Home-ownership and unemployment : revisiting the Oswald hypothesis from a regional heterogeneity perspective

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY-NC-ND 3.0 ES https://creativecommons.org/licenses/by-nc-nd/3.0/es/ Home-ownership and unemployment : revisiting the Oswald hypothesis from a regional heterogeneity perspective © Ediuno. Ediciones de la Universidad de Oviedo / Oviedo University Press Published version Tohmo, Timo; Viinikainen, Jutta Tohmo, T., & Viinikainen, J. (2023). Home-ownership and unemployment : revisiting the Oswald hypothesis from a regional heterogeneity perspective. Economics and Business Letters, 12(4), 342-355. https://doi.org/10.17811/ebl.12.4.2023.342-355 2023 342 12(4), 342-355, 2023 Economics and Business Letters 12(4), 342-355, 2023 Home-ownership and unemployment: Revisiting the Oswald hypothesis from a regional heterogeneity perspective Timo Tohmo 1,* • Jutta Viinikainen 1 1 Jyväskylä University School of Business and Economics, University of Jyväskylä, Finland Received: 22 November 2022 Revised: 1 February 2023 Accepted: 3 May 2023 Abstract Home-ownership may create both positive and negative externalities. As an example of a negative externality, the so-called Oswald hypothesis suggests that a high home-ownership rate creates frictions in the economy and thus increases the unemployment rate. We approach this hypothesis from a novel perspective by taking into account regional differences in population density and dwelling composition. Using municipality-level panel data, we find that although the phenomenon identified by the Oswald hypothesis may not be omnipresent, it may manifest itself, particularly in semi-urban areas where the share of large ownership dwellings is high. We also find that in-migration to these areas is lower, which is consistent with the view that the home-ownership rate may affect migration flows and, thus, economic dynamics. Keywords: Home-ownership; Unemployment; Dwelling size; Rural area; Urban area JEL Classification Codes: J20, J61, J64, R10, R20 1. Introduction High unemployment harms economies and people's well-being (Machin & Manning, 1999). Besides labor market features, such as labor market institutions, other societal characteristics are also likely to impact the unemployment rate. The Oswald hypothesis (Oswald, 1996, 1999, 2009) offers a potential explanation for the cross-country and cross-area differences in unemployment. According to this hypothesis, a high rate of homeownership in a particular area can increase the overall unemployment rate in that region. Oswald's hypothesis originated from the observation that as owner-occupied housing increased in Western countries, so did unemployment. Further, unemployment tended to be higher in countries with a high homeownership rate. Oswald conjectured that, because owneroccupied households are relatively immobile, the increase in homeownership reduces labor mobility, which in turn increases unemployment, resulting from the reduced ability of firms and workers to match each other (Oswald, 1996). Later, Oswald elaborated on five possible mech- * Corresponding author. E-mail: timo.j.tohm[email protected]. ORCID: https://orcid.org/0000-0003-0012-1657 Citation: Tohmo, T. and Viinikainen, J. (2023) Home-ownership and unemployment: Revisiting the Oswald hypothesis from a regional heterogeneity perspective, Economics and Business Letters, 12(4), 342-355. DOI: 10.17811/ebl.12.4.2023.342-355 T. Tohmo and J. Viinikainen Home-ownership and unemployment 343 12(4), 342-355, 2023 anisms that might explain the positive association between owner-occupation and unemployment (Oswald, 1999, 2009). According to Oswald, owning a home may increase the risk of unemployment because homeowners are less willing to move in response to labor market volatility. In addition to this direct effect, Oswald hypothesized that high homeownership may create externalities, which increase unemployment. First, high homeownership may block new workers from entering the labor market if the availability of suitable rental housing is lacking. Second, lower mobility can create a skill-job mismatch, raising production costs and reducing real income, which harms the economy and causes job losses. Third, high homeownership can fuel the "not