Jump to ContentJump to Main Navigation
Homeownership and the Labour Market in Europe$

Casper van Ewijk and Michiel van Leuvensteijn

Print publication date: 2009

Print ISBN-13: 9780199543946

Published to Oxford Scholarship Online: October 2011

DOI: 10.1093/acprof:oso/9780199543946.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2020. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 27 May 2020

Homeownership, Social Renting, and Labour Mobility across Europe

Homeownership, Social Renting, and Labour Mobility across Europe

Chapter:
(p.52) (p.53) 3 Homeownership, Social Renting, and Labour Mobility across Europe
Source:
Homeownership and the Labour Market in Europe
Author(s):

Thomas de Graaff

Michiel van Leuvensteijn

Casper van Ewijk

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199543946.003.0004

Abstract and Keywords

European unemployment is the main focus of this chapter. The chapter looks at various transaction costs such as financial, psychological, social, cultural, religious, and ethnic transaction costs. This chapter describes the differences in residence mobility patterns and social renting institutions in several countries. The chapter asserts that even if social leasing and privately owned residences encourage housing transaction expenditures, only renting heightens a person's probability of becoming unemployed. Potential justifications would be related to the relative returns of investment as well as the disadvantages perceived by both parties — homeowners and social renters. It would be interesting to explore the effects of these spending habits on the individual outflow of unemployment.

Keywords:   unemployment, transaction costs, residence mobility, social renting, labour market, homeownership

3.1. Introduction

The dominant adjustment mechanism is labor mobility, rather than job creation or job migration. Labor mobility, in turn, appears to be primarily a response to changes in unemployment, rather than in consumption wages.

(Blanchard and Katz 1992, p. 52)

This quote stipulates the importance of residential mobility regarding both individual labour market status and aggregate unemployment levels. Indeed, individuals find jobs easier when willing to move residence, and national employment levels are positively correlated with national migration rates.

The probability of moving residence differs widely across countries. Oswald (1997) regards this as one of the fundamental reasons for European unemployment levels being structurally higher than unemployment levels in the United States. Moreover, countries differ as well in the various transaction costs which enter the housing market—whether they are monetary, social, or psychological. Some of these transaction costs are explicit, such as transaction taxes for homeowners, which can be as high as 13–15 per cent of the residential property values for countries such as Belgium and Italy. Other transaction costs are less explicit, such as cultural, ethnic, and religious differences within countries, but may nevertheless hamper free migration between local labour markets. Especially since the seminal working papers of Oswald (1997, 1999), the impact of these transaction costs on being a homeowner have been extensively studied. Although the impact of homeownership on unemployment is still somewhat controversial, the impact of homeownership on residential mobility is not. Compared with private renters, homeowners are considerably less mobile in the housing market; in Europe, they move residence on average four to five times less than private renters.

(p.54) However, renters who rent from a public, municipal, voluntary, or non-profit agency—henceforth, social renters—face substantial moving costs as well. Usually, social renters have accumulated certain rights to rent residential property below market rents which are not transferable across municipalities or local labour markets. Thus, when moving to a new local labour market, i.e. to accept a job offer, social renters lose (the option value of) these rights, and, as a consequence, these add up to the transactions costs of moving residence.

In contrast to homeownership, not much research has been done on the relationship between social renting and labour market behaviour. Moreover, because the institutional framework regarding the housing market differs across countries—which accounts both for social renting and for homeownership—empirical research into the impact of transaction costs embodied by social renting and homeownership should focus on country-specific effects. Therefore, the main objective of this chapter is to look into the impact of social renting and homeownership on labour market mobility across Europe.

We do so by looking at the exit rates of workers to another job, unemployment, and non-participation. For exit rates out of unemployment by homeowners and renters under rent control we refer to Svarer et al. (2005), and Chapter 4 below, which have looked into this for the case of Denmark and, for homeowners, to Van Leuvensteijn and Koning (Chapter 8 below) and Van Vuuren and Van Leuvensteijn (2007) for the case of the Netherlands. The specific contribution of this paper to the literature is twofold: namely, (i) it takes a European perspective and performs cross-country analysis for most countries in Western Europe, and (ii) it includes data on the issue of social renting as well and analyses its impact on mobility. This chapter is explorative in nature; it considers the differences in homeownership and social renting institutions between the various countries, but we do not aim to develop a structural model explaining the differences in labour market effects.1 Neither do we take differences in commuting time—between countries and between residential tenure types—into account. We refer to the Appendix for an overview of the differences in housing market institutions across Western European countries.

The remainder of this chapter is organized as follows. The next section provides a survey of the literature. Because the literature on the impact of social renting is rather limited, the section emphasizes the impact of homeownership. However, one might argue that, theoretically, the process is to some extent similar, because they both involve residential transaction costs as a barrier to mobility. The third section deals with the data and provides an overview of some preliminary facts regarding residential mobility, homeownership, and social renting across fourteen countries in Western Europe. Thereafter we continue with the econometric specification followed by a discussion of the results for both homeowners and social renters in Europe. Although the results are (p.55) presented in a descriptive manner, the discussion provides an interpretation as well. The last section concludes.

3.2. Theory and literature

The impact especially of homeownership but also of social renting on residential transaction costs has already been well established in the literature. Moreover, it is also well established that transaction costs hamper mobility in the housing market, and, as a consequence, in the job market. However, high residential transaction costs do not seem necessarily to entail high probabilities of becoming unemployed. The following subsection therefore first argues why—for individual workers—the theoretical impact of reduced mobility on the probability of becoming unemployed is ex ante not clear. Thereafter, it reviews both the micro- and macroeconomic literature concerning the impact of homeownership and social renting on labour market mobility.

3.2.1. Theoretical impact of reduced mobility on labour market behaviour

Higher transaction costs on the housing market clearly reduce residential mobility. Munch et al. (Chapter 4 below) show in a search theoretical framework that is ex ante not clear what sign the impact of a reduced residential mobility has on unemployment. This depends on the labour market behaviour of the worker herself and the characteristics of the labour market. Basically, there are two competing hypotheses:

  1. 1. Workers who face higher transaction costs on the housing market, such as social renters and homeowners, are inclined to decline more job offers: usually, because some of these job offers originate from different regional labour markets and are therefore not acceptable (if commuting fails to be an option) or because social renters and homeowners demand higher wages to compensate the transaction costs to move residence. Therefore, social renters or homeowners may search longer and have longer unemployed spells or enter unemployment faster than, e.g., private renters who are more inclined to move residence to accept a suitable job offer.

  2. 2. Or those workers are—ceteris paribus—more committed to their current job as long as it is within their specific local labour market. They have more to lose when forced to move residence, especially when the transaction costs are sunk (for homeowners) or when the transaction consists of losing substantial benefits or a large option value (for social renters). Such job commitment could manitest itself in longer (unpaid) working hours, lower reservation wages, and higher productivity levels. This also entails that they leave unemployment faster, because their reservation wages for job offers in the local labour market are lower. Thus, compared with private renters, their job spells are longer, and their unemployment spells are shorter.

(p.56) Although both homeowners and social renters are affected by transaction costs, the impact may differ empirically because of the size of the transaction costs. Apart from that there are several other possible reasons why the outflow into unemployment is different between homeowners and social renters.

