Introduction

Deciding to buy a home involves a fundamental economic trade-off: the accumulation of housing wealth comes at the expense of geographic flexibility. While homeownership offers stability, it creates significant transaction costs that may anchor workers to specific locations. This paper investigates whether this "lock-in" effect traps individuals in declining local labor markets, thereby inhibiting their ability to climb the job ladder or escape unemployment.

Despite this suggestive evidence, we lack a unified framework that traces how an individual’s housing history, including both their homeownership decisions and residential locations, influences labor supply decisions and migration outcomes.

This paper fills that gap by constructing a novel linked dataset of the French population. By merging fiscal housing records with employment panels, I am able to observe the joint distribution of homeownership status and labor market transitions for millions of workers, moving beyond aggregate correlations to identify individual-level frictions.

This analysis reveals that the friction is geographically heterogeneous. As illustrated in Figure 1, homeownership rates in France are not randomly distributed; they are inversely related to local economic dynamism. Ownership is lowest in high-wage metropolitan areas like Paris (<40%) and highest in rural or post-industrial zones (>70%). This creates a potential "trap": the regions where workers most need the flexibility to leave are exactly the regions where homeownership makes them most immobile.

Do these spatial frictions translate into labor market penalties? Using this new data infrastructure, I document a severe "lock-in" effect: homeowners are 3.7 percentage points less likely to migrate and 4.2 percentage points less likely to change jobs than renters. Consequently, they experience 1.6 percentage points lower annual wage growth, results consistent with the hypothesis that housing rigidities dampen the efficient allocation of labor.

Figure 1: Geographic Heterogeneity in Homeownership. The map displays the share of homeowners by Zone d'Emploi. Ownership rates are highly spatially concentrated, ranging from under 40% in dynamic urban centers (like Paris) to over 70% in rural departments. These high-ownership zones are strongly negatively correlated with average local wages. 

Literature

This paper bridges the gap between the macroeconomic literature on housing frictions and micro-econometric evidence on labor mobility. The "Oswald Hypothesis" (Oswald, 1996; Blanchflower and Oswald, 2013) famously points out that high homeownership rates drive structural unemployment by reducing labor mobility. However, tests of this hypothesis have yielded mixed results, often finding that homeowners exhibit lower hazard rates of unemployment despite their immobility (Munch et al., 2008). I advance this literature in three ways. First, I overcome the historical lack of common identifiers in French administrative data by implementing the Bayesian Record Linkage methodology of Enamorado et al. (2019). This allows me to construct a novel linked dataset merging the universe of fiscal housing records (FIDELI) with detailed employment histories (Panel Tous Salariés), observing individual transitions that were previously invisible. To address the endogeneity of the decision to buy, I propose exploiting the 2014 DMTO transaction tax reform as a natural experiment to causally identify the downstream impact of transaction costs on labor mobility and wage growth (Bérard and Trannoy (2018) previously analyzed the impact of the reform regarding transaction volumes). Finally, I move beyond reduced-form estimates by developing a unified structural framework of spatial labor search and homeownership choice. I aim to integrate the asset market frictions of housing with the search frictions of the labor market (building on Head and Lloyd-Ellis, (2012); Rupert and Wasmer, (2012)), which I will calibrate to quantify the specific welfare losses generated by the "lock-in" effect. 

Data

To analyze the interaction between homeownership and labor outcomes, I leverage two complementary administrative sources produced by INSEE. The first is FIDELI (Fichiers Démographiques sur les Logements et les Individus), an exhaustive fiscal database covering the universe of French housing units and their occupants (2015-2023). Crucially, in addition to homeownership status and precise geolocation, FIDELI reports annual labor income and unemployment benefits, allowing for the observation of basic labor market transitions and residential mobility. However, it lacks the granularity required to observe specific employment conditions, such as hourly wages, contract types, or firm identifiers. To capture these detailed labor market histories, I utilize the Panel Tous Salariés (All-Employee Panel), a longitudinal sample tracking 1/12th of the workforce from 1976 to 2022. While this panel provides a rich set of employment variables, including hours worked, occupation, and sector, it contains no information on homeownership status. The probabilistic matching strategy described below serves to bridge these two sources, combining the universal housing coverage of FIDELI with the granular job histories of the Panel. 

Methodology: Bayesian Record Linkage 

A fundamental problem in this analysis is that FIDELI and the Panel Tous Salariés do not share a unique personal identifier (such as the social security number) due to privacy restrictions. However, since both datasets describe the same underlying population, I infer matches based on a vector of common attributes (K): fiscal/net income, age group, gender, zone of residence, and local-born status. To account for slight definitional differences between fiscal and labor data, continuous variables like income are discretized into bins or quantiles to robustify the comparison.

To solve this "many-to-many" matching problem, I adopt the Bayesian Record Linkage (BRL) framework developed by Enamorado et al. (2019) . The procedure operates in three stages:  

1. Blocking: To manage computational complexity, the data is blocked by Year and commuting zone, restricting the search space to individuals co-located in the same labor market.