in my backyard" (NIMBY) attitude that affects e.g., zoning regulations and prevents new businesses from entering the area. Finally, if homeowners are unable to find suitable jobs in their area, they may need to commute. This may cause transportation congestion, which imposes monetary and psychological costs, reducing the benefits of having a job and increasing the attractiveness of not working. Later, two other mechanisms explaining the link between homeownership rates and unemployment were proposed (Laamanen, 2013, 2017). First, high mortgage payments may force homeowners to cut down on other expenditures, reducing aggregate demand in the area. Second, to manage mortgage payments, homeowners may increase their job search activity and lower their reservation wages leading to the displacement of other workers in the region and creating negative externalities on the local labor market. Empirical studies, using aggregate-level data, have typically found a positive relationship between homeownership and unemployment rates (e.g., Blanchflower & Oswald, 2013; Isebaert et al., 2015; Laamanen, 2017; Oswald, 1996, 1999, 2009; Pehkonen, 1999). These findings have been challenged by individual-level empirical results, which suggest that homeowners are typically less vulnerable to unemployment compared to renters (Broulíkova et al., 2020; Coulson & Fisher, 2002, 2009; Laamanen, 2017; Morescalchi, 2016; Munch et al., 2006, 2008; Van Leuvensteijn & Koning, 2004). These findings suggest that it may not be the direct link between homeownership and unemployment which explains the Oswald’s observation but rather the negative externalities of homeownership that increase regional unemployment. Further endorsing the role of such externalities, Blanchflower and Oswald (2013) found that fewer new businesses were created in areas with a high rate of owner-occupied housing, and areas with higher rates of homeownership had lower labor mobility and longer commute times. Simultaneously, Laamanen (2013, 2017) independently found empirical support for the view that reduced consumption and increased job competition may partly explain the conflicting findings at the individual and aggregate levels. Despite all of the previous research that has examined Oswald's hypothesis, there has been limited exploration of how distinct regional characteristics influence the relationship between homeownership and unemployment. For example, the level of urbanization may affect this relationship in several ways. First, urban areas have more job opportunities; rural areas being more dependent on specific industries can face significant unemployment if these industries experience a downturn. Therefore, labor mobility is likely to be less of an issue in urban areas, where there are better possibilities of finding new employment, and low migration resulting from homeownership may impact the unemployment rate in urban areas less than in rural areas. Second, housing conditions differ between urban and rural areas. Urban areas have higher housing density and smaller dwellings with higher prices, while rural areas have larger dwellings with lower prices. The high share of large owner-occupied dwellings in rural areas can contribute to a higher unemployment rate, as residents may be reluctant to move to the city even if they lose their job fearing a potential decrease in their standard of living. Negative housing equity, which is more common in rural areas, can also reduce homeowners' intentions to move, even when the labor market situation deteriorates (Henley, 1998). Third, political influence opportunities can vary between urban and rural areas. Urban areas tend to have a larger and more diverse population, leading to more complex political dynamics, while rural areas tend to T. Tohmo and J. Viinikainen Home-ownership and unemployment 344 12(4), 342-355, 2023 have a smaller and more homogeneous community making it easier for individuals to engage with local officials. This can result in greater potential for NIMBY-type behavior in rural areas. In conclusion, the high homeownership rate in rural areas may have a more pronounced effect on the unemployment rate there, resulting from both direct and indirect effects. This study takes a novel perspective on Oswald's hypothesis by