First of all, there are external effects associated with homeownership and social renting. As Dietz and Haurin (2003) have shown, most external effects of homeownership that are not directly related to labour market behaviour are beneficial. Homeowners are, e.g., more attached to their residence than others, and therefore willing to invest more in their residential property and the surrounding neighbourhood. For social renters, higher attachment to the residence has not been reported, but is likely to be less than for homeowners. Disregarding for the moment selection effects amongst workers to the social renting sector, it might even be that social renting causes a negative external effect because social renters are, e.g., usually not responsible for the maintenance of the residence.

Secondly, there may be social security or fiscal reasons for homeowners or renters to avoid unemployment as much as possible. Subsidies or unemployment benefits may be lower for homeowners than for renters. For example, several Western European countries still allow (partly) for mortgage interest deduction from income taxation. When homeowners enter unemployment, monthly mortgage payments increase substantially.

Finally, landlords may be more forgiving in terms of monthly payments than mortgage providers, especially when homeowners are highly leveraged.

Which of the two hypotheses above is correct is mainly an empirical question, both for homeowners and for social renters. The next subsection first reviews the mostly empirical literature considering the effects of owning a home on labour mobility. Subsequently, it looks into the smaller empirical literature concerning the effect of social renting and rent control.

3.2.2. Previous literature

The theoretical literature on the relationship between homeownership on the one hand and job mobility and unemployment on the other hand does not predict a clear ex ante outcome with respect to the direction of this relationship. From a macro perspective, Green and Hendershott (2001b), e.g., offer three additional explanations for the fact that homeownership may cause reduced labour market mobility. First, when the economy is in a downturn, housing becomes a very illiquid asset, causing homeowners to be reluctant to sell their homes and search for appropriate jobs outside their local labour market. Secondly, high interest rates may cause homeowners to be locked in as well, with similar consequences for residential mobility. And finally, high transaction costs usually associated with homeownership may also cause reduced residential mobility.

(p.57) Dohmen (Chapter 1 above) looks at the consequences of homeownership in a microeconomic theoretical framework with search and moving costs and derives, ceteris paribus, that homeowners are less mobile than renters. However, high-skilled homeowners may be more mobile than low-skilled renters if the income loss associated with unemployment exceeds the income loss associated with moving house. Along the same line of thought, Van Vuuren and Van Leuvensteijn (2007) analyse the relationship between expected labour market outcomes and the housing market in a search theoretical framework, and argue that the empirically often observed negative relation at the micro level between unemployment durations and housing tenure boils down to an unobserved heterogeneity problem: namely, workers only become homeowners when expectations on the labour market (e.g. within their current job) are favourable. And the researcher usually observes realized outcomes instead of expectations.

Regarding the individual relation between residential and job mobility, several modelling frameworks exist, mostly in the context of job search theory (see, e.g., the study of Van den Berg and Gorter 1997). Theoretical predictions for the impact of residential mobility on job mobility are less ambiguous than for the impact of housing tenure. If workers face substantial (monetary) costs in changing residence, job mobility is severely hampered. These moving costs often result from particular types of housing tenure, but may also stem from household characteristics (like coordination problems in two-earner households in combination with higher commuting costs, as in Van Ommeren 1996).

Thus, because of the theoretical ambiguity, the relationship between home-ownership and labour mobility and unemployment is mainly an empirical issue. Unfortunately, macroeconomic and microeconomic empirical analyses show contradicting outcomes. On a macroeconomic level, most contributions show that homeownership increases unemployment. Nickell (1998) analyses the relationship between homeownership and unemployment, using a panel of twenty OECD countries, from 1989 to 1994. With these data, Nickell shows that the unemployment rate is (seemingly) positively correlated with the homeownership rate, with an elasticity of 0.13. Green and Hendershott (2001b) estimate an elasticity of 0.18, using aggregated data for the different states of the United States for the period 1970–90. This estimate is close to the estimate of Oswald (Chapter 2 above), with an elasticity equal to 0.2. He analyses the relationship between homeownership and unemployment, using panel time series data for nineteen OECD countries, from 1960 to 1990. He actually found this relationship not only between countries, but also between the regions of France, Italy, Sweden, Switzerland, the US, and the UK as well. In line with these results, Murphy et al. (2006) show that strong housing market conditions can prevent movement since expensive housing can deter migrants and make commuting more attractive as an alternative to moving residence.

(p.58) Contrary to the findings presented above, several microeconomic contributions show that homeownership actually diminishes the probability of becoming unemployed. Van Leuvensteijn and Koning (Chapter 8 below) and Munch et al. (Chapter 4 below) have analysed the effect of homeownership on job mobility and unemployment for, respectively, the Netherlands and Denmark. They find no effects on job mobility but identify a small negative effect on the probability of acquiring a job outside the local area. Vander Vlist (2001) studied the Dutch situation, and concludes that homeownership has a small positive effect on changing jobs. Barcelo (2003) analysed for five major European countries the effects of homeownership on unemployment and found that owners are more reluctant to move than renters. Interestingly, for the Australian situation, Flatau et al. (2003) found as well that homeowners become less unemployed than renters. However, this accounts only for homeowners with mortgages which supports the hypothesis that workers who are leveraged are more committed to their jobs. Finally, using US household data, Green and Hendershott (2001a) found that unemployed homeowners indeed find jobs at a slower rate than renters, but with an impact of only an eighth of what is found for aggregate data.

Following Dietz and Haurin (2003), the conclusion from the above literature is that the empirical results concerning the effect of homeownership on labour market mobility are ambiguous. It seems that, in general, studies using micro data tend to reject the Oswald hypothesis—i.e. homeownership increases unemployment—while studies using macro data tend to support it. This might point to the existence of a spurious relation or aggregation effects at the macro level or omitted variables at the micro level (as Van Vuuren and Van Leuvensteijn (2007) suggest).

The theoretical literature on the relationship between social renting and labour market behaviour is very much in line with that of homeownership. Like homeowners, social tenants receive an implicit subsidy through lower rental prices and their place on the waiting list. The size of the subsidy depends in most countries (like Denmark and the Netherlands) on the age of the rental home, or the length of the tenure, the location of the residence (rural or urban), and the distance to the city centre. This subsidy or the place on the waiting list is diminished or lost if a social renter moves to another residence. This opportunity cost thus limits the labour mobility of social renters living in residences under rent control.

Basically, there are two strands of literature. The first is primarily a UK literature which reports both a reduced residential mobility of social renters and an increased probabilityof becoming unemployed amongst social renters (Hughes and McCormick 1987; Minford et al. 1988). More recently, Flatau et al. (2003) found as well that social renters (and those living rent-free) are more likely than private renters to become unemployed.

(p.59) The second is a Danish literature which focuses on rent control. Munch and Svarer (2002) show that tenancy mobility is severely reduced by rent control. For a typical household in the private rental sector, tenancy duration is found to be six years longer if the apartment belongs to the 10 per cent most regulated units than if it belongs to the 10 per cent least regulated units. Svarer et al. (2005) find that rent control benefits reduce housing mobility and hence labour mobility. Furthermore, they find that renters subject to rent control are more likely to accept job offers in the local market than job offers in the non-local market.

The empirical evidence for social renters and rent control schemes is clearly mixed—at least more than the empirical evidence for homeownership—and probably depends upon country specific effects, such as type of social renting, institutions on the labour and housing market, and the national probability of migrating across regional labour markets.

