Homeownership’s effect on wealth accumulation in the Netherlands.

University of Amsterdam, Amsterdam Business School

Master in International Finance

Master Thesis

Homeownership’s Effect on Wealth Accumulation in

the Netherlands

September 2013

Abstract

Few empirical studies have been conducted that actually test the adage that becoming a

homeowner is a surefire path to wealth. Using the DNB Household Survey this paper tests the

independent wealth effect of homeownership on accumulated wealth with three models which

control for typical demographics. The first two models which test data sets from before and after

the crisis find that current homeownership is positively and significantly related to current wealth

levels despite the economic crisis. The third model tests a group of renters over time and

compares the wealth of those who chose to become homeowners in the sample period to those

who remained renters. This model finds that homeownership is not significantly related to wealth

rather the tendency to save is. The tendency to save is a unique variable controlled for in these

models that could prove reverse causation answering the question: “Are homeowners savers to

begin with?” The Netherlands provides an interesting backdrop for this test due to its government

intervention in the housing market artificially inflating the market, historic political gridlock on

the generous housing tax rebate, and feeble attempts to prop up a depressed real estate market

literally and figuratively teetering on stilts.

Elisabeth Schön

1 Table of Contents

1 Introduction ... 3

1.1 The state of the Dutch housing market ... 5

2 Literature Review ... 9

2.1 Homeownership’s effect on society ... 10

2.2 Homeownership’s effect on the economy ... 10

2.3 Homeownership as a portion of a household’s investment portfolio ... 11

2.4 Homeownership versus stocks and bonds ... 11

2.5 Homeownership versus renting costs ... 12

2.6 Homeownership as a hedge against inflation ... 12

2.7 Homeownerhship’s effects on wealth ... 12

3 Methodology and data ... 14

3.1 2007 and 2012 models ... 14

3.2 Di et al.’s model for homeownership in the long-run ... 17

3.3 Data ... 19

3.4 Descriptive statistics ... 20

3.5 Graphs based on descriptive statistics for 2007 and 2012 samples ... 22

4 Empirical Results ... 30

4.1 Net wealth 2007 and 2012 models ... 30

4.2 Non-housing wealth 2007 and 2012 models ... 31

5 Di et al. model ... 33

5.1 Descriptive statistics ... 33

5.2 Di et al. model empirical results ... 36

6 Conclusion ... 37

1 Introduction

It is a common belief that homeownership paves a surefire path to greater wealth. The statement, “Why throw your money away on rent?” is uttered by the man on the street and taken at face value. A proper checklist for life usually runs in the following order: go to college, so you can get a good job, so you can buy a house and raise a family in it. Any data set that encompasses household assets and liabilities does show that on average homeowners do indeed have more net wealth than renters and that most of that wealth does indeed come in the form of home equity. Governments have pushed policies to promote homeownership based on these statistics with faith that this would lead their nations to prosperity. But as the subprime crisis has revealed homeownership in and of itself may not be what makes a household wealthy.

This question is gravely relevant in light of the recent recession initially caused by the most systemic housing market boom and bust in history. More households own homes now than ever before.

Governments promoted homeownership through a variety of tax benefits and subsidies. Policymakers pushed mortgage lenders to provide products to extend homeownership to all income classes, a beautiful ideal to lift the veil of poverty with rows of property. People are now questioning the infallible conviction that homeownership is the answer to society’s poor or even a guaranteed path to wealth for those who can afford it.

Policymakers believed they could stimulate virtuous behaviors by stimulating homeownership. “But mere ownership doesn't magically make people financially wise, capable, or virtuous. Instead, in a country with property and other basic rights, wise and virtuous behavior makes it possible for people to accumulate enough capital and credibility to be able to get a home loan.” Richards (2013) addresses a key question about homeownership’s relationship to wealth. Are homeowners savers and investors to begin with? Although homeownership is attributed to many positives, Richards calls this a classic case of confusing correlation and causation.

This chicken-or-egg question may explain why there are actually very few empirical studies out there that test homeownership’s sole effect on wealth. There are numerous studies that correlate wealth with

everything from civic pride and duty to GDP growth but surprisingly few that question society’s assumption that owning a home will secure wealth and prosperity to private households. I would like to add to the small body of empirical research on the actual effects of choosing to own a home and the duration of that ownership on household wealth particularly in a housing market where government interventions play a major role and housing has become an economic force to be reckoned with.

Existing literature reveals several ways that homeownership contributes to wealth. It is looked at as a form of forced savings. Paying off a mortgage every month which directly translates into home equity

eventually provides households with a substantial asset. The costs that renters save themselves by not becoming homeowners are notorious for getting lost in a cloud of consumption. A house is also a leveraged asset so any increase in value amplifies capital gains. Mortgages are also a source of cheap funding so homeowners have access to lower-cost financing in the form of second mortgages. Home values also tend to keep pace ahead of rising prices so it is considered an effective inflation hedge. Mortgage interest is often tax-deductible so homeowners with larger incomes benefit from a tax credit. Lastly, a household can profit if they happen to sell their home at the right time in the market cycle but obviously a household risks losing wealth if they are forced to sell in a downturn. This is a concern for many as this stagnant housing market immobilizes households.

In this paper I will analyze the effect of homeownership on wealth, specifically tenure choice and duration of current ownership on net household wealth using descriptive statistics and regression testing on data from the DNB Household Survey (DHS), a representative panel of about 2,000 households in the

Netherlands. First I will examine data sets on household wealth from 2007, before the crisis, and 2012, the most recent panel survey since the crisis. I will compare the individual results of both samples to see whether regression results differ before and after the crisis. There are a variety of factors that influence wealth and will be controlled for in this paper. The standard demographic factors controlled for are age, income, education, marital status, children, and the region and level of population density the household resides in. Race is often controlled for but this variable is unavailable in the DHS.

Using the one paper I could find that tested tenure choice and duration of homeownership on net

household wealth as a blueprint I will also examine a sample of renters from 2003 and compare the wealth of the renters who decided to become homeowners to renters who remained so throughout the study period which runs from 2003 to 2012, a period of one decade. In addition to the demographic variables controlled for above, the regression on this longitudinal sample will control for starting net household wealth level and average annual household income rather than current household income.

This paper from Di, Belsky, and Liu controls for some additional variables which are present in all of the regressions tested here. Inheritance and the tax advantage that comes with having a mortgage are

controlled for. Inheritance is a direct addition to wealth and is often used immediately for consumption purposes rather than saving purposes though. The mortgage interest tax deduction has an indirect affect on wealth by decreasing taxable income as a tax credit. It is an added incentive that motivates many Dutch households with high tax rates to own homes. The Di et al. paper particularly highlights its control for a savings tendency. In their paper this is done by dividing the difference in wealth at the beginning and end

of the five-year period prior to the period being tested by total income received in those preceding five years including inheritances and settlements. I was unable to repeat this measure without cutting the sample size down to a negligible level. Instead I use a psychological variable included in the DHS survey which asks the household head what their current financial situation is. For the longitudinal sample running from 2003 to 2012 I will test the financial situation given at the beginning of the paper as a proxy for a household’s initial savings tendency. I also do not control for settlements as the Netherlands is not as litigious of a society as the United States and settlements are rare.

