Home ownership under changing labour and housing market conditions: tenure preferences and outcomes among freelancers and flex workers

Abstract

Increasingly, policy-makers regard flexible labour as a condition for a well-functioning economy, while they also tend to regard home ownership as the superior tenure. These two goals appear to be contradictory, as mortgage lenders prefer clients with a permanent, uninterrupted income stream. For the Dutch context, multivariate analysis shows that flexworkers on temporary/zero hour contracts have smaller chances of moving into home ownership than those on permanent contracts. They also tend to express less preference for home ownership. Because flexworkers often experience spells of unemployment, risk aversion appears to play a role. Our findings show, in the Netherlands at least, that self-employed freelancers do not experience too many problems in accessing home ownership, possibly because of more stable and higher incomes. However, the role of flexible labour is on the rise and policy-makers might consider methods to promote access to home ownership, ranging from mortgage guarantee schemes to mortgage payment insurances. Our research findings may not always be valid for other countries because of international variation in institutional arrangements, such as unemployment benefits, mortgage insurance and guarantee schemes, etc., but, nonetheless, sheds considerable light on this policy issue.

Keywords:

• Home ownership
• flexiblilisation of labour
• the Netherlands

Introduction

Since the late 1970s, national economies have become increasingly globalised. Western nations have become part of the dynamic international economy with many companies preferring a more flexible workforce that can better cope with the variable workload emanating from global economic cycles. Furthermore, innovations have followed one another at an increasing pace, creating an environment where products and services quickly become outdated through new innovations. Many economists and international think tanks have made the case for more labour market flexibility, because rigid employment protection legislation (EPL) makes it very time consuming and costly to fire under-used employees (OECD, 2016). Flexibilisation of the labour market may have its positive dimensions for commercial enterprises, but one downside is a decline in income security because of variable wages and temporary unemployment. At the same time, many Western governments have pursued an increase of owner occupied housing consumption. This appears to be somewhat at odds with flexibilisation of labour because it can result in un(der)employment spells and income variability, while mortgage lenders usually demand a stable income stream from secure and uninterrupted employment.

This article intends to shed empirical light on the relation between home ownership and flexible labour at the household level. Research on this topic has become increasingly relevant in the aftermath of the Global Financial Crisis, which stimulated new rounds of labour market flexibilisation. Although this topic generated research in Britain in the late 1990s (see Ford & Wilcox, 1998), recent systematic analyses of this issue are rare. Scarcity of integrated micro databases that combine labour market position with housing variables may be one reason. For the Netherlands, such data have now been made available. As a way to meaningfully analyse this data, in the rest of the article that follows, we first provide an introductory framework including a comparative overview of the development of labour market flexibilisation in a selection of Western European countries. This will also include some tentative explanations of international variation in housing finance and the meaning of tenure. We then turn to the micro context where we further elaborate on the theoretical relation between home ownership and household characteristics as found in the international literature.

The empirical part of this article focuses on an analysis of tenure preferences and tenure outcomes by labour market position. Is it actually possible to find an autonomous effect of tenure preferences stated by those in flexible labour as opposed to workers on a permanent contract? Furthermore, we investigate whether these preferences also match actual tenure outcomes. In seeking clues on the effects of macro-economic changes on tenure preferences and tenure outcomes, we compare results during a prosperous economic period (around 2006), with outcomes during a crisis (around 2012). We focus on these outcomes and variables knowing that the risk of losing a home for people in the flexible labour force, while socially relevant, is a complex research issue that arguably lies beyond the scope of a single article.

The remainder of this article is organised as follows. The next section presents a macro, international comparative view on flexible labour market practices and the main changes in the last years. It then continues in more detail with the theoretical relations between flexible labour, the life cycle and tenure choices at the household level. It proceeds further with a section on the data and methodology and then continues with the results of the quantitative analysis. The last section reflects on the research findings.

Convergence or divergence: labour relations, housing finance and tenure in a globalising world

Before we continue, it is useful to provide a clear definition of flexible labour as it involves several practices. A classic study by Atkinson (1984) on labour flexibilisation describes the ‘shifting employment practices’ of that era. Outcomes are often, next to reduced employment protection, practices of temporal flexibility and wage flexibility. Temporal flexibility refers to variable hours or temporary contracts, we refer to this as flexworking. Another form of labour flexiblisation is self-employment without personnel, the so called freelancer or ‘own account worker’. These persons act as (sub)contractors and agree to provide a contractually agreed service. Unlike flexworkers, they are not employees. In many cases, freelancers agree to perform a job for a fixed sum of money, but there are also freelancers who ‘bill’ every worked hour.

Labour relations

The literature shows that flexibilisation of labour markets has not been one uniform process throughout Western Europe (see for example Gebel & Giesecke, 2011). In fact, Table 1 shows significant variations in EPL levels in the mid-1990s. The large difference of EPL levels between the United Kingdom and Ireland on the one hand and Western European continental nations on the other hand, are explained by the varieties of capitalism approach (Hall and Soskice, 2001). Estevez-Abe, Iversen, and Soskice (2001) indicate that labour market policies in these liberal market economies (LME) are based on the liberal principle of ‘employment at will’, where companies can fire employees in cases where they do not function or when economic conditions demand so. The other nations in Table 1 are coordinated market economies (CME). CMEs provide a higher general level of EPL than LMEs, but there are substantial differences between CMEs. Some CMEs combine relatively low levels of EPL with generous unemployment benefits (UBs) (see Table 1). In order to keep generous UBs affordable, they usually last for a few years and are backed up with active labour market policies, which stimulate and assist the unemployed in finding new employment (Esping-Andersen, 1990). This approach results in more flexibility for companies, while redundant workers do not experience too great an income decline. Other CMEs combine high EPLs with very limited UB. In these countries, the high EPL provides income security to workers, but the downside is that it can lead to losses for enterprises which cannot restructure their workforce in reaction to an economic downturn. Many studies also indicate that high EPLs, such as the ones in Southern Europe, correlate with a higher incidence of temporary contracts and informal labour (for example Gebel & Giesecke, 2011). A third group of CMEs combine relatively high EPLs with relatively high UBs, which gives a high degree of income security to workers. The above serves as a general introduction on labour market variations and more backgrounds to these differences can be found in the varieties of capitalism literature (see Estevez-Abe et al., 2001). There has been a suggestion that the subdivision of EPL and UB levels in West-European CMEs runs parallel to Esping-Andersen's welfare regimes (Sapir, 2006), but in our data we do not see a strong link. For instance, Sweden and Italy, main representatives of specific welfare regimes (Scandinavian and Mediterranean), do not fit the picture that well. Table 1 shows that Sweden is closer to the continental European countries, while Italy is a case on its own rather than representing the Mediterranean countries.