3.3. Data and stylized facts

The dataset used in this chapter is derived from the European Community Household Panel (ECHP). The ECHP survey is based on a standardized questionnaire that involves annual interviewing of a representative panel of households and individuals in various European countries.2 The questionnaire covers a wide range of topics like income, health, education, housing, demographics, and employment characteristics, which makes the database especially suitable for a micro-econometric analysis of differences in residential and labour mobility between countries. The ECHP covers the period 1994 to 2001. In the first wave, i.e. in 1994, a sample of some 60,500 nationally representative households were interviewed in the twelve member states at that juncture, which equals approximately 130,000 adults aged 16 years and over. Austria, Finland, and Sweden joined the project in respectively 1995, 1996, and 1997.3

The ECHP accommodates several types of residential ownership: namely, homeownership, private and social renting, and accommodation that is provided rent-free (occupants do not have to pay rent; this happens predominantly when renting from family and friends). To be more precise, homeownership is defined in the ECHP as owner-occupied housing, and social renting is defined as when the accommodation is rented from a public, municipal, voluntary, or non-profit agency.4

Table 3.1 looks into the distribution of the various types of residential ownership across the various European countries in the ECHP. The percentages have been calculated by summing up the residential ownership types of all observations. Note that we thus assume that there is no selection in attrition from the panel. In other words, social renters and homeowners have the same (p.60) probability of leaving the panel. Because the Swedish Living Conditions survey does not provide information about specific characteristics when the accommodation is rented, we leave them out of the analysis, although we include them in the table for completeness.

Table 3.1. Percentage of homeowners and various types of renters across Europe during 1994–2001

 

Homeowners

Renting

 

 

Private

Social

Rent-free

Germany

0.44

0.32

0.21

0.03

Netherlands

0.61

0.04

0.34

0.01

France

0.63

0.18

0.15

0.05

Austria

0.68

0.12

0.14

0.06

Sweden

0.69

 

 

 

Denmark

0.71

0.12

0.17

0.00

Portugal

0.72

0.14

0.04

0.10

UK

0.74

0.06

0.18

0.02

Belgium

0.74

0.17

0.06

0.03

Luxembourg

0.76

0.21

0.00

0.03

Finland

0.77

0.09

0.12

0.01

Italy

0.78

0.11

0.06

0.06

Greece

0.83

0.13

0.01

0.03

Spain

0.84

0.08

0.02

0.05

Ireland

0.88

0.03

0.08

0.01

Total

0.71

0.14

0.11

0.04

Clearly, owner-occupied housing is the dominant form of housing across Europe. More than 70 per cent of the households in our sample report that they own their home. The percentage homeownership is smallest in Germany and highest in Spain and Ireland. However, during our sample years percentage homeownership has grown in all countries (for some countries, such as France and Denmark, even by more than 10 per cent).

Renting is divided into private, social renting, and rent-free renting, and the shares of various renting types differs considerably across Europe. Social renting is most prevalent in the Netherlands, Germany, the UK, Denmark, France, and Austria and seems to be completely absent in Greece, Spain, and Luxembourg—and to a lesser extent in Portugal and Italy. Obviously, the number of social renters in a country coincides with the geographical location (the north-western part of Europe) and the pervasiveness of the welfare state.

With its average of 4 per cent, rent-free accommodation seems to be a relatively minor issue in Europe. It occurs mostly in countries such as Portugal, Austria, and Italy, and to a certain degree it seems to correlate negatively with the presence of social renting within a country.

It is now insightful to look at the country-specific residential mobility for each of the various types of housing. We specify residential mobility here as (p.61) one when a household has moved residence in the year prior to the interview and average this over countries. This residential mobility can be seen as a proxy for a household’s probability of moving, but is most likely an underestimation because households can move residence multiple times during a year. A household’s probability of moving is not identical to an individual’s probability of moving—especially because most individuals’ residential moves coincide with a change in household composition. Table 3.2 presents residential mobility across Europe.

Table 3.2. Percentage of households that moved residence in the year prior to the interview

 

Total

Type of housing

 

 

Owner

Social renter

Private renter

Ireland

0.02

0.01

0.03

0.26

Austria

0.03

0.01

0.04

0.08

Greece

0.03

0.02

0.09

0.10

Italy

0.03

0.02

0.02

0.08

Portugal

0.04

0.03

0.05

0.06

Spain

0.04

0.03

0.06

0.12

Netherlands

0.05

0.05

0.06

0.12

Belgium

0.06

0.03

0.09

0.16

Sweden

0.06

0.03

 

 

Luxembourg

0.06

0.04

0.08

0.12

Germany

0.07

0.03

0.07

0.11

France

0.08

0.04

0.10

0.21

UK

0.08

0.06

0.09

0.36

Denmark

0.10

0.06

0.15

0.26

Finland

0.10

0.05

0.25

0.34

Total

0.05

0.03

0.08

0.14

Clearly, European private renters move residence most. And European homeowners move residence the least of all. So, countries that have high percentages of homeowners in Table 3.1, such as Ireland, Spain, Greece, and Italy, have low percentages of households moves. Households in Finland, Denmark, the UK, and France move the most, which clearly leads to a North–South divide again in terms of residential mobility.

Private renters move much more than homeowners (up to four to five times). Especially private renters in the UK, Ireland, Denmark, and Finland are mobile. However, social renters move much less than their private counterparts. Indeed, although social renters move more than twice as much as homeowners, they also move almost half as much as private renters.

Obviously, those are descriptive statistics and do not reveal much about possible causal impacts of transaction costs of owner-occupied or socially rented houses, especially because two possible sources of selection effects are present. First of all, households may select themselves out over types (p.62) of housing. Individuals who expect to settle down, such as to start families, may opt for homeownership, while people who expect large life-cycle changes, such as students or singles, may opt for private renting. And, of course, those individuals who (initially) need housing assistance choose social renting if possible. However, one might argue that this selection effect is partly caused by the existence of transaction costs. If households settle down, they are willing to pay large sunk costs as investments in their housing career. Households that are uncertain about their future are not prepared to invest those costs.

A second source of selection may be attributed to the accommodation itself. It might well be that the characteristics of owner-occupied houses are different from those of social or private rented houses. Differences may occur in quality, neighbourhood, and (lot) size of the house. This may result in larger attachments to owner-occupied housing than to private rented housing.

As argued above, and as Figure 3.1 shows clearly, homeowners do face large transaction costs when buying a residence. These transaction costs consist not only of taxation on moving residence, but also of real estate agent fees, notary fees, and the like. Especially, homeowners in Italy, Belgium, Portugal, and Greece seem to have paid large sunk costs to buy their home. This surely makes them less mobile than if they had rented privately.

                      Homeownership, Social Renting, and Labour Mobility across Europe

Figure 3.1. Taxation on moving residence and real estate agent fees for several European countries. Source: Belot and Ederveen (2005)

(p.63) Social renters do not face transaction costs as such, but instead a loss of subsidies or a loss of the option on subsidies when on a waiting list. Thus, social renting normally involves some kind of subsidy associated with the dwelling and not with the household. Households in general are not inclined to move when this entails a loss of (the option on) their subsidy. However, this loss of (the option on) subsidy usually occurs only when moving between municipalities. This is in contrast to formal transaction costs for homeowners, which always occur when moving residence. Obviously, because institutions, finance of social dwellings, and supervision differ across European countries, the impact of social renting on labour mobility should be researched in a country-specific manner (for a detailed analysis of the institutional setting of social housing in the Netherlands, France, Belgium, Germany, the UK, Sweden, and Denmark, see Priemus and Boelhouwer 1999).