After controlling for all these variables, duration of homeownership is positively related to wealth in the 2007 and 2012 models. This would prove that homeownership indeed is a main driver behind wealth accumulation. For the model testing the period from 2003 to 2012, starting wealth level and the tendency to save are the only factors that are positively related to wealth in 2012. This would prove that a

household’s ability to save is the main driver behind their wealth rather than being homeowners for any extended period of time, which would prove reverse causation that homeowners are indeed savers to begin with.

The first two models test larger samples making their results more significant but the sample that tests the period between 2003 and 2012 is interesting because it tests households over a longer period of time. Unfortunately the sample is much smaller. Nonetheless the results it provides remain interesting as the time horizon provides a more complete financial picture of the households and the effect of a change in tenure choice can be analyzed. It also brings to light the importance of a household’s capacity to save over time. Although homeownership has a larger impact on the 2007 and 2012 models the inability to save does have a significant negative effect on current net wealth. As Richards (2013) concludes, “In a healthy housing market, people get a mortgage loan because of what they have already done — they've worked hard, kept their jobs, paid their debts, delayed gratification, and saved for a down payment. In this virtuous circle, wise behavior makes it possible to acquire a home, and acquisition of a home reinforces wise behavior.”

1.1 The state of the Dutch housing market

Pre-crisis in the Netherlands, house prices went nowhere but up. A whole generation of homebuyers and real estate agents knew anything else. Post-crisis, however, is a whole different ball game where house prices are struggling to stay constant. In this brave new world are homeowners still paving that path to wealth?

This question is constantly being asked by the Dutch government, market, and media. From 1996 up until the crisis national wealth grew an average of 185 billion Euros a year. Since the crisis that figure has shrunk to 73 billion, not even half of what it was. Household wealth makes up the lion’s share of national wealth at 67 percent. In 1996 the value of homes and land made up 35 percent of total net household wealth. That figure grew to 58 percent by 2008. Since the crisis household wealth has shrunk by 54 billion Euros. The decrease took place fully in non-financial assets which is composed of homes and pensions. (Bhageloe-Datadin 2011) Figure 1 below illustrates the growth, decline, and present stagnation of Dutch housing market. (CBS)

House prices more than doubled between 1985 and 2007. In some European countries it even tripled. (Notten 2011) Homeownership increased. Leverage ballooned as mortgages became increasingly aggressive and creative by maximizing the tax benefit. Long-term debt in the Netherlands grew from 56 percent of GDP in 1995 to more than 125 percent in 2011. More than 90 percent of that debt comes in the form of mortgages totaling 653 billion Euros. (CBS 2011) The Netherlands has the highest level of long-term debt in the Eurozone, which grew at a faster rate than even Spain’s. (Notten 2011) The frenzy was propped up by the government with a generous tax incentive. In countries where house price growth was similar sans the tax advantage, debt increased at a much slower rate with homeowners instead choosing to pay off their mortgages sooner. (Notten 2011) After the music stopped, the Netherlands has been left with a debt hangover, a stagnant supply of real estate, immobility, and an unsustainable tax policy.

The most worrying effect is the Dutch household’s now high sensitivity to asset price movements in their homes. This increased risk profile can affect everything from asset markets to savings rates to economic growth. In down markets like this one, risk-aversion increases and people tend to spend and invest less,

0 50 100 150 200 250 300 T ho us a nds

Figure 1: House sale transactions and

average home sales price

save more, and hold depressed assets ever hopeful that if they wait they can at least break even eventually. In some cases, homeowners who want to sell may not even be in a financial position to absorb the losses caused by the binding combination of depressed house prices and maxed-out mortgages. Homeowner risk profiles are often measured by the loan-to-value (LTV) ratio which has been increasing since the crisis as house prices decrease. (Denneman 2011) The numerator tends to stay constant since the Dutch avoid decreasing the size of their mortgage in order to maximize the tax benefit. (DNB 2002) LTV ratios tend to decrease with age as longer time horizons often work in a homeowner’s favor. The most vulnerable age group is under 35 having an average LTV ratio greater than one and the crisis has not dampened this trend. (Denneman 2011). The Netherlands has one of the highest average LTV ratios in Europe. (Vandevyvere and Zenthöfer 2012)

One interesting thing to point out is that Dutch homeowners report that buying works out to be cheaper than renting for them. A 2002 report from the DNB states:

It is particularly striking that a large number of both tenants and buyers indicate that the way they live is cheapest for them. Those in the lowest income brackets often indicate they are better off renting because of rent subsidies, while higher-income families frequently cite favourable tax treatment as a reason for buying their homes. All this suggests that government housing policies have a major impact on the way the Dutch choose to live.

This generous tax treatment is unique to the Netherlands and plays a major role in the Dutch housing market. This leads to a hypothesis that tax treatment will have a major impact when testing for homeownership’s role in wealth accumulation.

Since the crisis a host of regulatory changes have taken effect to control mortgage lending. Interest rates are suppressed in an effort to stimulate the economy as a whole although mortgage lenders maintain a high margin. The maximum mortgage factor used to be between four to six times annual income. It has now been reduced to three to five-and-a-half times annual income. So-called ‘top’ mortgages were common before the crisis, where lenders were willing to write mortgages 125 percent of the actual sales prices of a home. At present this practice has been capped at 106 percent with the ultimate goal of disallowing any mortgages to be greater than the value of the house. Before the crisis it was also easy to obtain an interest-only mortgage of 100 percent the transaction price of a house. That has now been capped at fifty percent. In 2010, 56 percent of all outstanding mortgages in the Netherlands were interest-only mortgages. (Vandevyvere and Zenthöfer 2012) Newly-originated interest-only mortgages may no longer be deducted either. Only thirty-year annuity mortgages are now tax-deductible. This is a major product shift in the mortgage market as annuity mortgages have been rare in the Netherlands since the 90s. Before the crisis

Table 1: Timeline of regulatory changes in the Dutch housing market

Date Regulatory change Pre-crisis At present Goal

1 July 2009 NHG Guarantee temporarily increased to €350,000. €265,000 €290,000 €265,000 by 1 July 2014

29 March 2010 Budget exhausted for national home buyer’s subsidy.

Subsidy determined by income level.

No plans to re-introduce the subsidy.

AFM assessment of bank mortgages and lending standards September 2009

Stricter lending ratios: Closer adherence to LTI ratios and introduction of LTV ratios (maximum of 112% with portion above 100% paid off in first 7 years). Separation of consumption and mortgage loans.