Table 1. Labour market protection: employment protection legislation (EPL) for permanent and temporary contracts and unemployment replacement rate in selected EU countries.

	EPL permanent contracts		EPL temporary contracts		Unemployment replacement rate average first two years
Flexworkers (temporary contract)	1990	2013	1990	2013	2009
Netherlands	3.04	2.82	1.38	0.94	65%
Germany	2.58	2.68	3.25	1.13	56%
France	2.34	2.38	3.06	3.63	66%
Sweden	2.80	2.61	4.08	0.81	64%
Denmark	2.18	2.20	3.13	1.38	68%
Norway	2.33	2.33	3.13	3.00	72%
Italy	2.76	2.68	4.88	2.00	19%
Portugal	4.83	3.18	3.38	1.81	79%
Spain	3.55	2.05	3.75	2.56	67%
Ireland	1.44	1.40	0.25	0.63	50%
United Kingdom	1.10	1.10	0.25	0.25	28%

Sources: EPL = extracted from OECD.org data-sets on EPL, regular contracts (individual and collective) and temporary contracts; unemployment replacement rations from OECD (2009, p. 76).

So then, in the wake of calls for more flexibilisation of national labour markets since 1995, what has happened to EPL in the respective nations? Do we see any convergence toward a very flexible labour market throughout Western Europe? It should first be noted that different EPL applies to permanent and temporary contracts. The EPL for permanent contracts (and collective dismissal) refers to ‘procedural difficulties’ and direct costs such as severance payments to dismissed employees (Gebel & Giesecke, 2011). The EPL for temporary contracts refers more to the restrictions on the use of temporary contracts (Gebel & Giesecke, 2011). A high EPL for temporary contracts indicates that it may not be allowed to hire employees on subsequent temporary contracts or it imposes restrictions on hiring zero-hour contract workers on a prolonged basis. Interestingly, the data show that EPL has remained mostly unchanged with regard to permanent contracts (Table 1). However, a majority of CMEs have responded by reducing the strictness of EPL for temporary contracts. The contemporary literature indicates that this variation runs along lines of workers’ skill-levels (see Bonoli and Natali, 2011; Gebel & Giesecke, 2011). Highly skilled workers often have permanent contracts because they are indispensable and also require significant (training) investment by their employers, while less-skilled workers are more easily replaced at less cost. So overall, it seems that CMEs have responded to the ‘forces of globalisation’ by giving employers more opportunities to rely more on less restrictive, temporary contracts. As such, many countries show an increase of flexworkers in the labour force (Table 2). The main exceptions here are Denmark and Norway, where the labour market was already relatively flexible and the necessity to increase the number of temporary jobs appears to have been less urgent. Also the LMEs (UK and Ireland) show little change because their EPL was already quite relaxed from the start.

Table 2. Percentage of freelancers and flexworkers below in labour force in 11 selected West-European countries.

	Flexworkers			Freelancers
	1995	2015	Change percentage points between 1995 and 2015	1995	2013	Change percentage points between 1995 and 2013
Netherlands	10.8	20.0	9.2	4.5	8.6	4.0
Germany	10.4	13.2	2.8	2.6	3.9	1.3
France	12.2	16.6	4.4	4.0	4.2	0.2
Sweden	13.0	16.3	3.3	5.1	4.1	−1.0
Denmark	12.1	8.7	−3.4	2.9	3.3	0.4
Norway	13.2	8.0	−5.2	3.5	3.5	0.0
Italy	7.2	14.1	6.9	5.9	8.8	3.0
Portugal	10.1	22.0	11.9	10.5	6.3	−4.3
Spain	35.0	25.2	−9.8	7.6	6.7	−0.9
Ireland	13.2	8.7	−4.5	7.4	6.6	−0.8
United Kingdom	6.9	6.1	−0.8	6.3	8.3	2.0

Source: Eurostat. Labour Force Survey tables lfsa_esgais (Authors recalculation, with population data) and lfsa_etpga.

Turning to the Netherlands, the focus of our empirical research, the data reveal that EPL for permanent contracts has declined slightly, while a more significant decline is visible with regard to EPL for temporary contracts. In the next paragraph, we will elaborate more how these macro level, institutional changes play out at the household level and how they might relate to tenure choices.

Housing finance

In addition to considerable national variations in EPL, the literature on housing finance (mortgaged loan provision) reveals some variations that appear to run in parallel to the LME and CME divide. Interestingly, in many CMEs a higher degree of income security derived from high EPLs and/or generous UBs does not coincide with much more generous mortgage lending than in LMEs (see Schwartz & Seabrooke, 2008). In fact, LMEs often tend to have high rates of home ownership and highly accessible mortgage markets, which at times provide ‘subprime’ loans to households on insecure incomes. One explanation relates to LME banking sectors off-loading risky loans by means of securitisation and liberal regulators allowing them to do so (see Schwartz & Seabrooke, 2008). Indeed, the Global Financial Crisis revealed that this practice can rebound on vulnerable home owners, as well as on the entire financial sector. In general, financial sectors in CMEs are much stricter (towards flexworkers) and often demand down payments. Exceptions are found in Scandinavia and the Netherlands where mortgage guarantee programmes stimulate lower down- payment practices. In the Netherlands, banks set somewhat stricter criteria for flexworkers, but proof of a good past performance with regard to income stability will lift the largest barriers.