3.4. Estimation strategy

To study the impact of private renting, social renting, and homeownership on cross-country labour market mobility differences, we focus on the probability of leaving a job for various exit destinations. We assume that workers only end their current job for three possible reasons. First, they may find another job (whether more suited to the worker or not). Secondly, workers may become unemployed. And thirdly, workers may leave the labour force altogether because of retirement, to raise a child, to look after disabled family members, for study purposes, and so on. If labour mobility is hampered, this may show up in two ways. First, a worker may remain longer in his or her current job and move at a slower rate to a new one. This indicates that a worker is hampered in his or her upward career mobility. Secondly, a worker may end up faster in unemployment or leave the labour force faster. This happens when workers face difficulties in finding a job in the local labour market and when it is too costly to move to a different regional labour market with more attractive job opportunities.

The model we construct consists of a competing risk duration framework for the various exit rates of employment in combination with a discrete choice model for the probability of buying or renting—privately or socially—a house. The fundamental assumption we make is that the decision about the type of housing is correlated with labour market behaviour through observed and unobserved factors. First, as mentioned above, we allow labour market mobility to be directly related to housing tenure, by incorporating type of housing in the job duration model. This entails a direct test of whether homeowners or social renters are more or less mobile on the labour market relative to renters.

Simultaneously, we allow for unobserved heterogeneity to control for those unobserved factors that drive both job mobility and housing tenure. In our (p.64) case, unobserved heterogeneity may occur because of unobserved skills and job commitment. For example, workers with short-term contracts are less likely to buy a house compared to workers with tenure. And workers who have diminished prospects on the labour market may end up earlier in social housing.

The modelling approach we adopt in this paper is a direct extension of De Graaff and Van Leuvensteijn (2007), and closely resembles that of Van Leuvensteijn and Koning (Chapter 8 below) and Munch et al. (Chapter 4 below) as well, so little attention is spent on technical details. The first subsection deals with the econometric model. Subsequently, we pay some attention to the issue of identification. The last subsection combines all components and specifies the complete likelihood function to be estimated.

3.4.1. The econometric model

To model the probability of leaving a job we use a duration analysis framework. The basic concept in duration analysis is the hazard rate θ, which is defined here as the rate at which workers leave their current job in the time interval [T, T + dt] given that these workers occupy their job at least up to T. The probability that someone leaves employment within an interval dt after t can be denoted as Pr(T 〈 t 〈 T + dt|t≥T) (see e.g. Lancaster 1990). Dividing this probability by dt, we get the average probability of leaving employment per unit time period:

(3.1)                       Homeownership, Social Renting, and Labour Mobility across Europe

where the subscript b ∈{e, u, o} indicates the exit destination, which in our case can be employment (e), unemployment (u), and out of the labour force (o). Note that if dt → 0, we have an instantaneous rate of leaving per unit time period at t.

We use a proportional hazard rate specification, indicating that we assume that the impact of individual characteristics is proportional to the impact of the elapsed time of the job spell. Further, each destination-specific hazard is a function of a set of observed characteristics, such as age, sex, being married, and educational attainment, which may vary over time, X t, a time-varying indicator for housing tenure; h t, a function which measures duration dependence for a specific exit destination; λ b(t), and unobserved characteristics, v b. Thus, the hazard rate of a specific destination may be written as:

(3.2)                       Homeownership, Social Renting, and Labour Mobility across Europe

Often, λ b(t) is also referred to as the baseline hazard. We adopt here a non-parametric flexible specification in the form of a piecewise constant specification. So, duration dependence is assumed to be constant within duration intervals.

(p.65) We assume that the dichotomous—homeowners versus renters and social versus private renters—housing tenure variable h t conforms to the following logit specification:

(3.3)                       Homeownership, Social Renting, and Labour Mobility across Europe

where h is 1, if the worker owns or socially rents his current residence, and zero if the worker rents it socially or rents it privately, respectively. Y t denotes a set of variables that characterizes the particular choice of housing tenure. Note that the set of variables Y t may partly overlap with the set of variables X t, which is used to model job duration spells. Finally, to account for unobserved heterogeneity, we incorporate an additional unobserved random component, denoted by μ h.

In contrast to regression models, unobserved heterogeneity causes an estimation bias in duration modelling. Therefore, several modelling approaches have been developed to control for unobserved heterogeneity. We adopt here the often used non-parametric approach proposed by Heckman and Singer (1984). Basically, this boils down to the assumption of a discrete distribution, denoted G, with a pre-specified number (say K) of mass points. In addition, we assume v e, v u, v 0, and μ h to be correlated. Together with K mass points, this leaves us with 4K possible combinations between the K mass points, each with a separate probability, which has to be estimated simultaneously. When using constant terms, the distribution is identified by normalizing the first point of support to {0,0,0,0}, so that the number of mass points to be estimated reduces to (K–1) × 4.5

As shown above, our model consists of two parts: the housing model and the job duration model. If not for the correlation between the unobserved heterogeneity components, these two parts can be estimated separately. Allowing for correlation creates a mixture model which has to be integrated over the entire distribution of unobserved variables, G{v e, v u, v 0, μ h} (see Van den Berg 2001 for more details on the application of mixture distributions in duration models).

3.4.2. Identification

A key issue in the literature on the relationship between housing tenure and labour market mobility is the identification of the causal effect. Housing tenure may cause a change in labour market mobility, but the reverse relation is, a priori, just as likely. Those workers who have good prospects of long job spells (or lower probabilities of ending up unemployed) are the ones most likely to buy a house, and those workers who have high probabilities of entering (p.66) unemployment are the ones most likely to opt for social housing. The literature distinguishes two approaches to deal with this endogeneity. The first uses instrumental variables, where variables that affect housing tenure but not labour market mobility are incorporated in the housing model to control for endogeneity. Van Leuvensteijn and Koning (Chapter 8 below) proposed using regional homeownership as an instrument, while Munch et al. (Chapter 4 below) used homeownership of the parents in 1980 and the proportion of homeowners in the municipality where the individual was born. Usually, however, the impact of these instrumental variables is rather low, indicating that these models are already fairly well identified or that the performance of the chosen instruments is rather weak.

We choose a second approach by using multiple spells for identification (cf. Van Vuuren and Van Leuvensteijn 2007 and Munch et al. (Chapter 4 below))—the latter pay much attention to the intuition for this strategy (see Abbring and Van den Berg 2003 for a formal argumentation for this identification strategy). To summarize their arguments, it is not difficult to see that making repeated observations of one individual removes all interpersonal variation.6 Thus, if there are multiple job spells available to a specific individual, and if her housing tenure status varies as well over these spells, then the effect of housing tenure on labour market mobility is theoretically identified.7 Identification is then based on a sub-sample with multiple spells and changes in housing tenure status, where the existence of multiple spells ensures that the unobserved heterogeneity components capture the ‘within person’ effects (Chapter 4 below).

3.4.3. The log-likelihood function

To construct the log-likelihood, we introduce some additional notation. As in Lancaster (1990), let there be B binary destination vectors d b, where d b is such that there is a transition to state b and zero otherwise. Because we do not observe all job duration spells as ending, we model right-censored job duration spells as well. We do this by treating right-censoring theoretically as an additional dummy state. Thus, the set of possible destination vectors B now consists of employment, unemployment, out of the labour force, and censoring.8 Thus, given that individuals have an elapsed duration time T and job exit destination b, and conditional on their observed characteristics, housing tenure, and mass point vb, the log-likelihood for job durations may be written as:

(3.4)                       Homeownership, Social Renting, and Labour Mobility across Europe

where φ b is shorthand notation for the parameter vector {βb, λb(t), γb}. Note that the first part of Eqn. (3.4) displays the hazard rate of the transition to (p.67) destination b, while the second part denotes the probability of survival of the job spell until time T.