Lending standards for mortgage financing effective 1 January 2011 Maximum mortgage factor (LTI) 4 to 6 3 to 5.5 Closer adherence to calculated LTI norms from NIBUD Introduction of LTV ratio at 110%. 125% 105% (Agreed in Budget 2013) Maximum interest-only loan 100% 50% Rumors Spring 2011 Effective July 2011

Sale transaction tax 6% 2% Initially

temporary. 2013 budget drawn

up on 26 April 2012. Details leaked out on 16 May.

Official press release at the end of May.

Maximum Loan-to-Value (LTV) ratio

125% 105% including 2%

transaction tax. LTV can be based on post-renovation value; energy-saving renovations increases LTV to 106%. From 2013 to 2018 LTV will slowly be lowered to 100%. Deductible mortgages Every mortgage was deductible.

Only 30-year annuity or linear mortgages are deductible. Maximum

interest-only mortgage

100% Initially 0% but increased to 50% in February 2013.

Raising rents Inflation cap Inflation + 4% Encourage out flow of renters motivated by cheap rent. Loss on sale of house Do not have to pay off mortgage and interest deductible

for 10 years.

Starters Allow a larger mortgage if salary increase is expected. Revised Budget:

Debates throughout September, final on 29 October 2012

Mortgage interest tax deduction lowered for highest income rate.

42 – 52 % Max lowered to 51.5% 38%, lower by 0.5% annually.

mortgage interest was deducted at every tax rate level. The deduction has now been capped at 51.5 percent with the ultimate goal of reducing that cap to 38 percent. Government-sponsored programs such as

“starter” mortgages (mortgages for young people entering the real estate market for the first time) and “buyer” subsidies have been halted.

In an attempt to keep the housing market from going completely bust the government reduced the sale transaction tax from six percent to two percent. It was initially a temporary move but has since been maintained with no plans to raise it in the immediate future. The government also increased the national mortgage guarantee (NHG) from 265,000 to 350,000 with a timeline to reduce it back to its original level by 2014. This has been cause for concern as it has increased the government’s potential liability pot from 109 billion in 2009 to 149 billion in 2012. Defaults have increased with divorce being a major cause for NHG losses suggesting fraud.

The government has been debating these regulations since the crisis. Pressure from the European Union has forced the government to finally reach an agreement concerning budget cuts. The debates have had a negative effect on Dutch consumer confidence as the Netherlands has yet to climb out of the recession, falling behind fellow countries reporting positive economic figures. Figure 2 below highlights the effect major regulatory changes have had on average house prices.

2 Literature Review

Di, Belsky, and Liu (2006) provide a sobering summary of the existing literature on the effect of homeownership on wealth. Data shows that homeowners have more wealth than renters and that a large

190 200 210 220 230 240 250 260 T ho us a nds o f E ur o

Figure 2: Events along average sale price curve

part of that wealth is tied up in home equity (Zywicki and Adamson 2009, Di 2003, Boehm and Schlottmann 2004) which leads to proclamations that owning a home leads to wealth. Based on studies from Engelhardt (1994), Haurin, Hendershott and Wachter (1996), and Rohe, McCarthy and Van Zandt (2002), “Homeownership is considered the most viable path to wealth creation….” (Beracha and Johnson 2012) Yet there are virtually no empirical studies that actually test homeownership’s effect on wealth. Self-selection bias could be the reason, that is: Could homeowners be wealthy to begin with? (Di et al. 2006) According to Beracha and Johnson (2012), “…homeownership self-selects to the wealthier individuals in society, while simultaneously absorbing a significant set of self-imposed savers that might otherwise spend income and wealth on non-wealth-enhancing consumption.” In Di, Belsky, and Liu’s (2006) words: “It is possible that observed higher levels of wealth among homeowners than renters reflects the propensity of wealthier and higher income households to own, as well as the propensity of those inclined to save and invest to prefer homeownership and to achieve homeownership earlier in life.”

2.1 Homeownership’s effect on society

There are studies on the social effects of homeownership such as increasing generational wealth in homeowners (Boehm and Schlottmann 2002). These studies are of particular interest to policymakers promoting homeownership’s positive effects on society. Reports on subprime lending particularly drew on this literature. (Zywicki and Adamson 2009) Homeownership’s equalizing effect on the distribution of wealth was the main driver behind government support of subprime but homeownership has been attributed to everything from civic pride and voter turnout (Rohe et al. 2002, Dietz and Haurin 2003) to less crime and a better family environment (Parcel and Haurin 2002). Zywicki and Adamson (2009) summarize the general government pitch for homeownership:

In addition to these direct benefits, homeownership apparently has a number of indirect benefits. For instance, homeownership is correlated with a substantial increase in one’s propensity to vote, dramatic improvements in children’s life outcomes, and improvements in labor market outcomes; homeownership also creates incentives to improve property, generally increases life satisfaction, and is correlated with a reduction in crime rates. (Herbert and Belsky 2006, Dietz and Haurin 2003)

2.2 Homeownership’s effect on the economy

Homeownership’s macroeconomic effects have also been closely monitored, particularly its effect on consumption and aggregate saving, essentially two sides of the same coin. Case, Quigley, and Shiller (2005) regularly measure the ‘wealth effect’ of housing wealth on consumption. Their studies show that

housing wealth is more positively correlated with consumption than stock market wealth. Skinner (1989) conducted a study on house price appreciation’s effect on aggregate saving. One of his tests concludes that housing wealth does indeed have an effect on savings. There are two possible reasons for this: 1) homeowners save for their children with the idea that house prices will continue to grow out of their reach in the future and 2) renters save more in anticipation of higher home prices as well.

2.3 Homeownership as a portion of a household’s investment portfolio

Due to the large portion that housing wealth takes up of the wealth pie, from a portfolio perspective housing looks like a risky asset whose risk-return profile is not as optimal stocks or bonds. Lack of diversification and its illiquid nature lead many researchers to view homeownership as a risky investment. This is especially a risk for low and middle-income families. McCarthy et al. (2001) concluded that “homeownership offers much better financial security for wealthy owners than for low- and moderate-income and minority owners,” because “lower-moderate-income and minority households hold more housing than is optimal in portfolio wealth, exposing them to higher risk.” Although other studies conclude housing is the only way for low-income households to achieve any level of wealth mainly due to leverage and access to leverage: “…

it is not home equity but, instead, the value of homes with mortgage financing that

turns out to be the most ‘equally’ distributed wealth. This means that less wealthy households

benefit more through financial leverage and enjoy better ‘homes’ relative to their income than do

more wealthy households.

