices
Variables Model B1 Model B2 Model B3 Model B4 Model B5 Model B6
Intercept -0.302 (0.0548) -0.0448 (0.0539) -0.0519 (0.0633) -0.0007 (0.0727) 0.1299* (0.0687) 0.2293 (0.1187) ∆ Mortgage debt -0.0285 (0.0390) ∆ Mortgage debt 1 year lead -0.0087 (0.0360) ∆ Mortgage debt 2 year lead 0.0356 (0.0340) ∆ Mortgage debt 3 year lead 0.0461 (0.0348) ∆ Mortgage debt 4 year lead 0.0987*** (0.0288) ∆ Mortgage debt 5 year lead 0.0394 (0.0413) Population growth -3.5539 (7.9472) -3.8335 (7.3655) -1.1804 (9.0501) -2.714 (10.7957) -9.5356 (10.1952) -26.3110 (18.5693) GDP growth 0.2204 (0.7491) 0.4710 (0.7209) 1.2961 (1.0220) 0.4383 (1.0208) -0.4175 (0.8468) -0.4124 (1.4345) Longterm interest rate -1.4915
(1.1967) -0.9422 (1.0779) -1.3798 (1.0933) -1.4686 (1.1870) -2.7242** (1.0718) -1.7862 (1.2785) Unemployment 0.9456* (0.5635) 0.7185 (0.5171) 0.7276 (0.5490) 0.6870 (0.6149 0.0954 (0.5805) -0.6010 (1.0412) Inflation 1.0657 (1.3597) 0.8379 (1.2372) 0.5109 (1.2675) 0.0316 (1.3699) 1.4433 (1.2715) 1.3977 (1.6330) Observations 68 56 43 31 20 9 Adj. R² -0.0175 -0.0274 0.0159 0.0288 0.4657 0.1767 P-value model 0.5677 0.6082 0.3741 0.3658 0.0194 0.4991
39 Table S4 Broadened model results of relation between mortgage debt as percentage of GDP and house prices
Variables Model B7 Model B8 Model B9 Model B10 Model B11
Intercept -0.0484 (0.0745) -0.0374 (0.0908) 0.0046 (0.1375) 0.0054 (0.2245) 0.1618 (0.5385) ∆ Mortgage debt 1 year lag -0.0588 (0.0641) ∆ Mortgage debt 2 year lag 0.0665 (0.0905) ∆ Mortgage debt 3 year lag 0.0114 (0.1750) ∆ Mortgage debt 4 year lag 0.1190 (0.3601) ∆ Mortgage debt 5 year lag 0.2771 (1.0656) Population growth 2.1573 (12.2355) 5.3488 (15.5296) 16.3419 (23.1615) 21.8731 (34.9098) 16.5058 (89.0486) GDP growth 0.7895 (1.2475) 1.7049 (1.5013) 1.3477 (2.0776) 2.9503 (4.5734) 1.5644 (12.3355) Longterm interest rate -3.0388 (1.8728) -5.1970** (2.4882) -8.3636* (4.1867) -11.6600 (7.8238) -17.6970 (22.4764) Unemployment 1.9824** (0.8321) 2.5012** (1.0691) 2.9022* (1.5437) 2.9493 (2.2272) 5.2345 (5.2965) Inflation 0.5734 (2.1109) 1.1539 (2.6668) 2.4221 (4.3627) 6.2135 (9.7811) 6.3908 (28.5642) Observations 57 46 35 23 11 Adj. R² 0.0163 0.0226 -0.0381 -0.1827 -0.8375 P-value model 0.3454 0.3402 0.5839 0.8458 0.9403
40 Table S5 Extended model results of relation between mortgage debt as percentage of GDP and house prices
Variables Model C1 Model C2 Model C3 Model
C4 Model C5 Intercept 0.0350 (0.3093) -0.2708 (0.5960) -1.3255** (0.5142) -0.4798* (0.2507) -0.4877 (0.2829) ∆ Mortgage debt 4 year
lead 0.1025** (0.0327) 0.0872* (0.0421) 0.0730** (0.0265) 0.1191** * (0.0224) 0.1024*** (0.0240) Population growth -26.9579** (11.3114) -6.5538 (35.5461) -12.5841 (9.7693) -87.2136* ** (18.1764 ) -76.2124*** (17.9223) GDP growth 0.2633 (0.9651) 0.2982 (0.9987) 0.2536 (0.7236) 0.1080 (0.6477) 1.0789 (0.7556) Longterm interest rate -3.5137**
(1.3778) -3.2426* (1.4916) -0.9859 (1.3395) -0.7415 (1.1954) 0.0513 (1.5328) Unemployment 0.7003 (0.7714 0.8457 (0.8321) 1.4398** (0.6299) 1.2451** (0.5378) -0.7655 (0.7381) Inflation 2.6362* (1.4201) 2.9344* (1.5470) 1.9319 (1.0910) 0.3578 (1.1379) 0.4897 (1.2518) Female labour participation 0.3930 (0.2753) 0.1401 (0.5041) 0.9190*** (0.2722) 2.2165** * (0.5329) 1.2276*** (0.3370) Financial reform -0.1110 (0.2544) 0.2807 (0.6960) 0.7141* (0.3374) -0.4283* (0.1913) -0.0652 (0.1873) Western Europe -0.0826 (0.1359) Southern Europe 0.2178** (0.0735) Northern Europe -0.3330** * (0.0913) Eurozone 0.2366** (0.0765) Observations 19 19 19 19 19 Adj. R² 0.6066 0.5802 0.7789 0.8236 0.7882 P-value model 0.0153 0.0307 0.0024 0.0009 0.0020
41 Table S6 Extended model results of relation between mortgage debt
as percentage of GDP and house prices
Variables Model C1 Model C6 Model C7
Intercept 0.0350 (0.3093) 0.0161 (0.3232) -0.0783 (0.3431) ∆ Mortgage debt 4 year
lead 0.1025** (0.0327) 0.1025** (0.0341) 0.1088** (0.0341) Population growth -26.9579** (11.3114) -28.6951** (12.5024) -28.4920** (11.6458) GDP growth 0.2633 (0.9651) 0.4111 (1.0664) 0.7691 (1.1567) Longterm interest rate -3.5137**
(1.3778) -3.6658** (1.4825) -3.9815** (1.5107) Unemployment 0.7003 (0.7714 0.7442 (0.8119) 0.8311 (0.7999) Inflation 2.6362* (1.4201) 2.6739 (1.4850) 3.5559* (1.8237) Female labour participation 0.3930 (0.2753) 0.4093 (0.2899) 0.4873 (0.3022) Financial reform -0.1110 (0.2544) -0.1163 (0.2658) -0.0471 (0.2699) Global recession 0.0161 (0.0381) Twenty-first Century -0.0380 (0.0461) Observations 19 19 19 Adj. R² 0.6066 0.5714 0.5936 P-value model 0.0153 0.0331 0.0271