There is some consensus that both leverage and loan conditions in Western mortgage markets were relaxed from the mid-1990s until 2008 (Scanlon, Lunde, & Whitehead, 2011) and in some countries even led to subprime practices. To a great extent this applied to the Netherlands, but large-scale subprime lending never occurred. Since 2008 conditions have become more stringent for all loans (Scanlon et al., 2011). There is little solid evidence that Dutch mortgage lenders have put additional restrictions on loans to flexworkers and freelancers, but some qualitative research shows that it has become harder for both freelancers and those on flexible-temporary employment contracts (Companen, 2012).

Tenure

The main housing tenures (social rental, private rental and owner occupation) are differently perceived according to country, due to institutional and cultural variations. For example, Kemeny (2006) indicated that in liberal countries, social rental is mostly a residualised and stigmatised sector which only serves the most economically and socially vulnerable households. From this perspective, even those who are eligible for (means-tested) social rental housing might avoid it. In countries where the social rental sector serves a broad target group, it is much less stigmatised and more people will accept it as a viable alternative. Furthermore, a residualised social rental sector is often neglected in terms of maintenance quality and regarded as less attractive (see Haffner, Hoekstra, Oxley, & van der Heijden, 2009). Private rental also has a status that differs according to country. In countries where private rental competes with a high quality and accessible (subsidised) social housing sector, it makes for a relatively unattractive alternative, while it may be considered more attractive when the social rental sector is residualised. Interestingly, liberal countries often combine residualised social housing and limited tenure security in the private rental sector (see Haffner et al., 2009). Tenure insecurity can neutralise the competitive advantage of private rental over the residualised social housing sector. Tenure security in the owner occupied sector is strong in all countries, given that mortgage payments are not missed. Such institutional arrangements can lead to large variations in tenure preferences internationally. In fact, most LMEs have clearly established home ownership ideologies, which steers tenure preferences (Ronald, 2008), in spite of less income security. In some CMEs with a culture of self-provided owner-occupied housing, such as Germany, Belgium and Italy, owner occupation is strongly related to building the ideal house according to the tastes of the individual household (see Dol, Lennartz, & De Decker, 2012). In this context, owner occupation has (another) significant advantage over (private) rental. The Netherlands is a nation without much self-provision (self-build) and households buy relatively standardised dwellings from commercial developers (Dol et al., 2012).

Both the Dutch social and private rental sectors give a high level of tenure security with contracts of indefinite duration (Haffner et al., 2009). Furthermore, the Dutch social rental sector is not residualised, it is affordable and the dwelling quality is high compared to other, more residualised social housing systems. As such, the Dutch social rental sector is considered relatively attractive by many households. However, the social rental sector has become more restricted in terms of qualification and allocation (in terms of income-based means testing) (see Boelhouwer & Priemus, 2014). It still serves a broad target group, but not as broad as it used to be. Also, tenants have a right to a housing allowance, when their income is insufficient to pay the rent. This may be considered attractive for Dutch flexworkers. For owner occupiers, housing allowances are not available, but they do benefit from tax relief on mortgage interest.

It should also be noted that the degree of urbanisation often has a correlation with tenure (see for instance Helderman, 2007). In general, owner occupied dwellings are relatively scarce in urbanised areas, while there is relatively more supply in rural areas. In other words, the availability of owner occupied dwellings is expected to have a correlation with the propensity to move into owner occupation. Under conditions of ample demand for urban locations, this may lead to higher house price levels and crowding out of lower income households.

Labour position, the life course and tenure choice

Labour position and life course

The above suggests that labour market regulation, housing finance and the dominant perception of tenure shapes the (national or regional) context in which individual households make their tenure choice. For the Netherlands, we saw that owner occupation is a viable alternative because the housing finance system offers guarantees to first-time buyers, while unemployment will not cause dramatic income decline, that is, mostly for those that have a solid (permanent contract) employment history. Still, for many below median-income households, the social rental sector offers a reasonable housing alternative.

We now turn to the micro level in more detail. Within the boundaries set by the aforementioned national institutions and practices, individual household choices shape tenure outcomes. Again, income stability is crucial because it strongly determines the possibilities of gaining access to mortgaged loans. The literature indicates that permanent contracts and income stability have a strong relation with stable phases in the life course of households (see Remery, Van Doorne-Huiskes, & Schippers, 2002). Stable phases in the life course can be regarded as an outcome of choices by individuals (Remery et al, 2002). For instance, individuals decide to marry and raise children which leads to certain commitments. Such commitments are best met by means of solid employment contracts with stable income streams. Remery et al. (2002) indicate that a majority of households prefer a permanent contract because of the security and ‘a clear perspective on the future’. However, some individuals, such as young people who are exploring the labour market, may not be in a phase where they seek stability and in fact feel restricted by the inflexibility of terms of employment such as a longer period of notice. These are referred to as risk takers, who have fewer commitments to others (Remery et al., 2002). Also some specific family households which have many commitments may still prefer more risk because they expect higher returns from self-employment (freelancing) above permanent employment contracts. In many cases, freelancers are professional specialists who generate enough income to overcome temporary income decline. However, there is now a trend in the Netherlands where increasing numbers of intermediately skilled artisans such as carpenters and plumbers are self-employed. For the highly cyclical construction industry this can be very beneficial, but the latest crisis showed that it is risky for workers because their income levels do not allow for the creation of a personal financial reserve. After all, freelancers do not fall under any collective wage agreement in CMEs and can only rely on a basic state benefit which in the Netherlands varies from less than €700 net for single persons to about €1350 net for families.