2.4 Homeownership versus stocks and bonds

There are many studies that compare homeownership to investing in stocks and bonds. The consensus is that stocks provide better returns than real estate but that real estate provides better returns than bonds. Ibbotson and Spiegel (1984) report greater returns in stocks (11 percent) versus residential-property capital appreciation (7.4 percent) in the period from 1947 to 1982. Housing did better than bonds returning 3.98 percent. Goetzman (1993) reported the same for the period from 1971 to 1985. Bond and Stillabower (1987) analyzed the cash flows of renting versus purchasing a home and found that renter wealth grew to $100,000 over 30 years based on the assumptions that stocks grow at eight percent per year and bonds at two. Hurst et al. (1998), after comparing homeownership to an investment in stocks, found that home equity in 1984 had a negative effect on net wealth a decade later. Conversely, more recent research from Marjama (2002) found that average household home equity earnings were twice as much as average stockholder market earnings over the previous decade. Beracha and Johnshon (2012) question whether renting or owning is a better investment by setting up a market portfolio in which the renter invests the

surplus from not owning rather than using it for consumption, which is often the case. They explain that, “…while renting may be wise, any extra savings from renting might be spent on non-wealth-enhancing goods resulting in any benefits from renting versus owning disappearing in a cloud of consumption spending rather than savings. Renting (Goodman 1997, Beracha and Johnson 2012) is often the superior investment in these studies as stocks generally perform better than housing (Goetzman and Spiegel 2002) but there are some contrary findings depending on the time frames tested. Obviously depressed markets result in depressed returns in any asset class. Stocks have had a really great run over a long period of time in the United States.

2.5 Homeownership versus renting costs

A cost comparison between renting and owning estimated by Goodman (1997) found that homeowners paid six percent more for housing than if they would have rented the same property for the same period of time; the shorter the period of residency the higher these costs. But what sets homes apart from other assets is its utility value to owners. Unaccounted for homes look like risky assets due to house price volatility and the large portfolio allocation homes take up. (Peterba and Samwick 1997, Hurst et al. 1998, Sinai and Souleles 2005) The user-cost framework takes the costs of renting relative to owning into account. (Case and Shiller 1990)

2.6 Homeownership as a hedge against inflation

Other research shows housing as an effective hedge against inflation (Sinai and Souleles 2005). In a more persuasive study, Goetzmann and Spiegel (2002) claimed that returns to home investment only moderately exceeded inflation of this period, and homeownership, therefore, has high risks and is a poor investment tool. (Ibbotson and Siegel 1984, Goetzmann 1993)

2.7 Homeownerhship’s effects on wealth

Rohe et al. (2002) state, “Asset accumulation through house price appreciation is considered the main financial benefit of home ownership.” Di (2001) elaborates on the effects of housing on household wealth by identifying four roles homeownership plays in its contribution to wealth: as “equalizer,”

“accumulator,” “cultivator,” “protector,” and as a “double-edged sword.” Housing wealth is more equally distributed than other forms of wealth. This role as equalizer was the driver behind promoting housing among low-income households. It is generally believed that house prices appreciate. Housing’s role as a wealth accumulator is not that straightforward however. Real estate markets are highly cyclical and the timing of buying and selling a home plays a large role in which direction household wealth may take.

Increased access to funding from home equity growth leads to new opportunities to cultivate wealth. Many households refinanced or took out second mortgages to renovate homes, invest in stocks, or increase their level of education. These are considered investments in assets that will bring further wealth down the road. Fixed-rate mortgages also allow households to protect themselves against future increases in housing costs, such as raised rents. Capital gains in housing are also said to keep pace with the rate of inflation which has a direct effect on rents. However an extended holding period is necessary to cover the initial transaction costs associated with buying a home.

Di et al. (2006) provide a more complete summary of the how homeownership contributes to wealth: 1. Leverage: Housing provides households the unique opportunity to invest in a highly leveraged

asset that tends over long periods of time to grow at least slightly ahead of inflation (Goetzman and Spiegel 2002, Sinai and Souleles 2005) in most places. Therefore, even small percentage returns on the asset are magnified through leverage into large returns on invested capital. (Gates 2002)

2. Forced savings: Making mortgage payments is also a form of forced savings as the principal on the mortgage is paid off. Hence owners may save more than renters.

3. Increase in consumption: The literature suggests that the long-run marginal propensity to consume from each dollar increase in housing wealth in the United States ranges somewhere between 4 cents and 15 cents, but perhaps as much as 14 to 17 cents in other countries (Belsky and Prakken, 2004; Case et al., 2005; Haurin and Rosenthal, 2004).

4. Access to cheap funding through home equity loans: In addition, even though owners often elect to borrow against their home equity, home equity borrowing generally constitutes a lower cost way to finance consumption than the alternatives available to renters. Thus, this too frees up funds for potential saving and investment.

5. Inflation hedge: Furthermore, although not all costs of owning are fixed, homeownership does provide a hedge against future rent inflation (Sinai and Souleles, 2003). This also potentially leaves owners with more to save or invest over time.

6. Tax advantage: The last but not the least important advantage of homeownership comes from the tax code. Most homeowners benefit from mortgage interest and real estate tax deductions that exceed their standard deductions. The exception is low-income homeowners with small mortgages and property tax bills due to the progressive tax rate applied to the deduction. (Poterba 1992) 7. Market timing of sale and purchase: In addition to the effects of housing wealth on household

wealth that Di et al. list, duration of ownership and timing of sale effect household wealth. Timing affects the sales price and longer duration often translates into more price appreciation and a

larger amount of mortgage paid off (i.e. forced savings). A lower mortgage and higher sales price results in a profit. Contrary to this, Belsky and Duda (2002) found that, “Homeowners frequently sell homes for less than they paid for them in nominal terms and… large shares of them resell after experiencing real house appreciation insufficient to cover even transaction costs.” It is generally impossible to time the market and, “After a careful study of several local markets at different times, Case and Marynchenko (2002) concluded that the complex pattern of the real estate market cycles made it difficult to make any generalizations.”

Di et al. (2006) is the only paper that tests the direct effect of homeownership on net wealth and will be used as a blueprint for the testing done in this paper. Di et al.’s 2006 study finds that tenure choice and duration of homeownership do have a significant effect on future wealth. Though initial wealth had some impact as well their control for a tendency to save was not significant, meaning homeownership in and of itself has a greater effect on wealth than the propensity to save. Income however had the largest impact on the variation of the level of net wealth, meaning the capacity to save has the greatest effect.

In conclusion, the literature reveals that housing is generally an inferior investment to stocks but that market timing plays a major role on the level of household wealth. Though housing keeps pace with inflation and performs better than bonds the duration of the holding period must be long enough to offset any initial transaction costs. Seen solely from an investment point of view it is considered a risky asset but once its utility value is taken into consideration housing wealth’s role as a component of household wealth becomes more complex. There are a variety of benefits attributed to homeownership for the household such as tax advantages and cheap funding, not to mention its effects on society in general. Whether housing is a good investment for a household is not such an easy question to answer. The timing, the household’s level of income and general financial situation all have to be taken into account and at the end of the day a household needs a place to live.

3 Methodology and data

3.1 2007 and 2012 models

I examine households in 2007 and 2012 for a comparative check between pre and post-crisis years. The dependent variables modeled are net wealth and non-housing wealth. Home equity is subtracted from net household wealth to get household non-housing wealth. The independent variable tested is the number of years spent owning the current home using the reported year the current house was bought. Each year will be tested separately then compared.