The log-likelihood of choice of housing tenure h t during the total length of the job spell conditional on the observed characteristics and country-specific housing market variables follows immediately from the logit Eqn. (3.3), and is given by:

(3.5)                       Homeownership, Social Renting, and Labour Mobility across Europe

where φ h denotes the parameter vector {β h, δ h}. The joint log-likelihood is now formed by multiplying the likelihoods of (3.4) and (3.5)—given the discrete unobserved heterogeneity distribution—and integrating over the entire distribution of mass points G{v e, v u, v 0, µ h}. Allowing for the presence of multiple job spells, the joint log-likelihood for the contribution of an individual i can be written as:

(3.6)                       Homeownership, Social Renting, and Labour Mobility across Europe

where, j (j ∈ {1,…, N j }) stands for spell j, and N j for the total number of job spells of individual i. The log-likelihood in (3.6) basically states that the log-likelihood of job duration, as in (3.4), and the log-likelihood of housing tenure, as in (3.5), has to be integrated over the distribution of mass points, which raises an additional difficulty, in that we have to optimize not over a set of parameters, but over a probability distribution as well. To do this, we apply an expectation-maximization (EM) algorithm to solve for the parameters of Eqn. (3.6) that we are interested in (see De Graaff and Van Leuvensteijn 2007 for more details).

3.5. Estimation results

We have monthly information about each worker’s status and yearly information on all other characteristics. Thus, job tenure is measured in months, and housing tenure in years. In terms of exit destinations, we denote a job move when a worker changes job or apprenticeship, a job exit to unemployment only when the next activity is labelled unemployment, and entry into the out-of-labour force when a worker becomes retired, spends his or her time in (unpaid) housework activities, is doing community or military service, or ends up in other activities that are economically inactive.

(p.68) We use individual and household characteristics to control as much as possible for individual, household, and life-cycle effects that might influence the event of leaving the current job spell apart from residential mobility effects. First, we use age cohort dummies as age controls for life-cycle effects that might cause, e.g., individuals to enter an out-of-labour force status. Secondly, gender is included to control for the fact that females have a higher probability of looking after children and thus may leave a current job spell faster to become economically inactive. The same accounts for the dummy variables that control for the presence of children of different ages within the household. We include education—measured as low, medium, and high—to control for the fact that more educated workers earn higher wages and therefore show higher homeownership rates. Here, medium education denotes secondary level, and high education a university degree or above. Having a partner in the household and whether the partner earns an income is included, as these households usually have higher probabilities of owning a house as well. Finally, housing tenure is included to test the impact of reduced housing mobility at an individual level.

Identification is done by using the availability of multiple spells. A fair amount of multiple spells is present in the data. More than 30 per cent of the observed workers show two or more employment spells.

All variables are measured at the moment the worker leaves his or her current job. As mentioned above, to control for duration dependence, we adopt a non-parametric flexible specification. Here, duration dependence is assumed to be constant within the following duration intervals: within one year, between one and three years, between three and five years, and above five years. A specific approach to incorporate such a non-parametric specification is shown in Lancaster (1990).

Finally, we set the number of mass points (K) at two, which—in theory—leaves us with sixteen probabilities to be estimated. However, experiments with sub-samples show that a smaller number of these probabilities make computation not only considerably faster, but give (almost) the same estimation results as well. We therefore only use four of these probabilities.

3.5.1. The impact of homeownership

To analyse the impact of homeownership on labour mobility, we distinguish between homeowners and private renters. Because our dataset contains in total more than 200,000 employment spells, we use a 10 per cent random sample to reduce computing time, resulting in a total of 19,482 employment spells. Table 3.3 displays the estimation results of the joint model of being a homeowner and leaving the current employment spell for three possible exit destinations.

(p.69)

Table 3.3. Joint estimation of owning a home and competing risk model (standard errors given in parentheses)

Variable

Homeowner

θb(t)

 

 

Employment

Unemployment

Non-participation

Constant

−0.300 (0.05)

 

 

 

 Age dummies (baseline: age 〈 25)

Age 25–34

−0.190 (0.05)

−0.569 (0.04)

0.100 (0.06)

−0.864 (0.05)

Age 35–44

0.254 (0.06)

−0.974 (0.05)

−0.197 (0.07)

−1.674 (0.07)

Age 〉 45

1.090 (0.06)

−1.089 (0.04)

−0.245 (0.07)

−0.973 (0.06)

Female

−0.164 (0.04)

0.254 (0.03)

0.163 (0.04)

0.617 (0.04)

 Education dummies (baseline: education = low)

Education medium

−0.047 (0.04)

−0.012 (0.03)

−0.261 (0.04)

−0.019 (0.04)

Education high

0.050 (0.05)

−0.254 (0.04)

−0.557 (0.06)

−0.008 (0.05)

Spouse employed

0.287 (0.05)

−0.137 (0.04)

−0.157 (0.05)

−0.213 (0.05)

Living with partner

−0.680 (0.06)

−0.119 (0.04)

−0.275 (0.06)

0.051 (0.06)

 Children within the household (baseline: no children ≤ 18)

Children 11

0.345 (0.04)

−0.001 (0.03)

0.052 (0.05)

0.082 (0.05)

Children 12–15

0.420 (0.07)

−0.015 (0.05)

0.252 (0.06)

−0.031 (0.06)

Children 16–18

0.759 (0.07)

0.230 (0.04)

0.162 (0.06)

0.261 (0.06)

Homeowner

 

−0.219 (0.03)

−0.275 (0.04)

0.027 (0.04)

Baseline hazard

 

 

 

 

0–1 year

 

−2.951 (0.04)

−4.048 (0.06)

−4.037 (0.06)

1–3 years

 

−4.985 (0.05)

−5.862 (0.08)

−5.623 (0.07)

3–5 years

 

−4.994 (0.07)

−6.160 (0.10)

−5.670 (0.08)

〉5 years

 

−5.007 (0.06)

−5.557 (0.08)

−5.287 (0.08)

Unobserved heterogeneity distribution

Mass point

1.820 (0.04)

0.646 (0.03)

0.701 (0.04)

0.448 (0.04)

Probabilities

 

 

 

Pr(G = {0,0,0,0})

0.192

 

 

 

Pr(G = {1,0,0,0})

0.588

 

 

 

Pr(G = {0,1,1,1})

0.073

 

 

 

Pr(G = {1,1,1,1})

0.147

 

 

 

Mean log-likelihood

−3.690

 

 

 

No. of spells

19,482

 

 

 

Most estimated coefficients of the competing hazard rate model are significant, conforming with intuition and previous research. Our main variable of interest, whether someone is a homeowner, reduces the probability of changing jobs significantly (risk reduction 20 per cent, which can be calculated as exp(-0.219)). Further, homeownership ensures that workers face smaller probabilities of becoming unemployed (risk reduction of 24 per cent) and a (insignificant) higher risk of leaving the labour force (risk increase 3 per cent). These results are very similar to outcomes of previous microeconomic studies (see e.g. Van Leuvensteijn and Koning (Chapter 8 below) and Munch et al. (Chapter 4 below)).9 Basically, this confirms the hypotheses of, e.g., Dietz and Haurin (2003) and Van Leuvensteijn and Koning (Chapter 8 below), that homeowners have larger job commitment than private renters. This can partly be explained by the substantial monetary transaction costs (p.70) when forced to sell their house because of, e.g., unemployment (De Graaff and Van Leuvensteijn 2007).