Overall, a high degree of labour market flexibilisation, can certainly lead to involuntary flexwork for (family) households, who would actually benefit much more from permanent contracts (see De Cuyper, De Witte, & Van Emmerik, 2011). As indicated, permanent contracts are often granted to workers with high skill-levels and/or specialised knowledge, while those with lower skill levels increasingly rely on temporary contracts (Barbieri, 2009).

Safe as houses: relations between income, life course and tenure choice

A large body of international research shows that stable phases in the life course are related to owner occupation (see inter alia, Beer & Faulkner, 2011), even in those countries where owner occupation levels are comparatively low, such as in Germany (Tegeder & Helbrecht, 2007). This literature shows that the transition into home ownership is connected to having a stable relationship (marriage) and raising a family with children (see inter alia, Beer & Faulkner, 2011; Feijten & Mulder, 2002). Being single is often negatively related to home ownership. Young singles especially, are often in a dynamic life phase, searching for the ‘right job’ and the ‘right spouse’, which does not contribute to stability. Indeed young households, whether single or not, have a high degree of mobility and rental housing facilitates this better than home ownership (see Beer & Faulkner, 2011). Although stable income is crucial for access into home ownership, educational levels for young singles and young couples are also of great importance. In many cases, a high educational level of a young mortgage applicant is considered a lower risk, as these persons will often find permanent employment and their incomes are also expected to rise in the near future (Companen, 2012).

Furthermore, an above-average income can give a more solid base for home ownership, because it can reduce the impact of income decline and the possibility of mortgage payment arrears (see De Groot, Manting, & Mulder, 2013). In this sense, age also plays a role because incomes are often positively correlated with age. A Dutch study by Dol, Boumeester, and Mariën (2014) reveals that household incomes show no great differences between freelancers and households on permanent contracts. The main reason is that many freelancers are often skilled professionals with an employment history before starting their own business. In the Dutch case, many are already owner occupiers and have often become so in a phase when they had a permanent employment contract. This contrasts with flexworkers who are often young and have substantially lower household incomes (Dol et al., 2014).

Lastly, there is often path dependency in housing trajectories in the life course of households. For instance, a publication by Helderman (2007) shows that once a household has accessed home ownership, they will often make the same tenure choice in a subsequent move. Trading up to a higher quality dwelling is often supported by capital gains (and repayments) from the current dwelling and owner occupation then becomes a rational choice, rather than using the assets to support renting a larger dwelling.

A special tenure preference for flexworkers and freelancers?

The literature suggests that home ownership can have benefits for flexworkers. Many flexworkers are under-served in pension benefits, while home ownership can provide an asset in old age (see inter alia, Doling & Elsinga, 2013). In many countries, home ownership is regarded as a pension provision strategy, but the current Dutch pension system has good coverage for the majority of workers. It has a mandatory employment-related pension scheme, and the collective pension funds invest the pension contributions across the globe for future generations. However, for flexworkers this may not always lead to the desired retirement income. Where a major part of the working life has been spent in flexwork, unemployment spells and variable wage levels may put insufficient contributions into the (collective) pension fund.

For Dutch freelancers, there is no mandatory collective pension fund, compelling them to accumulate private retirement assets if they wish to receive more than the basic state pension. Freelancers have a variety of strategies at their disposal, ranging from a commercial pension fund to personal investment strategies. One of the personal investment strategies can be the purchase of an owner occupied dwelling as it provides rent free shelter at a later stage. In this scenario, the state pension and modest additional savings would be sufficient to have a decent life in old age. Moreover, a dwelling can be sold in order to release equity: the retiree would then ‘trade down’ to a rental dwelling or to a smaller owner occupied apartment. Furthermore, the Dutch financial markets have started to provide reverse mortgage products, which can be used to release equity, while not having to sell the dwelling.

In sum, it can be expected that many freelancers are willing to invest in an owner occupied dwelling, but possibly especially those who have a relatively stable business. Although we have to speculate to some extent, we assume that for flexworkers, the choice of owner occupation may still be less attractive. This can be substantiated by a recent report for the Ministry of Social Affairs and Employment, which shows that flexworkers often have spells of unemployment and therefore often rely more on social security than other employed households (SEO, 2013). During a spell of unemployment, the Dutch central government's system of income-related rental allowances provides a good hedge against income fluctuations. In sum, the owner occupied sector may not be the best option for flexworkers on low and variable incomes. This idea is to a great extent shared by the Dutch mortgage providers, who set special criteria for flexworkers. In the next section, we will explore mortgage provision for flexworkers and freelancers in more detail.

Hypotheses

Flexworkers often have characteristics that are negatively correlated with home ownership such as youth, small household composition, and low income. However, even when controlling for these characteristics, we assume that their more precarious labour market position has an additional negative effect on home ownership preferences. This may be partly a result of constraints set by mortgage lenders, but the specific Dutch institutional context may also have an effect on tenure choice. Given their spells of unemployment, it may be a better choice to be in the rental sector, where income decline can be compensated through housing benefits. Although flexworkers might prefer home ownership as an additional asset vehicle that compensates for gaps in pension build up, we assume the other factors are stronger drivers of a negative choice for home ownership.

Controlling for other determinants of home ownership at the household level, we expect that freelancers have a relatively strong preference for home ownership as it can be regarded as a part of investment for pension provision and possibly also as a collateral for business loans. Although they may have to meet special lenders’ criteria, we assume that, given their economic position, most will not meet many barriers during normal economic circumstances. Many of them are professionals who have established business networks to guarantee stable income. This situation might change during economic decline, when contracts become scarcer.