considering urban-rural differences and paying attention to the heterogeneity in the local housing structure. Using municipal-level panel data from Finland, we examined the heterogeneity of the relationship between homeownership and unemployment among regions with different population densities and dwelling compositions. Specifically, we explored how the share of different-size-owned dwellings in rural, semi-urban, and urban regions relates to the local unemployment rate. Based on our results, Oswald’s hypothesis seems to apply to semi-urban areas where the share of large home-owned dwellings is high. 2. Methods For our main analyses, we used the natural logarithm of the municipality unemployment rate as the dependent variable. To explore migration flows, we measured in-migration and out-migration as the share of in-migrants (out-migrants) relative to the municipality's total population. Followinng Blanchflower & Oswald (2013), the explanatory variable of interest was the logarithm of the homeownership rate in the area. To explore housing structure's role, we categorized homeownership by size, indicating the share of different-size-owned dwellings relative to the area's total housing stock: lower small-size (7–29 m2), upper small-size (30–59 m2), lower medium-size (60–89 m2), upper medium-size (90–119 m2), and large (≥120 m2). While carrying out robustness analyses with different geographic aggregations, dwelling size was measured by room count given the unavailability of surface area data. Other control variables were education level, the share of 16–24-year-old residents in the area, the municipality’s total net costs, and year and municipality fixed effects. The education measure, formulated by Statistics Finland, represents the average level of education in the area. Each year of education, beyond the compulsory level, is assigned a value of 100 and multiplied by the corresponding population share. The municipality’s net costs per capita were calculated by subtracting the operating income from the operating costs and dividing the result by the total population. Net costs reflect the municipality's financial pressures, which can affect the unemployment rate. We used the fixed-effects method to control for all possible time-invariant heterogeneity across municipalities. Given the high auto-regressivity of unemployment (Blanchflower & Oswald, 2013), our model included a lagged dependent variable. Following Blanchflower & Oswald's (2013) study, we considered four lags in the homeownership rate in our estimations. We estimated separate models for university municipalities (n = 10), other urban municipalities (n = 51), semi-urban municipalities (n = 66), and rural municipalities (n = 209) to explore the heterogeneity in the unemployment-homeownership relationship. The classification was based on Statistics Finland's statistical grouping of municipalities, except for separating university municipalities from other urban areas. There are two main challenges related to our analyses. First, although high homeownership may cause unemployment reverse causality is possible. Some previous studies have used the instrumental variable (IV) method to address this concern (Coulson & Fisher, 2009; Wolf & Caruana-Galizia, 2015; Laamanen, 2013, 2017). As this study did not use the IV method, our estimates may be downward biased because of reverse causality although the use of lagged explanatory variables could partially mitigate this issue. Second, the appropriate geographic level for data aggregation, in our analysis, is unclear. Our main analyses are based on municipal-level, i.e., local administrative unit-2 (LAU-2) data, 1 which accounts for externalities within 1 LAU is a term used by the European Union to refer to a territorial subdivision (Eurostat, 2023). T. Tohmo and J. Viinikainen Home-ownership and unemployment 345 12(4), 342-355, 2023 municipal borders. Using municipality-level data is justified, for example, by the fact that zoning decisions are primarily made at this level. However, focusing on relatively constrained geographical areas, such as municipalities, may not be optimal for detecting outcomes resulting from spillover effects that extend to a wider region. Therefore, we also used wider regional aggregation, i.e., the LAU-1 data for additional analyses, which reduced the number of regions in each category as follows: university (n = 10), urban (n = 24), semi-urban (n = 23), and rural (n = 12). This classification of the LAU-1 areas was determined based on the category of the largest municipality (in terms of population) in the area. By using LAU-1 data, wider spillover effects may be uncovered, but the link between homeownership and unemployment could also be diluted, especially if the frictions only affect the unemployment rate in nearby areas. 