Factors that can influence wealth accumulation are controlled for. Race is a popular factor controlled for in American papers however race was an unavailable variable in the DHS data. This variable carries more weight in the United States as the population is more diverse with a variety of large ethnic groups. The Dutch population is far more homogenous with very small pockets of ethnic groups. As of 2009, 81 percent of all households in the Netherlands have a household head of Dutch origin and 89 percent of all owner-occupied housing has a household head of Dutch origin. (Denneman 2011)

Income is considered the main source from which households accumulate wealth. In this paper current income is measured. A proxy for permanent income would be a more reliable variable as some households have varying annual incomes. Education level is often used as a proxy for permanent income (Choudhury 2002; Johnson 2006) or financial literacy (Di et al. 2006). Choudhury (2002) found that education is highly correlated with long-term financial well-being and that homeownership patterns based on education are similar to homeownership patterns based on income. In this paper education is not used as a proxy for another variable but it is controlled for on its own merit.

I control for gross income and net income. They are highly correlated so the regression should throw out one unless it really has an impact on its own. Testing for net income may reveal some advantages that some low- or high-income earners have. For instance low-income earners may gain from subsidies and other government benefits or high-income earners may benefit from the mortgage interest tax deduction. Some households, usually older ones, may have large interest income. It is common practice to test the log of income but taking the log of income of the samples in this paper made the distribution even more skewed with higher kurtosis. Therefore I will not transform the income values in this paper.

Inheritance will also be controlled for as it has a direct effect on a household’s savings capacity.

Inheritance is often used for consumption purposes so I do not expect it to have much impact on the level of wealth. However there is an outlier in one of the samples, a household that received an unusually high inheritance and went on to buy property with it. I will exclude this case from the models as it distorts the average.

The amount of gross mortgage interest tax deductions is also controlled for. I expect this variable to have some impact and magnitude as the tax advantage of homeownership motivates high-income households to buy homes. Many households who do not qualify for subsidies or other government benefits due to income levels being too high find buying homes more cost-effective than renting in the private rental market. Gross mortgage interest tax deductions should be highly correlated to income as the level of deduction is based on the income tax rate.

Age will also be controlled for. Life-cycle affects housing choices. Young people tend to be renters, once established in their careers the prospect of homeownership becomes a possibility. Retired people over 65 may switch back to renting after owning for a variety of reasons such as they cannot keep up with the maintenance a home needs or the house is too big with all the children gone. Also, the older one is the longer a person could have owned a house, so age obviously effects duration of ownership.

Marital status is controlled for as many households with married couples are double-income households therefore increasing the capacity to accumulate wealth and costs go down due to economies of scale. Divorced households may suffer from a strain on resources due to the costs divorce entails.

Number of children is also controlled for. Some studies find that having children increases a household’s savings tendency as they save for their children’s future whereas others find children increase a

household’s expenses. The impact of having children on a household is unclear. (Di et al. 2006)

A proxy for a tendency to save can be controlled for. There is a variable in the psychological segment of the DHS survey which has the sample subject rate the importance of having money left over to save and another variable asking whether the household actually saved that year. This would only capture savings behavior and not investment behavior. There are variables in the survey that profile a subject’s risk appetite but not their actual investment behavior. There is also a measure for how a subject views his financial literacy. These variables are based on the subject’s personal opinion of himself rather than actual numbers. Nonetheless the more financially-literate or important savings was to households the greater their average net wealth in spite of average gross income remaining constant and evenly distributed among options. Nonetheless, the variable I have chosen to represent a savings tendency assesses a household’s

-50 0 50 100 150 200 250 300 350 400 450 Have debts Manage Use savings Save some Save a lot T ho us a nds o f E ur o

Figure 3: Current wealth and income

levels by current financial situation

Wealth 2012 Wealth 2007 Non-housing 2012 Non-housing 2007 Income 2012 Income 2007

present financial situation, whether they are currently paying off debts, tapping into savings to get by, managing to get by, or if they are able to save some or save a lot. This variable had the highest correlation to net wealth and was the most highly correlated with other savings proxy variables. For example, the likelihood that families actually saved in the current year when the savings options was chosen was higher. As with the other variables, wealth tends to be higher for any option involving savings despite income staying constant and fairly evenly distributed over the chosen options here as well (although there is a slight increase in income for those able to save). See Figure 3.

Geographic regions and how urban a location is are controlled for. House price appreciation differs per region and generally due to low supply and high demand the more urban a location the more expensive the property. In the Netherlands, the province of Holland in the West has a higher population density and is the economic hub of the country which increases the cost of living and in turn house prices in this area. Although wages are higher in this region so are costs. It will be interesting to see how these factors balance out in the regression.

Wealth is highly skewed with high kurtosis. The papers from Di use the log to normalize the skew. €1 is assigned to cases with negative net wealth and zero values to avoid the loss of cases. All numbers are adjusted to 2012 Euro values.

3.2 Di et al.’s model for homeownership in the long-run

I will also run another model based Di et al.’s empirical test. It uses panel data from the Panel Survey and Income Dynamics (PSID) which is a longitudinal survey. The researchers start with renters in 1989 and analyzed the impact the choice homeownership has had on the level of their net wealth in 2001. I will test from the period 2003 to 2012. The dependent variables are net household wealth and non-housing wealth in 2001 (2012 in my model) and the independent variable is the number of years spent owning a home. There is a duration and timing effect being picked up in the independent variable:

Given this, we allow for the possibility that duration of homeownership is non-linear in effect on wealth accumulation by introducing a variable that is the square of the number of years in homeownership. Theoretically one might expect the sign of the coefficient on this variable to be positive because the longer a household owns a home the greater is the likelihood that the owners will add to wealth not only through a leveraged investment in an appreciating asset but also through amortization of the loan. But in fact the influence of duration of homeownership on wealth accumulation is an empirical matter linked to the timing of home purchases and sales. This is because changes in the value of the home overwhelm the influence of any amortization,

especially in the first one to twelve years of a mortgage loan. Given that those with longer durations in the sample more likely endured a period of slow price appreciation or depreciation early in their tenure, one would expect the coefficient on this variable to be negative.

The sample tested in this paper will have the opposite scenario described above with house prices appreciating at the beginning of the period and stagnating or declining at the end of the period. Nonetheless, since there is a decline in the sample period I also expect a negative coefficient.

This paper by Di et al. (2006) examines a period of twelve years but my sample period begins in 2003 because the Euro was fully introduced at that point. In this way I avoid having to correct for any effect the Euro conversion may have had on wealth in the study period. Fixed mortgages prior to the Euro benefitted from not being affected by the increased price inflation that came with the Euro. These households may have more favorable home equity levels and distort the affect that duration of homeownership may have on wealth. In addition 2003 provided a larger sample compared to earlier periods.