The remaining coefficients for the housing model are as might be expected. Age tends to increase the probability of homeownership, just as being male, having received higher education, and having an employed spouse. Living together with an (unemployed) partner reduces homeownership, while having (older) children increases this probability again. The latter is probably a proxy for a life-cycle effect, where household heads have a smaller probability of being a homeowner when households are relatively recently formed.

The hazard rate for job-to-job transitions is found to decline with age, which is understandable, because younger workers are more mobile on the labour market searching for a suitable job. The same accounts for the probability of entering unemployment. However, hazard rates into non-participation seem to rise again for older workers (with age above 45). Females have in general higher hazard rates out of employment than males; in particular, females have a high risk of ending up as non-participants in the labour market (their risk of non-participation is about 85 per cent higher than that for males). Higher educational levels result in smaller hazards with respect to job changes or unemployment, although education does not affect the probability of becoming a non-participant significantly. Having an employed spouse or living with a partner diminishes the risk of changing jobs, becoming unemployed, or leaving the labour force altogether, although living with a partner does not significantly affect the latter. Finally, having older children increases the probability of changing jobs, becoming unemployed, or leaving the labour force. This might again point to a life-cycle effect.

The piecewise constant specification for duration dependence gives consistent and intuitively appealing outcomes. After the first year, all hazard rates out of employment drop significantly and continue falling with the worker’s job duration.

Finally, we turn to the unobserved heterogeneity distribution. All mass points are positive and very significant. Most probability is assigned to the combination with a low exit rate out of current employment and a high probability of owning a home. Interestingly, only a small part of our population—around 22 per cent—face higher exit rates out of their current job. The current estimation, where four segments are used, show that two segments account for almost 80 per cent of all individuals.10 The actual allocation of the probabilities to the segments depends upon the number of segments and the initial starting point of the algorithm, but experiments show that all combinations converge to the same log-likelihood, and that the largest group is usually the segment that contains all favourable mass points (thus the one denoted here as G = {1, 0, 0, 0}).

As mentioned above, European countries differ widely in their institutional setting, the functioning of the labour and housing market, and the general tendency of the population to move house. Therefore, it is insightful to look (p.71) at the impact of owning a home on labour market behaviour for each European country separately. Figure 3.2 displays for each country in our dataset the impact that homeownership has on the probability of leaving the current job for another job, unemployment, or leaving the labour market altogether.

                      Homeownership, Social Renting, and Labour Mobility across Europe

Figure 3.2. Impact of homeownership on labour mobility across Europe

Unsurprisingly, there is a large amount of country-specific variation present. This is primarily caused by the countries’ institutional differences in both the labour and the housing market. Overall, homeownership seems to diminish the probability of leaving the current job. The notable exceptions—Ireland, Greece, Spain, and Finland—all occur in countries with high percentages of homeownership. And for Greece, Spain, and Finland it truly does seem that homeownership does not impede residential mobility such that it hampers labour market behaviour; homeowners enter unemployment faster than renters. Thus, in Spain, Greece, and Finland private renters are doing better on the labour market than homeowners.

However, in most other European countries, such as the UK, the Netherlands, Germany, Belgium, France, and Portugal, homeowners seem to stay longer in their current jobs than non-homeowners. Assuming that our analysis has corrected sufficiently for self-selecting, then at least for these countries homeownership causes a higher job commitment than non-homeownership. A possible reason for this is that homeowners grow attached to their homes (and the surrounding neighbourhood) through both monetary and psychological investments.

The next subsection presents the empirical results for social renters and investigates whether being tied to a social home causes social renters as well to increase their job commitment.

(p.72) 3.5.2. The impact of social renting

To analyse the impact of social renting on labour mobility, we distinguish in this subsection between private and social renters. There are in total 72,292 employment spells in our database which can be attributed to renters. Analogously to Table 3.3, Table 3.4 displays the estimation results of the joint model of being a social renter and leaving the current employment spell for three possible exit destinations.

Table 3.4. Joint estimation of social renting and competing risk model (standard errors given in parentheses)

Variable

Social renter

θb(t)

 

 

Employment

Unemployment

Non-participation

Constant

 −0.978 (0.02)

 

 

 

Age dummies (baseline: age 〈 25)

Age 25–34

−0.138 (0.02)

−0.630 (0.02)

0.066 (0.03)

−0.829 (0.03)

Age 35–44

0.114 (0.03)

−1.011 (0.02)

−0.068 (0.03)

−1.407 (0.03)

Age 〉 45

0.484 (0.03)

−1.285 (0.02)

−0.339 (0.03)

−0.873 (0.03)

Female

0.016 (0.02)

0.067 (0.01)

0.001 (0.02)

0.469 (0.02)

 Education dummies (baseline: education = low)

Education medium

−0.212 (0.02)

0.060 (0.01)

−0.130 (0.02)

0.167 (0.02)

Education high

−0.629 (0.03)

−0.223 (0.02)

−0.391 (0.03)

0.295 (0.03)

Spouse employed

−0.219 (0.02)

−0.083 (0.02)

−0.188 (0.03)

−0.118 (0.03)

Living with partner

−0.152 (0.02)

−0.030 (0.02)

−0.014 (0.03)

−0.006 (0.03)

 Children within the household (baseline: no children ≤ 18)

Children 11

0.507 (0.02)

−0.012 (0.02)

0.171 (0.02)

0.132 (0.02)

Children 12–15

0.428 (0.03)

0.029 (0.02)

0.072 (0.04)

−0.159 (0.04)

Children 16–18

0.145 (0.03)

0.202 (0.02)

0.158 (0.04)

0.203 (0.03)

Social renter

 

0.034 (0.01)

0.113 (0.02)

−0.022 (0.02)

Baseline Hazard

0–1 year

 

−2.732 (0.02)

−4.120 (0.03)

−3.987 (0.03)

1–3 years

 

−4.806 (0.02)

−5.962 (0.04)

−5.736 (0.04)

3–5 years

 

−4.888 (0.03)

−6.080 (0.05)

−5.743 (0.04)

〉5 years

 

−4.838 (0.03)

−5.533 (0.04)

−5.281 (0.04)

Unobserved heterogeneity distribution

Mass point

1.534 (0.02)

0.109 (0.01)

0.094 (0.02)

0.07 (0.02)

Probabilities

Pr(G = {0,0,0,0})

0.414

 

 

 

Pr(G = {1,0,0,0})

0.215

 

 

 

Pr(G = {0,1,1,1})

0.195

 

 

 

Pr(G = {1,1,1,1})

0.175

 

 

 

Mean log-likelihood

−3.977

 

 

 

No. of spells

72,292

 

 

 

Again, most coefficients from the social renting and competing risk model are significant and conform to intuition. Most importantly, whether someone is a social renter seems to increase the probability of leaving the current job (versus private renters). Social renters have about a 3 per cent higher chance of leaving the current job for another job and a 12 per cent higher chance of (p.73) ending up in unemployment. The lower probability of leaving the labour force is not significant. Thus, the mobility costs associated with social renting seem to hamper labour market performance in terms of the probability of leaving employment. Job-to-job mobility, however, is higher for social renters than for private renters.

The other coefficients for the social renting model are as might be expected. Being a low-educated, older, single person with children increases the probability of ending up in the social renting sector. Living with a working partner increases the chance of renting in the private sector, just as being more highly educated does.