Data and methodology

We use data from the cross-sectional waves of the 2006 and 2012 Housing Research Netherlands database (Woon Onderzoek Nederland, WoON). The WoON, database of 2006 holds 55,958 cases, while 60,365 cases are included in the 2012 database. The original WoON-database does not break down the labour status into permanent contract, flexwork or freelance, but the national statistics service (Statistics Netherlands) have added those data from national (tax) registries. It includes the ‘zelfstandige zonder personeel’, self-employed without personnel which is the Dutch equivalent of the freelancer. They take on jobs for a contractually agreed amount of money. The other category is flexworkers, who either work with a temporary contract or who can be called upon for very short term jobs and/or jobs with variable hours. This includes so-called zero hour contracts, which are permanent contracts, but at variable hours.

The WoON data-sets are based on a detailed questionnaire on household characteristics, housing situation and house moving intentions. For those who intend to move, questions were asked about the preferred future housing situation. The WoON also includes data on households that have moved in the last two years. These recent movers provided additional information about their previous housing situation.

The data used are cross-sectional and not longitudinal. For recent movers in the last two years, this could lead to some issues in the data. For example, there is a ‘theoretical’ possibility that a breadwinner on a permanent contract takes out a mortgage and shortly afterwards becomes a freelancer. Such would be strategic behaviour to circumvent stricter mortgage lenders’ criteria for freelancers. We assume that any such behaviour would mostly apply to young breadwinners who aim to move out of a rental home into an owner occupied dwelling (first time buyers). For those who move from an owner occupied dwelling to another owner occupied dwelling, such a strategy is often not needed because many have equity in their dwelling, which forms good collateral. In fact, these ‘movers’ form the majority of freelancers who move to an owner occupied dwelling.

However, we have no reasons to assume that there are many first time buyers who give up a permanent contract for a freelancing career shortly after moving house. Indeed, a recent Dutch industry report indicates that very few employees on a permanent contract consider becoming self-employed (NRC, 2016). Any positive choice to change from a permanent job to freelancing is often associated with ‘older’ professionals who have good income prospects. Moreover, they are often long-time owner occupiers. So indeed, we might miss a few of those recent first time buyers who had a permanent position and then became a freelancer, but we expect that these are rare cases. With regard to flexworkers, we do not expect that there are many people who took out a mortgage, based on a permanent contract and then made a positive choice to become a flexworker after they moved into owner occupation. In the Dutch context, this is definitively considered as a step downwards in the labour market hierarchy.

The WoON database contains detailed income data of existing households from the tax registry, but income data for people that intend to move for the very first time, that is from the parental dwellings or shared student homes, are not available. Therefore, these ‘starters’ were not included in the analysis, but it would be very interesting to conduct research on this group when proper income data become available.

With regard to age, the literature suggests an inverted U relation between age and home ownership preferences and outcomes. This implies that a classification of age will lead to better model results. The data show a substantial increase after the age of 25 and a very gradual decline after 50 years. We classified the ages accordingly into three groups. Educational level is available in the WoON database. Secondary and professional refers to those that follow education until they are about 18 years old. Tertiary levels include higher professional education and university. About 20% of recent age cohorts finish a university degree, while about 30% finish a higher professional education (see Table 3). The WoON database gives a standard variable with the degree of urbanisation in five categories. Unfortunately, reliable data on the employment sector of respondents are not available in the WoON database. This can be a relevant factor, as the literature suggests that freelancing is increasingly shifting to occupations other than the highly paid, professional trades.

Table 3. Summary statistics of used variables (weighted).

	Intention to move		Recent movers
	2006	2012	2006	2012
Household composition*
Single person	27	28	24	31
Couple	29	31	32	31
Couple with child(ren)	44	41	44	38
Age head household
< 25	6	6	5	7
25–50	78	75	84	80
51–64	16	19	11	13
Educational level
Primary	19	13	16	11
Secondary and professional	35	36	34	34
Tertiary	46	51	50	55
Intended type of move
Movers from rent	54	46	55	63
Movers from ownership	46	54	45	37
Degree of urbanisation
Urban	13	11	10	10
Moderately urban	30	40	40	44
Moderately rural	22	14	15	14
Rural	25	23	26	23
Very rural	10	12	9	9
Labour position
Flexible-temporary contract	20	22	17	21
Freelancer	7	8	7	9
Permanent contract	73	70	76	70
Used cases in database (unweighted)	6702	8105	4215	2989

*For intention to move = household composition after intended move.

Source: WoON 2006 and 2012, enriched versions.

The method we use is a logistic regression where binary dependent variables are analysed, tenure of recent movers and tenure preferences of those who intend to move. As indicated, the databases are large and use weights that include an inflationary factor of around 100. This will make any regression output significant and therefore we eliminated the inflationary factor.

We estimate models for tenure preferences in 2006 and in 2012. Two other models are estimated for 2006 and 2012 on tenure outcomes after a (recent) move in the past two years. After selecting households who intend to move or who have recently moved, there are more than enough cases left to perform multivariate analysis (see Table 3). We only report significance at the 5% and 1% levels. The 10% level is too small for databases that we work with (see also Table 2).

Results

First, a model was estimated with those household characteristics that have been mentioned in the literature as having a relation with owner occupation. In a second model, we add the position in the housing market and the degree of urbanisation, as they are also expected to play a role (see literature review). These first two models serve as a general test of previous research findings. By adding the labour position variable in the third model, we test our main hypotheses on the relation between the labour market position and tenure outcomes (and choice).

With regard to the estimation of explanatory models, it is general practice to find the best model fit, which can be achieved by transformations of the data and/or adding interaction terms. This has been done in the explorative analysis but none of these transformations worked for all four models. For reasons of comparability, we chose to leave out any transformations or interactions. The results in Tables 3 to 6 need to be regarded as the best outcomes for comparability and not as the best fits.