Average annual income instead of current income is used to control for differences in household income. It more closely approximates a measure of permanent income which better indicates the capacity to accumulate wealth. Net household wealth at the beginning of the tested period is also controlled for. Di et al. controlled for a savings tendency, which they considered a unique key for proving causation in their study. The savings tendency was a measure of the change in wealth over total income in five years prior to the period modeled. I attempted to calculate a savings tendency variable in the same way but it reduced my sample size to such a negligible number that I was unable to control for this variable. There were too many missing variables and sample subjects that were not yet included in prior years. In their study the savings tendency had little impact or magnitude. Any attempt to test the savings tendency in my small sample produced similar results. In addition, probably as result of the crisis the savings tendency was predominantly negative. Many sample households experienced a decrease in net wealth. I even calculated the savings tendency for the 2007 and 2012 models above. I had the same issue with a vastly shrinking sample with each passing year considered and they were also predominantly negative. It also had no correlation between any of the savings tendencies calculated or wealth. As mentioned above there are variables in the Economic and Psychological Concepts questionnaire that can be used as a proxy for savings tendency, however the variable is based on the subject’s opinion of his savings behavior rather than actual numbers.

The remainder of the variables controlled for include the control variables mentioned above in the 2007 and 2012 models. An additional dummy variable included here is the change (i.e. improvement) in

education in addition to the level of education at the start of the study period. A similar dummy variable was added for the net change in the number of children in the household rather than the number of children in a given year in the study period. The percentage of time married and whether the household head had a divorce during the study period is also controlled for. Divorce generally reduces household wealth and is included as a dummy variable in this model. A variable that can be added since an entire period is being tested in the model is net wealth at the beginning of the sample period. This variable may also capture a tendency to save.

The equation for the model is as follows:

Log W = α + βh * H + βsh * SH + βc * C + ε

W = net household wealth in 2001 α = the intercept

βh = the coefficient for years of homeownership (H)

βsh = the coefficient for years of homeownership squared

βc = a vector of coefficients for the covariates (C)

ε = the residual

3.3 Data

Researchers commonly use household surveys to assess wealth. There are a variety of data sources on income, wealth, and mortgages such as those provided by the Centraal Bureau voor de Statistiek (CBS) or Rabobank but household surveys are one of the only ways to get a complete picture of a person’s financial situation. “A survey is a prime tool for gathering detailed information on various groups of households and their asset breakdowns, motives for holding various asset and liability categories and financial behavior.” (DNB 2002)

The DNB has been conducting an annual household survey since 1993. It provides longitudinal data on the financial behavior of a representative sample of individuals and their households in the Netherlands. The sample size consists of about 2,000 households. The DNB Household Survey (DHS) is available to researchers and students. It is made up of six questionnaires: 1) Work and Pensions, 2) Housing and Mortgages, 3) Income and Health, 4) Assets and Debts, 5) Economic Concepts, and 6) Psychological Concepts.

For the purposes of this thesis the focus will be on the questionnaires on Housing and Mortgages and Assets and Debts. The questionnaires on Income and Pensions provide additional insights. An obvious example would be the influence of income on wealth. The last questionnaire on Psychological Concepts can provide insights on what direction they feel their financial position will take. A proxy for a tendency to save will be pulled from this questionnaire.

The data set begins at 2,000 and tends to decrease with each additional variable tested and controlled for due to missing values. In addition, the models that tested over a period of time suffered from households not answering the questionnaires in some years and sample atrophy (i.e. sample subjects leaving the study). It seems the households with more complete data tend to have a better understanding of their financial situation and have higher wealth values and incomes and are likely to be homeowners. The homeownership rate is also higher than the national rate. Therefore complete data sets are biased towards financially-literate households.

3.4 Descriptive statistics

Table 2: 2012 sample descriptive statistics sans outliers (sample size 1059)

Variable Definitions Mean Median Minimum Maximum

Net wealth 2012

Net household wealth in

2012. 228,679 182,709 -669,598 2,530,000

Net wealth 2012 log

Log of net household wealth in 2012. 10.98 12.12 0.00 14.74 Non-housing wealth 2012 Non-housing wealth in 2012. 82,233 32,783 -332,364 2,064,000 Non-housing wealth log

Log of non-housing wealth in

2012. 9.6158 10.39767 0 14.54016

Years of ownership

Number of years since household bought its current house. 13.70 11.00 0.00 59.00 Years of ownership squared

The square of the number of years since household bought its current house.

358.45 121 0.00 3,481 Income gross

Current gross income in 2012. 42,729.79 39,249.50 0.00 276,557.50 Inheritance Total amount of inheritance

received by household. 1,114.87 0.00 0.00 251,000.00 Mortgage interest tax deductions

Amount of mortgage interest deducted when taxes filed in 2012.

2,981.49 0.00 0.00

48,528 Age Age of household head in

2012. 58.54 60.00 16.00 89.00

Children Number of children in household in 2012.

0.52 0.00 0.00

6.00

Table 3: 2007 sample descriptive statistics of categorical variables sans outliers (sample size 1059) Variables Share % Variables Share % Variables Share %

Living in areas: Region: Highest level of completed education:

Very high urban 14.4 3 largest cities 14.4 Special 0.2

High urban 26.7 Other West 28.1 Primary 2.5

Moderate urban 21.5 North 12.7 VMBO 26.4

Low urban 21.0 East 20.6 Secondary 10.8

Very low urban 16.4 South 24.2 MBO 15.4

Marital status: Financial situation: Technical university 27.3

Single 13.4 Have debts 2.9 Univeristy 16.7

Cohabitant 8.2 Using savings 18.3 Other 0.7

Registered 8.8 Manage to get by 20.4

Married 55.6 Save some 47.3

Divorced 7.2 Save a lot 11.0

Widow 6.8

Table 4: 2007 sample descriptive statistics sans outliers in 2012 Euros (sample size 1018)

Variable Definitions Mean Median Minimum Maximum

Net wealth 2007

Net household wealth in

2007. 212,745 152,883 -350,570 2,312,919

Net wealth 2007 log

Log of net household wealth in 2007. 10.77 11.94 0 14.65 Non-housing wealth 2007 Non-housing wealth in 2007. 71,952 32,746 -490,289 2,014,466 Non-housing wealth log

Log of non-housing wealth in

2007. 9.61478 10.39654 0 14.51586

Years of ownership

Number of years since household bought its current house. 10.68 7.00 0 54.00 Years of ownership squared

The square of the number of years since household bought its current house.

252 49 0 2,916 Income gross

Current gross income in 2007.

42,662

39,884 0 189,644

Inheritance Total amount of inheritance

received by household. 674 0 0 132,156

Mortgage interest tax deductions

Amount of mortgage interest deducted when taxes filed in 2007.

2,823 0 0

26,431 Age Age of household head in

2007.

53.26 54 24 92

Children Number of children in household in 2007.