3. Data Our longitudinal municipal-level data combined register information from three databases: 1) ALTIKA for unemployment and homeownership rates/housing structures at the LAU-2 level; 2) Statistics Finland's 'Dwelling and Housing Conditions' for housing at the LAU-1 level; and 3) SOTKA for education, age structure, and total net costs of public services per capita. The data cover the years from 1998 to 2011 in our main analyses. In the robustness analyses, using the LAU-1 level data, the data cover the years from 2005 to 2020 due to data availability. 4. Results Table 1 displays descriptive statistics by municipality type. Table 1. Descriptive statistics. University municipalities Mean (SD) Urban municipalities Mean (SD) Semi-urban municipalities Mean (SD) Rural municipalities Mean (SD) Home ownership rate 0.513 (0.039) 0.632 (0.074) 0.674 (0.055) 0.671 (0.057) Unemployment rate 12.571 (3.848) 10.279 (4.724) 11.183 (4.389) 10.567 (5.931) Educational level 349.050 (41.681) 300.869 (51.788) 264.431 (35.790) 229.740 (35.705) Municipal net costs per capita (€) 3 856.179 (777.674) 3 702.262 (815.966) 3 747.718 (809.641) 3 957.749 (969.665) Share of 16–24-year-olds 0.134 (0.011) 0.107 (0.013) 0.101 (0.013) 0.093 (0.016) Population 166 822 (145 905) 40 107 (36 386) 13 120 (5 415) 4 448 (3 055) Out-migration (share) 0.051 (0.006) 0.050 (0.013) 0.050 (0.031) 0.045 (0.012) In-migration (share) 0.053 (0.071) 0.051 (0.017) 0.048 (0.036) 0.040 (0.014) 7–29 m2 home-owned dwellings (share of total housing stock) 0.007 (0.006) 0.005 (0.003) 0.005 (0.002) 0.006 (0.003) 30–59 m2 home-owned dwellings (share of total housing stock) 0.122 (0.022) 0.108 (0.030) 0.099 (0.021) 0.090 (0.030) 60–89 m2 home-owned dwellings (share of total housing stock) 0.173 (0.013) 0.188 (0.029) 0.180 (0.021) 0.166 (0.022) 90–119 m2 home-owned dwellings (share of total housing stock) 0.122 (0.026) 0.178 (0.043) 0.209 (0.030) 0.210 (0.036) ≥120 m2 home-owned dwellings (share of total housing stock) 0.087 (0.023) 0.149 (0.059) 0.176 (0.050) 0.190 (0.051) Number of municipalities 10 51 66 209 Note. The means and standard deviations (SD) refer to the average values over the 1998–2011 period. T. Tohmo and J. Viinikainen Home-ownership and unemployment 346 12(4), 342-355, 2023 Homeownership rates in urban, semi-urban, and rural areas range between 60–70% but are lower (around 50%) in university municipalities. The percentage of upper-medium-size and large home-owned dwellings is also smaller in university municipalities. University municipalities have higher education levels, unemployment rates, and in-migration, as well as a younger population. Rural areas experience the lowest levels of both in and out-migration. First, we regressed the regional unemployment rate on the share of different-size home-owned dwellings using the LAU-2 categorization. The models were separately estimated for university (Table A1, Appendix), urban (Table A2, Appendix), semi-urban (Table 2), and rural (Table A3, Appendix) municipalities. Table 2. Homeownership rate and unemployment by dwelling size in semi-urban municipalities, 1998–2011. Dependent variable: ln unemployment (municipal unemployment rate in year t) (1) (2) (3) (4) Panel A: First lag Ln home ownershipt-1 7–29 m2 0.013 Ln home ownershipt-1 30–59 m2 0.064 Ln home ownershipt-1 60–89 m2 0.036 Ln home ownershipt-1 90–119 m2 -0.037 Ln home ownershipt-1 ≥120 m2 0.294** Panel B: Second lag Ln home ownershipt-2 7–29 m2 0.029 Ln home ownershipt-2 30–59 m2 0.054 Ln home ownershipt-2 60–89 m2 0.080 Ln home ownershipt-2 90–119 m2 0.0004 Ln home ownershipt-2 ≥120 m2 0.285** Panel C: Third lag Ln home ownershipt-3 7–29 m2 0.039 Ln home ownershipt-3 30–59 m2 -0.113 Ln home ownershipt-3 60–89 m2 0.091 Ln home