2. Comparison Vectors (γ): Within these blocks, the algorithm computes vectors γ(i,j) measuring the degree of agreement (exact, partial, or disagreement) across the matching variables for every potential pair (i,j). 

3. Mixture Model: These vectors are modeled as being drawn from a mixture of two distributions, true matches (M) and non-matches (U), whose parameters are estimated via the Expectation-Maximization (EM) algorithm. 

Crucially, this method does not force a deterministic "best match," which often leads to false positives. Instead, I calculate the posterior probability, denoted as ϵᵢⱼ, that any specific pair is a true match . For every worker i in the labor panel, I construct a probability-weighted homeownership variable. This estimator acts as a "soft" classifier, representing the weighted average of homeownership status across all potential matches: 

Stylized Facts: The Geography of Ownership and Mobility 

Before presenting the causal estimates, I describe the key patterns observed in the data. These stylized facts provide the empirical motivation for the hypothesis that homeownership acts as a constraint on mobility. 

The Mobility Gap: Using the full population data from FIDELI,  I show two distinct migration behaviors: 

• Fact 1: The "Trap" Effect: Renters are significantly more mobile than owners, but the gap is not constant. In zones with high homeownership rates (often poorer labor markets), the gap widens. As Figure 2 below illustrates, renters in these zones have very high outmigration rates, effectively "fleeing" the bad market. In contrast, owners in these same zones have flat or depressed migration rates. This divergence suggests that while renters respond to negative local conditions by leaving, owners remain "locked in." 

Figure 2: The Divergence of Mobility. Annual outmigration rates for renters (red) and homeowners (blue) plotted against the local homeownership rate. As the homeownership rate increases, for France, a proxy for weaker local labor markets, renters "flee" at higher rates, while owners remain immobile. The widening gap illustrates the "lock-in" effect in distressed regions. 

• Fact 2: Barriers to Entry The friction exists on the entry side as well. High-homeownership locations receive significantly fewer newcomers compared to areas with deeper rental markets. This suggests that high ownership rates may reduce overall market liquidity, making it harder for workers to enter these regions even if they want to. 

Figure 3: Barriers to Entry. Inmigration rates plotted against the local homeownership rate. High-homeownership zones exhibit lower labor market fluidity, receiving significantly fewer incoming workers compared to zones with deeper rental markets. This suggests that housing rigidities constrain not just exit, but also entry into these local markets. 

• Fact 3: Asymmetric Upward Mobility Do homeowners move to better jobs when they do move? I classify moves based on the wage differential between Origin and Destination zones. In Figure 4 below I distinguish between:

  • • Renters (Left Panel): Exhibit significant "upward" mobility, frequently moving from Low-Wage Zones to High-Wage Zones.

    • Owners (Right Panel): Are significantly less likely to make these arbitrage moves. Their migration flows are often lateral (Low-to-Low) or non-existent, suggesting their moves are less responsive to wage signals.

Figure 4: Missing Upward Mobility for Homeowners. A heatmap of migration flows classified by the wage level of the Origin and Destination zones. Renters (left panel) exhibit strong upward mobility, frequently moving from "Low Wage" origins to "High Wage" destinations (the red clusters). In contrast, homeowners (right panel) rarely execute these arbitrage moves, effectively capping their earnings potential. 

Empirical Results: Quantifying the Lock-In 

To isolate a conditional correlation of homeownership with labor market outcomes, while controlling from confounding factors (such as age, industry, or local amenities), I estimate the following linear model using the probabilistically linked dataset: 

In this specification, the key variable is the probability-weighted homeownership status. The model includes Zone d'Emploi (δₑ) and Year (δₜ) fixed effects to absorb local shocks, while the vector X controls for age, experience, past wage, sector, and family structure as well as job characteristics.

• Result 1: The Barrier to Migration Homeownership acts as a significant barrier to geographic migration. The results indicate that, all else equal, homeowners are 3.7 percentage points less likely to migrate across commuting zones compared to renters (SE: 0.0045). Given the low baseline migration rate in France, this represents a substantial reduction in mobility. Furthermore, when homeowners do move, they are significantly less likely to relocate to a region with higher wages (Logit coefficient: -0.433, SE: 0.019), results consistent with the hypothesis that housing frictions prevent optimal spatial arbitrage. 

• Result 2: Reduced Labor Market Fluidity This geographic rigidity translates into reduced professional fluidity. Homeowners are 4.2 percentage points less likely to make job-to-job transitions than renters (SE: 0.0007). This suggests that the "lock-in" effect restricts the effective search radius of workers: because they cannot easily move locations, they have fewer outside options and are less likely to switch employers, even within their local market.

• Result 3: The Wage Growth Penalty The most direct welfare cost of this immobility is reduced earnings trajectories. Homeownership is associated with 1.57 percentage points lower annual wage growth (SE: 0.0008). This finding supports the hypothesis that limited mobility reduces a worker's bargaining power; unable to credibly threaten to leave for a better offer in a different city, homeowners may be forced to accept lower wage increases.