0.62 0 0

5.00

Table 5: 2007 sample descriptive statistics of categorical variables sans outliers (sample size 1018)

Variables Share % Variables Share % Variables Share %

Living in areas: Region: Highest level of completed education:

Very high urban 16.9 3 largest cities 17.5 Special 0.4

High urban 26.9 Other West 29.1 Primary 3.4

Moderate urban 19.4 North 11.6 VMBO 25.7

Low urban 21.7 East 19.5 Secondary 10.9

Very low urban 15.1 South 22.3 MBO 19.9

Marital status: Financial situation: Technical university 24.1

Single 16.3 Have debts 3.9 Univeristy 14.6

Cohabitant 8.3 Using savings 16.2 Other 0.9

Registered 8.0 Manage to get by 23.1

Married 55.1 Save some 47.8

Divorced 7.3 Save a lot 8.9

Widow 5.1

3.5 Graphs based on descriptive statistics for 2007 and 2012 samples

The homeownership rate in the Netherlands has been growing over the last 40 years. The graph below illustrates how homeownership in the samples started growing in the 70s and continued to increase after the 1980 housing crash until the most recent years preceding the crisis.

,0 1,0 2,0 3,0 4,0 5,0 6,0 19 53 19 59 19 63 19 66 19 69 19 72 19 75 19 78 19 81 19 84 19 87 19 90 19 93 19 96 19 99 20 02 20 05 20 08 20 11

Figure 4: Percentage of households that

bought their house in a certain year

2007 2012

Housing wealth is the greatest contributor to net wealth and is highly correlated with it. Home equity makes up 64 percent of sample net worth and 67 percent of homeowner net worth in both 2007 and 2012. To compare, home equity in the United States makes up 40 percent of net worth. (Zywicki 2009, Di 2003) Normally the lower the income the more net wealth is made up of home equity as households with less wealth have less to diversify. This is the trend in the graph below in 2012 but in 2007 owner wealth has no trend with the lowest income group having the lowest amount of home equity making up its net wealth. In contrast, the 2012 low-income group is much more dependent on home equity to pump up its net wealth value with more than three-quarters of its wealth made up of home equity.

On average, owners had about €260k (in 2012 Euros) and €247k more net wealth than renters in 2007 and 2012 respectively. See figure 6 below. The difference between owner and renter net wealth fell slightly since the crisis. Owner non-housing wealth barely grew but still averaged more than twice as much than tenant wealth. Renter wealth increased at a higher rate than owner wealth having 18 percent more wealth after the crisis whereas owner wealth and non-housing wealth increased by only 5 percent.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% All households Owners Below 20k 20-50k 50k+

Figure 5: Home equity / net wealth

2007 2012

In general wealth tends to increase with age. See figure 7 below. Older people have more wealth than younger people. Renters do not achieve as much wealth as owners in any age bracket and the upward trend in wealth accumulation by age seems to stop for renters once they no longer have an income, that is older renters once retired cease growing their wealth. Broken down by age the graph shows that the oldest homeowners experienced a decrease in wealth after the crisis perhaps because older household net wealth was largely made up of home equity. They benefitted the most from the prolonged upward trend in house prices having longer durations of ownership and often having bought homes early in the market upswing. As mentioned above, older homeowners once retired may also have fewer channels for accumulating wealth. 50 100 150 200 250 300 350

Renter Owner non-housing Owner

T ho us a nds o f E ur o

Figure 6: Average wealth renter vs

owner

Figure 7: Average wealth by age

Below 35 35-65 65+

Figure 8 below shows no matter what the level of income owners again have more wealth. In general, the higher a household’s income is the higher its wealth. However in 2007 the lowest income group of owners’ wealth exceeded the middle-income group of owners’ wealth. Initially I guessed this represented a large group of baby boomers retiring early (hence the low income) benefiting from home equity

windfalls over the several decades they were homeowners. However more than a third of non-housing wealth in this group comes in the form of real estate not used for accommodation and most of these households also have business equity. These households have the highest net wealth in this group and are responsible for driving up the average. Perhaps real estate was an attractive business investment before the crisis and perhaps a reluctance to disclose income information due to high taxes puts this group in the lowest income bucket. Business owners and landlords also tend to have variable incomes so some years they may do better than others. Nonetheless this group is notorious for maneuvering around tax laws. Only one household from this group remains in the survey in 2012 so this may not be a representative sample of low-income earners.

Homeownership seems to have the effect of equally distributing wealth. Renter households earning an income above €50,000 have more than double the amount of wealth than other renter income groups. The disparity in wealth between top-earners and low-earners among owners is lower than in the case of renters.

Figure 9 below shows that the portion of homeowners in the sample increases with income. Starting at the lowest income there is a 50/50 chance in 2007 and 60/40 chance in 2012 that the household owns a home. The homeownership rate increases with income hitting 70 percent by the €30,000 - €45,000 range. The homeownership rate for the entire sample before the crisis is 69 percent and 75 percent after the crisis

50 100 150 200 250 300 350 T ho us a nds

Figure 8: Average wealth by income

Below 20k 20k-50k 50k+

which is higher than the national homeownership rate of 59 percent (CBS WoOn 2009). Despite the crisis it seems that homeownership is still going strong.

Across all durations of ownership, owners had more wealth than renters before and after the crisis. See figure 10 below; the longer a person owns a home the higher their wealth. However, it looks as though real estate market cycles have more of an impact on the curve than duration. In the following home equity graph (see figure 11), which lines up the curves based on what year the house was bought rather than how long the house has been owned, the 2007 and 2012 data series show a similar trend. Is it the effect of duration (purely the time one owns a home) that increases wealth or is it the effect of the housing market becoming overcrowded that decreased housing wealth over the last two decades? Note how the home equity curves below show that homeowners who bought before 1990 had an erratic yet high level of home equity whereas homeowners who bought after 1990 did not have as high a level of home equity. In the wealth difference graph (figure 10) this trend is translated into a smaller amount of excess wealth over renter wealth the shorter the duration of ownership. Nonetheless being a homeowner even after the crisis results in greater net wealth compared to non-owners. (Note in 2012, there are a couple households driving up average.) 0% 20% 40% 60% 80% 100% 120%

Figure 9: Percent owners vs renters by

income

owners 2007 renters 2007 owners 2012 renters 2012

The last time house prices dipped before the most recent crisis was in the early 1980s. Anyone who bought a home then has been able to build up a large amount of home equity as house prices have since grown far beyond that point. The forced savings principal leads to higher home equity levels through the amortization of loans over time but the downward trend after the 1980s in the figure above could be because of the trend in rising house prices. Waiting meant buying at higher and higher prices with each passing year. In addition, new homeowners started maxing out their mortgages in the 1990s as new mortgage structures took full advantage of the generous mortgage interest tax deduction available in the Netherlands. At the same time top-up mortgages were easily accessible. This increase in leverage has eroded home equity levels. Figure 12 below shows that loan-to-value ratios grew dramatically since the 1990s. In the home equity graph (figure 11) the 2012 home equity curve has increased since the crisis.