ownershipt-3 90–119 m2 -0.094 Ln home ownershipt-3 ≥120 m2 0.256** Panel D: Fourth lag Ln home ownershipt-4 7–29 m2 0.032 Ln home ownershipt-4 30–59 m2 -0.132 Ln home ownershipt-4 60–89 m2 0.064 Ln home ownershipt-4 90–119 m2 -0.015 Ln home ownershipt-4 ≥120 m2 0.274** Observations 858 792 726 660 R2 (within) 0.891 0.869 0.855 0.845 Note. All models controlled for the first lag of unemployment rate, education and age structure, total net costs of public services, year fixed effects and municipality fixed effects. * Significant at 10%; ** significant at 5%; *** significant at 1%. Homeownership rates and unemployment showed either no significant relationship in university, urban and rural areas or, with a few exceptions, the relationship was negative, contradicting Oswald's hypothesis. However, in semi-urban areas, the results indicated a robust positive association between the share of large (≥ 120 m2) home-owned dwellings and unemployment. Thus, in semi-urban areas, with a high proportion of large-size home-owned dwellings, the results align with Oswald's hypothesis. We also estimated the Table 2 model, without the net costs variable, as it has not been commonly used in prior research and may be considered a bad control (Angrist & Pischke, 2009: 64–68). The results (Table A4, Appendix) were consistent with those presented in Table 2. T. Tohmo and J. Viinikainen Home-ownership and unemployment 347 12(4), 342-355, 2023 Table A5 (Appendix) presents the results using the LAU-1-level categorization. In urban regions, there seems to be a positive correlation between homeownership and the unemployment rate, which supports Oswald's hypothesis. However, while we found that at the LAU-2 level, a higher share of large, owner-occupied dwellings was related to higher unemployment in semiurban municipalities, we no longer find consistent significant relationships between the shares of different dwelling sizes and unemployment in either urban (Table A6, Appendix) or semiurban (as presented in Table A7, Appendix) areas. The results in Table 3 show the association between the share of large home-owned dwellings and migration flows in semi-urban, i.e., the LAU-2 municipalities. We did not find a significant relationship between the share of large home-owned dwellings and migration flows. Although a higher proportion of large home-owned dwellings was associated with a higher unemployment rate in semi-urban areas, the migration flows at the aggregate level did not appear to account for this relationship. Table 3. Homeownership rate and migration flow in semi-urban municipalities, 1998–2011. ln out-migration ln in-migration Ln home ownership (≥ 120 m2 dwellings) 0.221 -0.254 Observations 910 910 R2 (within) 0.132 0.352 Note. All models controlled for the first lag of unemployment rate, education and age structure, total net costs of public services, year fixed effects, and municipality fixed effects. * Significant at 10%; ** significant at 5%; *** significant at 1%. Information on migration was not available for one semi-urban municipality, resulting in a reduction of 14 observations. 5. Concluding remarks In this study, we examined the Oswald hypothesis, which states that a high homeownership rate in an area increases unemployment. Specifically, we considered how the proportion of ownership dwellings of different sizes in rural, semi-urban, and urban areas is linked to the local unemployment rate. Our results indicated that in semi-urban municipalities, with a high share of large home-owned dwellings relative to total housing stock, the unemployment rate was higher. Notably, migration at the aggregate level did not seem to explain this relationship. The semi-urban municipalities in this study are regions with 5,697–29,065 inhabitants. 