Home ownership preferences of households that intend to move

The results for the models on home ownership preferences are presented in Tables 4 and 5. The relative chances of preferring home ownership are shown for each variable. The first two columns give B-parameters and significance levels. The standard error (SE) is listed in the third column. Where the odds ratio (OR) is smaller than 1, the predicted odds are smaller than the reference category and when the OR is larger than 1, the odds are greater. The general variables mentioned in the literature as drivers of home ownership, show mostly the expected results. Only the stronger preference of single persons for home ownership than that of families with children contradicts the expectation. Households in the ages from 25 to 50 have the highest preference for home ownership. Furthermore, there is a positive relation between household income level and educational level on the one hand and home ownership preferences.

Table 4. Binary logistic regression on households that prefer owner occupation and that intend to move within two years, 2006.

	Model 1				Model 2				Model 3
	B	Sign	SE	OR	B	Sign	SE	OR	B	Sign	SE	OR
Household composition (after intended move)
Single person	.310	**	.080	1.364	.417	**	.088	1.571	.407	**	.089	1.502
Couple	.767	**	.084	2.153	.825	**	.091	2.282	.816	**	.092	2.262
Couple with child(ren) (ref)
Age head household
< 25	1.283	**	.142	3.608	1.960	**	.160	7.098	2.039	**	.162	7.682
25–50	1.450	**	.091	4.265	1.912	**	.111	6.768	1.943	**	.111	6.980
51–64 (ref)
Educational level
Primary	−1.103	**	.083	.332	−1.160	**	.093	.313	−1.165	**	.094	.312
Secondary and professional	−.467	**	.073	.627	−.649	**	.080	.523	−.672	**	.081	.511
Tertiary (ref)
Household income	.091	**	.004	1.095	.056	**	.004	1.057	.053	**	.004	1.055
Intended type of move
Movers from rent					−2.718	**	.105	.066	−2.710	**	.105	.067
Movers from ownership (ref)
Degree of urbanisation
Urban					−.617	**	.149	.539	−.606	**	.150	.546
Moderately urban					−.642	**	.134	.526	−.649	**	.134	.522
Moderately rural					−.262		.142	.770	−.272		.142	.762
Rural					−.187		.141	.830	−.212		.142	.809
Very rural (ref)
Labour position
Flexworkers									−.344	**	.082	.709
Freelancers									−.276		.148	.758
Permanent contract (ref)
Constant	−2.588	**	.167	.075	.304		.232	1.355	.459	*	.235	1.583
N	6702
Nagelkerke R square	.340				.501				.504
Initial log likelihood	8189
Model log likelihood	6349				5264				5244
Log likelihood change (Chi2), Df, p-value	1840 df = 7, p = 0.0				1085 df = 12, p = 0.0				20 df = 14, p = 0.0

*< 0.05; **p = <0.01.

Table 5. Binary logistic regression on households that prefer owner occupation and that intend to move within two years, 2012.

	Model 1				Model 2				Model 3
	B	Sign	SE	OR	B	Sign	SE	OR	B	Sign	SE	OR
Household composition (after intended move)
Single person	.071		.070	1.073	.245	**	.078	1.278	.271	**	.078	1.312
Couple	.330	**	.069	1.391	.438	**	.075	1.549	.455	**	.076	1.576
Couple with child(ren) (ref)
Age head household
< 25	.530	**	.121	1.698	1.181	**	.134	3.257	1.303	**	.138	3.682
25–50	1.071	**	.071	2.918	1.538	**	.083	4.658	1.581	**	.083	4.861
51–64 (ref)
Educational level
Primary	−1.124	**	.080	.325	−1.244	**	.090	.288	−1.241	**	.090	.289
Secondary and professional	−.319	**	.058	.727	−.451	**	.065	.637	−.448	**	.065	.639
Tertiary (ref)
Household income	.056	**	.002	1.058	.033	**	.002	1.033	.032	**	.002	1.032
Intended type of move
Movers from rent					−2.043	**	.069	.130	−2.025	**	.070	.132
Movers from ownership (ref)
Degree of urbanisation
Urban					−.759	**	.125	.468	−.735	**	.125	.479
Moderately urban					−.792	**	.103	.453	−.773	**	.103	.462
Moderately rural					−.358	**	.122	.699	−.335	**	.122	.715
Rural					−.162		.113	.851	−.145		.113	.865
Very rural (ref)
Labour position
Flexworkers									−.237	**	.072	.789
Freelancers									.247	*	.115	1.280
Permanent contract (ref)
Constant	−1.791	**	.131	.167	.243		.169	1.355	.222		.171	1.248
N	8105
Nagelkerke R square	.279				.433				.435
Initial log likelihood	10,300
Model log likelihood	8486				7274				7256
Log likelihood change (Chi2), Df, p-value	1814 df = 7, p = 0.0				1212 df = 12, p = 0.0				18 df = 14, p = 0.0

*< 0.05; **p = <0.01.

Adding the intended type of move and degree of urbanisation in Model 2 gives a strong effect. This becomes clear from the increase in Nagelkerke pseudo R² and the significant decline of the log likelihood. Those households who are in home ownership already, have a much higher preference for home ownership than renters. This comes as no surprise as simple tables show that around one third of renters prefer owner occupation, while this percentage is around 90 for owner occupiers who wish to move. The intended type of move impacts on age and shows that the current position on the housing market is strong. Overall, it can be said that younger and/or single households who are already home owners, have a strong chance of preferring to be a home owner in their next housing move than those who are not (yet) an owner occupier. Furthermore, there is a negative correlation between preferred degree of urbanisation and home ownership preferences. Households intending to move and with a preference for moderately urbanised areas have smaller OR than households that prefer non-urban areas.

Comparing the Model 2 results for 2006 with 2012 shows that the general patterns do not differ that much. It seems that the magnitude of the B-parameters becomes somewhat smaller in 2012. This indicates that differences between reference categories and the other categories are less pronounced, but overall the results are significant.