As might be expected, the coefficients for the competing risk model are, in a qualitative sense at least, rather similar to those in Table 3.3, although the absolute values change of course because of the different sample selection (only renters). The unobserved heterogeneity distribution is, however, rather different. Where the results in Table 3.3 show a clear, almost bimodal distribution, the results in Table 3.4 show a more homogeneous distribution. Especially, there does not seem to be much unobserved heterogeneity present in the labour market. There is a clear group who seem to leave the current job 10 per cent faster for whatever reason; but this is not as strong as presented in Table 3.3.

This does not entail that results are homogeneous across countries as well. Again, because European countries differ widely in their institutional setting, we distinguish the impact of social renting on each country separately in our dataset in Figure 3.3.

                      Homeownership, Social Renting, and Labour Mobility across Europe

Figure 3.3. Impact of social renting on labour mobility across Europe

(p.74) When comparing Figure 3.3 with Figure 3.2 it is obvious that social renting has a less homogeneous impact on labour market behaviour then homeownership. In general, social renting seems to increase the probability of changing jobs, especially for countries such as Germany, Denmark, Ireland, Italy, and France. In countries such as Belgium, France, the UK, and to a lesser extent Portugal, social renting causes employees to change jobs less frequently. More distinctly, however, social renters have higher probabilities of entering unemployment than private renters, except for countries such as the Netherlands, Austria, and Luxembourg. Social renting in these countries even decreases the overall probability of ending a job spell. Note that of these countries there is a sizeable social renting sector only in the Netherlands and Austria.

Thus, overall one can conclude that social renting does not increase job commitment, or at least not to the extent that homeownership does. This might be because social renters do not invest—both monetarily and psychologically—as much as homeowners in their residence. Usually there is a landlord who maintains the dwelling (see as well the appendix for Western European cross-country differences in social renting institutions). This might cause social renting to hamper labour market performance in the sense that a larger social renting sector increases aggregate unemployment.

3.6. Conclusion

This chapter has dealt with the impact of homeownership and social renting on labour market behaviour. We have argued that homeownership and social renting both increase transaction costs and therefore should both have an impact on labour market behaviour. To what extent social renting and home-ownership increase transaction costs or affect labour market behaviour differs widely across European countries. Thus, the main goal of this chapter was to distinguish between various European countries in our analysis, to control for the various national institutional settings.

In general, residential transaction costs hamper individual mobility because they limit individual choices. However, it is difficult to make ex ante predictions about the impact on (un)employment probabilities. For higher residential transaction costs may as well increase job commitment or lower reservation wages instead of increasing the probability on, e.g., unemployment.

Indeed, previous research such as that of Van Leuvensteijn and Koning (Chapter 8 below) and Munch et al. (Chapter 4 below) find that home-ownership benefits individual labour market behaviour in the sense that homeowners change jobs less frequently but leave unemployment faster. This chapter confirms those findings for most countries in Europe. Exceptions are countries with a high level of aggregate homeownership, such as Spain and Greece.

(p.75) The results for workers in the social renting sector are less clear, in that they vary much more across European countries. However, overall we can say that social renters display a worse labour market performance than private renters, except for countries such as the Netherlands (note that the amount of private renting is rather low in the Netherlands compared with social renting), Luxembourg, and Austria.

Therefore, we may conclude that although both social renting and home-ownership increase residential transaction costs, only social renting increases individuals’ probabilities of entering unemployment. A possible reason might be that homeowners invest more then social renters—both monetarily and psychologically—in their houses and neighbourhoods. If true, then one might also hypothesize that homeowners will make larger commutes than social renters: for they may be willing to sacrifice leisure time to commuting time to avoid paying residential transaction costs. Another consequence would be that homeowners are willing to accept lower wages than private renters, where social renters would be more hesitant to do so. Or, alternatively, homeowners are willing to work more hours for the same (monthly) wage.

Apart from this, it would be interesting as well to look at the impact of transaction costs on the individual outflow of unemployment. Although this has already generated a large literature, it has never been done on a micro level (except for the study of Barcelo 2003) for a variety of European countries. Moreover, the impact of social renting on unemployment deserves more attention, together with the need for more information on the various forms of social renting across Europe and the institutions that drive them.

A.1. Homeownership in European countries

Table 3.5 presents some selected characteristics of homeownership throughout Europe. The percentage of homeowners is the mean of 1994–2001 data found in the European Community Household Panel by Eurostat. Property transfer taxes (usually paid by the buyer), real estate agent fees (these vary somewhat), and mean duration of transactions are taken from Belot and Ederveen (2005).

Typical loan-to-value ratios and average typical terms are found in IMF (2008) and are more recent. The former figure is known to have increased in most countries in the last decade as mortgage requirements became less strict.

Finally, the last two columns represent the possibility of interest tax deduction and a taxation on the property itself (Wolswijk 2005). Both characteristics are known to vary quite a bit, as legislation changed drastically in most European countries during the 1990s and 2000s. For example, Sweden abolished interest tax deduction at the marginal tariff (of 70 per cent) but only allowed deduction at the tariff of 30 per cent in 1991. A similar change occurred in Finland in 1993.

A.2. Social renting in European countries

Table 3.6 presents some selected characteristics of social renting in Europe. Because social renting and its associated institutions vary significantly across Europe, a direct comparison of the characteristics in Table 3.6 should be made very carefully. In this table social renting is defined as renting from a public, municipal, voluntary, or non-profit agency.11 Unfortunately, we only have information about social renting in a limited number of countries. As source of information we have used Eurostat (ECHP) and Whitehead and Scanlon (2007).

As Table 3.6 clearly shows, social renting varies hugely across European countries. Social renting seems to be especially prevalent in the north-western part of Europe. Moreover, who may rent where seems to be especially decided at a local (municipal) level. If taken into account that most people are eligible for social housing, that rents usually do not increase when income increases, and that for most social housing there is a significant waiting list (Whitehead and Scanlon 2007), then—clearly—large transaction costs can be associated with social renting.

(p.77)

Table 3.5. Characteristics of homeownership in Europe in the late 1990s and early 2000s

Country

Percentage homeowners (mean 1994–2001)

Property transfer tax (%)

Real estate agent fees (%)

Duration of transaction (weeks)

Typical loan-to-value ratio (%)

Average typical term (years)

Interest tax deductibility (yes/no)

Wealth tax on housing (yes/no)

Germany

0.44

4.3–4.7

3–5

 

70

25

no

no

Netherlands

0.61

6

1.5–3

4–6

90

30

yes

no

France

0.63

4.89

5–10

6–13

75

15

no

yes

Austria

0.68

4.5

3–5

 

60

25

no

no

Sweden

0.69

1.5

0.5–5

4–6

80

25

yes

yes

Denmark

0.71

0.6

3

6

80

30

yes

no

Portugal

0.72

10

3–5

6–8

 

 

no

no

UK

0.74

1

1–3

12

75

25

no

no

Belgium

0.74

12.5

3–5

 

83

25

yes

no

Luxembourg

0.76

9

3

 

 

 

yes

yes

Finland

0.77

4

4–8

 

75

17

yes

yes

Italy

0.78

11

3–8

 

80

15

yes

no

Greece

0.83

9.5–11.5

4

 

75

17

yes

no

Spain

0.84

6.5

3–5

 

70

20

no

yes

Ireland

0.88

6

1.75

 

70

20

no

no

(p.78)

Table 3.6. Social renting in Europe in the late 1990s and early 2000s

Country

Percentage social renting (%)

Ownership municipality/housing association (%)

Eligible at entry (% of population)

Happens when income exceeds limit

Determines eligibility

Assigns household

Germany

0.21

 

20

rent unchanged

local government

landlord

Netherlands

0.34

1

40

rent unchanged

landlord

landlord

France

0.15

36

30–80

small supplement

landlord

landlord

Austria

0.14

40

80–90

rent unchanged

local government and landlord

local government and landlord

Sweden

 

100

100

NA

landlord

landlord

Denmark

0.17

5

100

NA

local government and landlord

local government and landlord

Portugal

0.04

 

 

 

 

 

UK

0.18

54

100

NA

 

 

Belgium

0.06

 

 

 

 

 

Luxembourg

0.00

 

 

 

 

 

Finland

0.12

 

 

 

 

 

Italy

0.06

 

 

 

 

 

Greece

0.01

 

 

 

 

 

Spain

0.02

 

 

 

 

 

Ireland

0.08

88

restricted

rent rises

local government

landlord

References

Bibliography references:

Abbring, J., and Van den Berg, G.(2003), ‘The non-parametric identification of treatment effect in duration models’, Econometrica, 71, 1491–517.

