Abstract
Our empirical study stems from previous research on the inter-relations between residential status and microeconomic labor market outcomes. It focuses on employees and assesses the a priori ambiguous effect of homeownership on job-match quality. We use the French data set of the 1995–2001 European Community Household Panel to build a subjective measure of job downgrading. We estimate a recursive trivariate probit with partial observability that simultaneously models the residential status choice, its impact on the probability of being downgraded, and the selection into employment. Taking into account, the double selection process, into employment and into homeownership, and controlling unobserved individual heterogeneity, we find that private renters have between 30% and 40% higher probability of subjective downgrading than homeowners. Mortgage constraints increase the downgrading probability, but their effect is of a limited scope (around +2% percentage points for mortgagers compared with outright owners). We show that these results are robust to various specifications and instruments choice. Consequently, homeownership seems not to be associated to some harmful effects on the job-match quality. Our conclusions are consistent with recent microeconometric studies which call into question Oswald’s hypothesis.
Notes
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 In our paper, we will employ indifferently these two terms.
2 For example, ‘the use of realized matches is often regarded as inferior to the other two methods’ (Leuven & Oosterbeek, Citation2011).
3 We exclude households constituted by families implying indirect ascendants, or more than one direct ascendant, as well as, households which are not families.
4 We do not investigate potential discrepancies between qualifications and skills mismatch: see Quintini (Citation2011) for a review on the literature on this topic.
5 For the 1995 survey, Alba-Ramírez & Blázquez (Citation2003) find an overqualification rate equal to 53.8% in Spain against 59.4% for Budria & Moro-Egido (Citation2009) for the 2001 survey.
6 One instrument is the regional homeownership rate (Munch et al., Citation2006; van Leuvensteijn & Koning, Citation2004), albeit it was heavily criticized (de Graaf et al., 2009; Coulson & Fisher, Citation2009; de Graaff & van Leuvensteijn, Citation2013). Other instruments used in the literature are past residential status of parents (van Leuvensteijn & Koning, Citation2004), homeownership rate in the city where the individuals was born (van Leuvensteijn & Koning, Citation2004), age dummies (Flatau et al., Citation2003), father’s job (Battu et al., Citation2008), average distance to jobs (Brunet & Lesueur, Citation2004), the percentage of households in the area living in multifamily housing (Coulson & Fisher, Citation2009; Morescalchi, Citation2016), an indicator capturing whether the two first-born children in the household have the same sex (Coulson & Fisher, Citation2009; Morescalchi, Citation2016), the state marginal tax rate that applies to the mortgage interest deduction (Coulson & Fisher, Citation2009), the user cost of owning compared to renting (Brunet & Lesueur, Citation2004) and the relative cost of owning with mortgage versus renting (Morescalchi, Citation2016).
7 Results from all other models are available from the authors upon request.
8 Again, results from all other models are available from the authors upon simple request. Note that we also successfully test that our results are robust to the inclusion of individual time-means of other time-varying variables (namely education and urban are size), following Semykina (Citation2018).
9 For the sake of completeness, we also check that results remain the same when we exclude the Parisian area.
10 Similar to the homeownership equation, family structure and the spouse’s labor market status are not significant in the employment equation with the ‘Semykina-like’ estimation.
11 The highest absolute value is actually reached when the Parisian observations are discarded: in this case, the coefficient associated with the homeownership variables equals -0.911 (results are available upon request).
12 The marginal effect of homeownership is computed as the difference in the probability of downgrading, between an outright owner and a renter, taking into account the whole estimated system and the correlations between equations. The marginal effect of mortgage constraints is similarly evaluated, and defined as the difference in the probability of downgrading, between an outright owner and a mortgager.
13 The individual with the sample mean characteristics.
14 The modal individual has explanatory variables set a their sample mode.
15 This view is nonetheless challenged by Andersson & Mayock (Citation2014) or Ferreira et al. (Citation2010, Citation2011).