100 200 300 400 500 600 0,5 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 T ho us a nds o f E ur o

Years of ownership in current house

Figure 10: Wealth difference owners vs

renters

Year current house was bought

Figure 11: Average home equity

2007 2012

Could this be the effect of stricter mortgage lending rules translating into less leverage through more conservative home loans? LTV ratios have also decreased since the crisis returning to levels prior to the 1990s.

Average LTV ratios in 2007 and 2012 are 38% and 33% respectively again illustrating the decrease in leverage since the crisis. The average LTV ratio curves in figure 12 above show this decrease in leverage over the last five years. The graph below shows that the younger the household the higher the LTV ratio. Obviously the older the household the more time it has had to pay off its mortgage but this could also reveal the lax lending terms younger generations have been privy to the last couple decades.

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9

Year current house bought

Figure 12: Average LTV ratio

2007 2012 0 0,2 0,4 0,6 0,8 1 1,2 Below 25 25-35 35-45 45-55 55-65 65+

Figure 13: LTV ratio by age

2007 2012

As shown in the average net wealth graph (figure 6) above homeowners have twice as much non-housing wealth as tenants. Figure 14 below shows that the difference in non-housing wealth between owners and tenants generally remains above zero. (I excluded one outlier that had an unusually large consumer credit liability for this sample.) Movements in house prices seem to have an effect on non-housing wealth as well. The graph shows an upward trend in non-housing wealth in the late 1970s and a dip in the early 1980s. Another upward cycle surfaces in the 1990s with a downward cycle surfacing at the turn of the century. The last upward trend in the 2000s seems to end in the lead-up to the crisis. Surprisingly the development in the last few years since the crisis is very positive. There seems to be a housing effect on all sources of wealth. Perhaps all markets are booming or homeowners feel richer and invest more or have more access to funding that they manage to transform into wealth, for example by buying more property or investing in the stock market. This funding often comes in the form of a second mortgage and the Dutch tend to use that money for home improvement projects. Other popular uses of top-up mortgages were speculating in the stock market and consumption, particularly cars and holidays. As a result, the government stopped allowing mortgage interest tax deductions on funding used for purposes other than home improvement. Over time, fixed housing costs decrease and free up funds for savings and investment as well. This is the positive effect of duration. Although the curve in the graph below is erratic it seems the earlier a household purchased a home the higher the chance that its non-housing wealth is larger.

-50 50 100 150 200 250 300 20 12 20 09 20 06 20 03 20 00 19 97 19 94 19 91 19 88 19 85 19 82 19 79 19 76 19 73 19 70 19 67 T ho us a nds

Year current house was bought

Figure 14: Non-housing wealth

difference between owners and renters

2012 2007

4 Empirical Results

4.1 Net wealth 2007 and 2012 models

Years of ownership in current home has the largest impact on net wealth in both 2012 and 2007 causing 16 and 21 percent respectively of the variation in the level of net wealth. Years of ownership squared has the effect of increasing the coefficient of years of ownership. In 2012 years ownership and years of ownership squared accounted for 21 percent of the variation in the level of wealth. Combined with having debts the R-square is 25 percent. In 2007 years of ownership squared adds a similar amount of variation to the level of wealth however having debts had more of an impact on the R-square in the pre-crisis sample. Together the three top variables raise the R-square of the 2007 sample to 37 percent.

Years of ownership squared has a negative coefficient of very low magnitude. This is in line with Di et al.’s result. They explain that due to the downturn in their sample period a negative coefficient was expected. Perhaps previous downturns are reflected in the value of this sample as well or perhaps it means that once the timing of the purchase is controlled for duration of homeownership actually has a negative effect on wealth although the magnitude is practically zero.

Table 6: R-square impact and regression coefficients predicting household wealth in 2007 and 2012

Dependent variables 2007 Change in R-square % 2007 Coefficients 2012 Change in R-square % 2012 Coefficients Intercept 7.75548*** 7.59609*** Years of ownership 20.76 0.26321*** 16.40 0.20961***

Years of ownership squared 6.60 -0.00522*** 5.04 -0.00372***

Have debts to pay 9.26 -4.88151*** 3.34 -3.31363***

Age 1.82 0.03235*** 2.32 0.03614***

Manage to get by financially 1.41 -0.98871*** 1.02 -0.79217***

Annual gross income 0.35 0.00001*** 0.71 0.00001***

Mortgage interest tax deduction 0.23 -0.00004 ** 0.97 -0.00006*** Living in a very high urban area 0.99 -1.05545*** 0.48 -0.62959***

University 0.58 0.59248 ** 0.45 0.61075*** Using savings 0.81 -0.65810*** 0.36 -0.53007 ** Divorced 0.30 -0.78594 ** Single 0.38 -0.62398 ** Registered partner 0.37 0.74807 ** **p <. 05; ***p < .01

Age raises the R-square to 27 percent in 2012 and 38 percent in 2007. Age is highly correlated with years of owning. The older you are the longer you could have owned a home. Age also has a decent-sized magnitude.

Income increases the R-square to 29 percent in 2012. In 2007 its impact is low. It is responsible for some slight variation in the level of net wealth but the magnitude of its impact is very low. In the Di et al. model income plays the most important role in growing wealth, perhaps because income levels tend to be higher in the United States with a wider range between low and high-income earners. In the case of net income, researchers in America do not take it into consideration. Perhaps because the low tax rates and lack of government benefits do not allow for much disparity between gross and net income. In my model gross income also has more impact on the R-square than net income. I also decided to take net income out of the model in the end because its impact and magnitude were minor but it was shrinking the sample size by a large amount. Mortgage interest tax deductions raise the R-square slightly but have little magnitude on net wealth. So although the generous mortgage interest tax deduction motivates high-income earners in the Netherlands to become homeowners and has some impact on the variation in the level of net wealth it does not contribute to wealth accumulation by a significant amount. Surprisingly the coefficient leans toward the negative although it is basically zero. Large tax deductions come from large mortgages so perhaps the negative effect of a high debt load outweighs the positives of the tax advantage.

Living in very urban areas has a negative effect on wealth. The magnitude of the coefficient is very high in both 2007 and 2012. The higher cost of living in urban areas could attribute to its negative effect on wealth. Going to university also has impact and a very high magnitude in both the 2007 and 2012 models. Being in a registered partnership further raises the R-square and produces positive coefficients of a very high magnitude in the 2007 model. Other forms of marital status besides being married have a negative effect on wealth, particularly being single or divorced in the 2012 model, both having a very high

magnitude as well. The savings tendency surfaced in both models. Having debts, just managing to get by, or tapping into savings all had very high magnitudes particularly having debts.

4.2 Non-housing wealth 2007 and 2012 models

The next model tests non-housing wealth as the dependent variable instead of all of net wealth. The housing component is extracted by subtracting home equity.

Having debts accounts for 12 and 7 percent of variation in the level of non-housing wealth in 2007 and 2012 respectively. Together with age and gross income these variables account for 20 and 14 percent of the variation in the level of non-housing wealth in 2007 and 2012 respectively. The coefficient of having