Barcelo, C. (2003) ‘Housing tenure and labour mobility: a comparison across European countries’, CEMFI Working Paper no. 0302.

Belot, M., and Ederveen, S. (2005), ‘Indicators of cultural and institutional barriers in OECD countries’, CPB Memorandum (The Hague).

Blanchard, O. J., and Katz, L. R. (1992), ‘Regional evolutions’, Brookings Papers on Economic Activity, 11, 1–75.

(p.80) De Graaff, T., and Van Leuvensteijn, M. (2007), ‘The impact of housing market institutions on labour mobility: a European cross-country comparison’, CPB Discussion Paper no. 82 (The Hague).

Dietz, R. D., and Haurin, D. R. (2003), ‘The social and private micro-level consequences of homewnership’, Journal of Urban Economics, 54, 401–50.

Dohmen, T.J. (2005), ‘Housing, mobility, and unemployment’, Regional Science and Urban Economics, 35, 305–25. Reproduced as Ch. 1 above.

Flatau, P., Forbes, M. Hendershott, P. H., and Wood, G. (2003), ‘Homeownership and unemployment: the roles of leverage and public housing’, NBER Working Paper 10021.

Green, R. K., and Hendershott, P. H. (2001a), ‘Homeownership and the duration of unemployment: a test of the Oswald hypothesis’, Working Paper, University of Aberdeen.

——(2001b), ‘Homeownership and unemployment in the US’, Urban Studies, 38/9, 1509–20.

Heckman, J. J., and Singer, B. (1984), ‘A method for minimizing the impact of distributional assumptions in econometric models for duration data’, Econometrica, 52, 271–320.

Hughes, G., and McCormick, B. (1987), ‘Housing markets, unemployment and labour market flexibility in the UK’, European Economic Review, 31/3, 615–45.

IMF (2008), World Economic Outlook: Housing and the Business Cycle (Washington: IMF).

Lancaster, T. (1990), The Econometric Analysis of Transition Data (Cambridge: Cambridge University Press).

Minford, P., Ashton, P., and Peel, M. (1988), ‘The effects of housing distortions on unemployment’, Oxford Economic Papers, 40, 322–45.

Munch, J. R., and Svarer, M. (2002), ‘Rent control and tenancy duration’, Journal of Urban Economics, 52, 542–60.

——Rosholm, M., and Svarer, M. (2006), ‘Are homeowners really more unemployed?’, Economic Journal, 116, 991–1013. Reproduced as Ch. 4 below.

Murphy, A., Muellbauer, J., and Cameron, G. (2006), ‘Housing market dynamics and regional migration in Britain’, Nuffield College, UK.

Nickell, S. (1998), ‘Unemployment: questions and some answers’, Economic Journal, 108/448, 802–16.

Oswald, A. J. (1997), ‘The missing piece of the unemployment puzzle: an inaugural lecture’, mimeo, University of Warwick.

——(1999), ‘The housing market and Europe’s unemployment: a non-technical paper’, mimeo, University of Warwick. Reproduced as Ch. 2 above.

Priemus, H., and Boelhouwer, P. (1999), ‘Social housing finance in Europe: trends and opportunities’, Urban Studies, 36, 633–45.

Svarer, M., Rosholm, M., and Munch, J. R. (2005), ‘Rent control and unemployment duration’, Journal of Public Economics, 89, 2165–81.

Van den Berg, G. (2001), ‘Duration models: specification, identification and multiple durations’, in Handbook of Econometrics (Dordrecht: Elsevier), 3381–453.

——and Gorter, C. (1997), ‘Job search and commuting time’, Journal of Business & Economic Statistics, 15, 269–81.

Van der Vlist, A. J. (2001), Residential Mobility and Commuting, Thela Thesis, Tinbergen Institute Research Series 248 (Amsterdam).

(p.81) Van Leuvensteijn, M., and Koning, P. (2004), ‘The effect of homeownership on labor mobility in the Netherlands’, Journal of Urban Economics, 55, 580–96. Reproduced as Ch. 8 below.

Van Ommeren, J. N. (1996), Commuting and Relocation of Jobs and Residences: A Search Perspective, Thela Thesis, Tinbergen Institute Research Series 130 (Amsterdam).

Van Vuuren, A. P., and Van Leuvensteijn, M. (2007), ‘The impact of homeownership on unemployment in the Netherlands’, CPB Discussion Paper no. 86 (The Hague).

Whitehead, C., and Scanlon, K. (2007), Social Housing in Europe (London: LSE).

Wolswijk, G. (2005), ‘On some fiscal effects on mortgage debt growth in the EU’, Working Paper, European Central Bank, no. 526.

Notes:

(1) . For an explanation of the differences between the impact of homeownership on labour market mobility we refer to De Graaff and Van Leuvensteijn (2007).

(2) . The countries involved in the ECHP are Germany, the Netherlands, France, Austria, Sweden, Denmark, Portugal, the UK, Belgium, Luxembourg, Finland, Italy, Greece, Spain, and Ireland.

(3) . The data for Sweden have been derived from the Swedish Living Conditions Survey and transformed into ECHP format.

(4) . The term ‘public housing’ is used as well for this particular tenure type. However, public housing is usually associated with government authorities, whereas social housing can also be provided by other organizations, e.g. by the church.

(5) . Note that this leaves the maximum number of probabilities still to be estimated as 4K.

(6) . However, as one referee rightfully observed, this is true only if unobserved individual heterogeneity is constant over time. Because in our case of job and housing mobility this assumption may be a bit strong, we incorporate as many variables as we can in models (3.2) and (3.3) that might reflect changes in preference structures, i.e. because of life-cycle effects.

(7) . That is, apart from possible changes in her preference structure, which may well arise if, e.g., life-cycle effects are not properly accounted for by the exogeneous variables.

(8) . To avoid confusion, we do not model censoring as another competing risk. In other words, transitions to state b do not include censoring, while the destination vector d b does include censoring.

(9) . The cited risk into the out-of-labour force is somewhat inconsistent across studies. In our case, the coefficient is (marginally) significantly positive. Van Leuvensteijn and Koning (Ch. 8 below), e.g., find for the Netherlands that the coefficient is insignificantly positive as well. In any case, the coefficient is small, pointing to the limited effect of homeownership on leaving the labour force.

(10) . Actually, estimations with only two segments—one with all mass points and one with no mass points—result in almost the same log-likelihood and coefficient estimates.

(11) . Opposed to definitions that are concerned with who constructed the dwelling, whether rents are below market levels, or what the relevant funding or income streams are.