However, the main interest of this article is the effect of the labour position on home ownership preferences and choice. Both in 2006 and 2012, the model improves by adding the labour market variable, but it is no great leap. Home ownership preferences of flexworkers are decidedly smaller than for workers on a permanent contract, with smaller ORs in 2006 (0.709) and in 2012 (0.789). So given the other household characteristics, it is possible to state that being a flexworker has a negative effect on home ownership preferences. As the literature suggests, this should be explained by the insecurity of the labour market position, which includes income gaps and temporary unemployment. For freelancers, the OR is smaller in 2006 (0.758), but not significant. For a large database like the one used here, this should be interpreted as not significant. It thus appears that there is no large difference in the home ownership preferences between freelancers and those on permanent contracts. In 2012, the OR (1.280) for freelancers is higher compared to households on permanent contracts and the effect is significant at the 5% level. There is thus no strong proof for a specific autonomous preference of freelancers for owner occupation.

Recent moves to the owner occupied sector

The results for the models of recent moves into home ownership are presented in Tables 6 and 7. Model 1 shows that household composition, age, educational levels and household income all have a role to play. All the control variables show the expected results. In 2006, the explanatory value of this model, measured in Nagelkerke pseudo R², is similar to the ones of home ownership preferences. The previous position on the housing market and the degree of urbanisation of the living environment are also dominant factors in explaining the recent moves into home ownership. The analysis for 2012 shows some changes as well as the influence of conditions during the crisis period. Couples without child(ren) in particular have less chance of moving into home ownership compared to couples with children, but these results are not significant. Younger households have significantly lower odds than other age groups of moving into home ownership. Income has a significant and positive effect on the propensity to move into owner occupation. Adding type of move and degree of urbanisation improves the model significantly (Model 2). Renters and people in more urbanised areas have lower odds of moving into home ownership.

Table 6. Binary logistic regression on households that moved into owner occupation the two previous years, 2006.

	Model 1				Model 2				Model 3
	B	Sign	SE	OR	B	Sign	SE	OR	B	Sign	SE	OR
Household composition
Single person	−.129		.100	.879	.142		.107	1.152	.136		.108	1.145
Couple	.179	*	.091	1.195	.468	**	.097	1.596	.523	**	.098	1.687
Couple with child(ren) (ref)
Age head household
< 25	−.362		.198	.696	.284		.212	1.328	.623	**	.220	1.864
25–50	.205		.117	1.228	.708	**	.129	2.030	.807	**	.130	2.241
51–64 (ref)
Educational level
Primary	−.549	**	.106	.578	−.651	**	.112	.522	−.645	**	.114	.525
Secondary and professional	−.023		.084	.978	−.167		.090	.846	−.180	*	.091	.835
Tertiary (ref)
Household income	.081	**	.004	1.084	.071	**	.004	1.074	.066	**	.004	1.069
Type of move
Movers from rent					−1.457	**	.087	.233	−1.505	**	.089	.222
Movers from ownership (ref)
Degree of urbanisation
Urban					−1.056	**	.185	.348	−1.055	**	.186	.348
Moderately urban					−.570	**	.150	.566	−.529	**	.151	.589
Moderately rural					−.338	*	.170	.713	−.306		.172	.737
Rural					−.360	*	.159	.698	−.342	*	.161	.711
Very rural (ref)
Labour position
Flexworkers									−.888	**	.104	.412
Freelancers									.061		.165	1.063
Permanent contract (ref)
Constant	−1.864	**	.196	.155	−.730	**	.244	.482	−.530	*	.250	.589
N	4215
Nagelkerke R square	.300				.390				.407
Initial log likelihood	5466
Model log likelihood	4430				4063				3987
Log likelihood change (Chi2), Df, p-value	1036 df = 7, p = 0.0				367 df = 12, p = 0.0				76 df = 14, p = 0.0

*< 0.05; **p = <0.01.

Table 7. Binary logistic regression on households that moved into owner occupation the two previous years, 2012.

	Model 1				Model 2				Model 3
	B	Sign	SE	OR	B	Sign	SE	OR	B	Sign	SE	OR
Household composition
Single person	−.086		.112	.879	.056		.119	1.058	.021		.121	1.021
Couple	−.108		.102	1.195	.178		.108	1.195	.197		.110	1.218
Couple with child(ren) (ref)
Age head household
< 25	−.452	*	.200	.696	−.136		.209	.873	.104		.215	1.110
25–50	.214		.126	1.228	.420	**	.134	1.523	.517	**	.135	1.677
51–64 (ref)
Educational level
Primary	−.651	**	.140	.578	−.791	**	.146	.453	−.836	**	.148	.434
Secondary and professional	−.064		.092	.978	−.219	*	.097	.804	−.272	**	.099	.762
Tertiary (ref)
Household income	.049	**	.003	1.084	.039	**	.004	1.040	.034	**	.004	1.034
Type of move
Movers from rent					−1.183	**	.098	.306	−1.169	**	.100	.311
Movers from ownership (ref)
Degree of urbanisation
Urban					−1.311	**	.214	.270	−1.237	**	.217	.290
Moderately urban					−.584	**	.173	.558	−.527	**	.175	.590
Moderately rural					−.532	**	.197	.587	−.559	**	.199	.572
Rural					−.365	*	.184	.694	−.350		.186	.705
Very rural (ref)
Labour position
Flexworkers									−1.010	**	.113	.364
Freelancers									−.210		.155	.810
Permanent contract (ref)
Constant	−1.445	**	.209	.236	.010		.273	1.010	.305		.280	1.357
N	2989
Nagelkerke R square	.246				.322				.350
Initial log likelihood	4088
Model log likelihood	3483				3268				3185
Log likelihood change (Chi2), Df, p-value	605 df = 7, p = 0.0				215 df = 12, p = 0.0				83 df = 14, p = 0.0

*< 0.05; **p = <0.01.

The last model adds labour market position. Controlling for all the other variables, flexworkers have smaller chances of moving into home ownership than households on a permanent contract in 2006 and in 2012. In both years, flexworkers have smaller odds (around 0.400) of moving into home ownership compared to households on a permanent contract. This does not always need to be interpreted as a result of constraints on the mortgage market, because flexworkers also have smaller preferences for home ownership than households on a permanent contract (see Table 3 and 4). However, we assume that flexworkers’ preferences for rental are also influenced by the safety nets within the Dutch rental system as rental housing benefits for lower incomes cushion the impact of income decline.

In 2006, freelancers have similar chances of moving into home ownership (OR = 1.063), to households on permanent contracts. This seems rather to contradict the results for home ownership preferences, that were somewhat smaller. Possibly, the favourable conditions on the housing and mortgage markets, i.e. rising house prices and readily available credit, may have stimulated this behaviour. However, in 2012 the chances of moving into home ownership have become smaller for freelancers (OR = 0.810) but the effect is still insignificant. The somewhat smaller effect may be related to a more critical attitude of mortgage lenders towards freelancers during uncertain economic times. Other factors that might play a role are a decline in contracts for freelancers during the crisis: as a result freelancers themselves may have become more reluctant to invest in housing. Overall the effect is still insignificant for freelancers.

Conclusions

This article has sought to shed some new light on the effect of changing labour markets on home ownership. Our investigation first shows that many Western European countries have increasing percentages of workers on temporary contracts and freelancers (own account workers). At the same time, the EPL for those on temporary contracts has been relaxed in many countries, while those on permanent contracts still have a high degree of employment protection. This leads to more income variability for a larger group of households, while the financial sector prefers stable incomes when providing mortgaged loans for prospective home buyers. What does this mean for access to home ownership? Is home ownership indeed less accessible for households with non-permanent employment positions?

We have investigated the situation in the Netherlands where home ownership rates have risen sharply in the past two decades and are now close to 60%. At the same time, the Netherlands is a European front-runner in working on variable hour contracts or temporary labour contracts, while there are now also many freelancers: that is self-employed persons without any personnel. EPL for Dutch temporary contracts has decreased in the past few years, while it can also be somewhat difficult to gain access to mortgaged loans for those who are on temporary contracts and variable income (freelancers). As such the Netherlands is an interesting case, but the outcomes should not be directly translated to any other Western European country.

For the study, it was hypothesised that flexworkers may have less preference for owner occupation because of the risk. Many experience income fluctuations and in the Dutch context, spells of income decline can be compensated through housing allowances in the rental sector. Such subsidies are not automatically provided for owner occupiers. These expectations hold true in an analysis of micro data for the years 2006 and 2012. Flexworkers who intend to move indeed have less preference for home ownership, controlling for other relevant variables. The same applies to flexworkers who recently moved into home ownership. Both income insecurity and a stringent attitude of mortgage providers may play a role in this. The main conclusion for the Dutch context is that home ownership and flexworkers do not match well. Any further increases in the number of flexworkers may not be beneficial to home ownership.

With regard to freelancers, the general expectation was that they might have higher preferences for home ownership because they have to provide for their own pension, which makes home ownership attractive as a pension asset. Furthermore, home ownership could serve as collateral for small business loans. Unlike flexworkers, we expect that their more stable income might make their home ownership preferences comparable to households on permanent contracts. However, the results show that, controlling for other relevant variables, there is no proof that freelancers who intend to move have a significantly stronger preference for owner occupation than households on permanent contracts. The actual moving behaviour in 2006 shows that freelancers have the same relative odds as those on permanent contracts of moving into home ownership. In 2012, this parameter changes into somewhat smaller odds of freelancers moving into home ownership, but the effect is insignificant and does not allow for a firm conclusion about any crisis-related effect.

The results for freelancers appear somewhat at odds with the discussion about this topic in the Dutch media. Especially during the crisis, popular media suggested that many freelancers experienced problems in taking out a mortgage. We would argue that this is much more the result of specific household characteristics, but not so much freelancer status. For instance, websites of the main Dutch banks clearly indicate that for freelancers, they take into account factors such as experience (age) and occupational level (education as an indicator of prospects) in lending decisions. Indeed, such controls were also included in the analysis for flexworkers, but it is likely that their weak position gives a negative autonomous effect on home ownership access. Indeed, secondary literature points out that flexworkers have quite a bit of income insecurity (SEO, 2013), possibly much more than freelancers. To flesh out these details, more (qualitative) research is needed.

A last comment is that the results for freelancers may not always hold in the future. Freelancers are traditionally found in the professional trades, such as consultants, lawyers and medical specialists, and are often less vulnerable to income fluctuations. However, labour statistics show that there are increasing numbers of freelancers in the blue-collar sectors, including construction. Such changes in the structure of the freelance population might also change their fortunes in the owner occupied sector.

The problematic paradox of stimulating home ownership and flexibilisation of the labour market at the same time might be addressed by the development of new policies. A very obvious measure goes back to the trade-off suggested in the literature review (Sapir, 2006). In this trade-off, income decline as a result of flexibilisation should be cushioned with generous income replacement for all and not solely for those who have had long employment histories. It would be beneficial to support such measures with active labour market policies such as skills-adjustment or skills-improvement. However, such measures are often found in the Northern European welfare regimes and may not fit well with the more liberal welfare regimes that expect much more self-reliance of their inhabitants. We still expect that liberal countries pursue market solutions and they may think more in lines of private unemployment insurances, mortgage credit insurance or mortgage payment insurances. However, some findings in the English context show that these products are often costly (Doling & Ford, 2003). The financial market could also consider more flexible repayment schemes, where missed payments are added to the outstanding loan. Another solution that may fit better with non-liberal regimes is for instance a special government backed mortgage guarantee scheme, but in some countries generous UBs would also cushion the risk of home loss. Furthermore, policy-makers in countries that have (generous) systems of housing allowances for tenants in the rental sector, might consider extending these to the owner occupied sector. Any of the above measures can assist people on non-permanent contracts to access and sustain owner occupation.

Disclosure statement

No potential conflict of interest was reported by the authors.