2 In such regions, the labor market typically relies on a few large employers, population size is declining, and housing prices and liquidity are lower than those in bigger cities. This often results in declining dwelling prices and negative housing equity, which could prevent homeowners from moving even if they become unemployed. Although our results indicated that migration flows at the aggregate level were similar in semi-urban areas with high proportions of large home-owned dwellings and other semi-urban areas, migration patterns may differ among residents in different-size dwellings. Individual-level data would be required to further investigate whether migration patterns of those who live in large homes in semi-urban areas relate to unemployment. Finally, small populations, homogeneous communities, and direct contact with local decision-makers may have resulted in NIMBY behavior in semi-urban areas. This could have hindered the development of these regions in the past, making them more vulnerable to economic shocks over time. We did not find support for Oswald’s hypothesis in remote rural areas, possibly because of the different structure of labor markets in these regions, particularly the central role of agriculture. When using wider regional aggregation at the LAU-1 level, we found support for Oswald's 2 In 2011, the total population of Finland was 5.4 million (Official Statistics of Finland, 2012). T. Tohmo and J. Viinikainen Home-ownership and unemployment 348 12(4), 342-355, 2023 hypothesis in urban regions but no relationship between dwelling sizes and unemployment in either urban or semi-urban areas. The former finding may be due to some of the LAU-2 semiurban regions being categorized as part of urban regions at the LAU-1 level. Additionally, the inclusion of semi-urban regions in urban regions most likely dilutes the proportion of large owner-occupied dwellings, which may account for the latter finding. Overall, the relationship between large, owned dwellings and unemployment appears regionally restricted. Countries often provide subsidies to promote homeownership, as it is believed that homeownership can generate local amenities, increase social capital, and improve well-being (DiPasquale & Glaeser, 1999). Due to potential issues of reverse causality and potential omitted variables, our results may not establish a causal relationship, and more research is needed to confirm this finding. However, our findings do suggest that negative frictions from homeownership can partly offset the benefits, which should be taken into account when homeownership subsidies are considered. Acknowledgements The authors would like to thank Petri Böckerman for helpful comments. References Angrist, J.D., and Pischke, J-S. (2009) Mostly harmless econometrics: an empiricist’s companion. Princeton: Princeton University Press. Blanchflower, D.G., and Oswald, A.J. (2013) Does high home-ownership impair the labor market?, NBER working paper number w19079. National Bureau of Economic Research. Cambridge, MA. Broulíková, H.M., P. Huber, J. Montag, and Sunega, P. (2020) Homeownership, mobility, and unemployment: Evidence from housing privatization, Journal of Housing Economics, 50, 101728. Coulson, N. E., and Fisher, L. M. (2002) Tenure choice and labour market outcomes, Housing Studies 17 (1), 35-49. Coulson, N. E., and Fisher, L. M. (2009) Housing tenure and labor market impacts: The search goes on, Journal of Urban Economics 65 (3), 252-264. DiPasquale, D. and Glaeser, E.L. (1999) Incentives and social capital: Are homeowners better citizens?, Journal of Urban Economics, 45(2), 354-384. Eurostat 2023. Local administrative units (LAU). https://ec.europa.eu/eurostat/web/nuts/localadministrative-units Retrieved 29.3.2023. Henley, A. (1998) Residential mobility, housing equity and the labour market, The Economic Journal, 108(447), 414-427. Isebaert, D., F. Heylen, and Smolders, C. (2015) Houses and/or jobs: ownership and the labour market in Belgian districts, Regional Studies, 49(8), 1387-1406. Laamanen, J.-P. (2013). Home-ownership and the Labour Market: Evidence from Rental Housing Market Deregulation. Tampere Economic Working Papers, 89/2013. Laamanen, J.-P. (2017) Home-ownership and the labour market: evidence from rental housing market deregulation, Labour Economics, 48, 157-167. Machin, S., and Manning, A. (1999) The causes and consequences of longterm unemployment in Europe, Handbook of labor economics 3: 3085-3139. Morescalchi, A. (2016) The puzzle of job search and housing tenure: a reconciliation of theory and empirical evidence, Journal of Regional Science, 56(2), 288-312. Munch, J. R., M. Rosholm, and Svarer, M. (2006) Are homeowners really more unemployed?, The economic journal 116 (514), 991-1013. Munch, J. R., M. Rosholm, and Svarer, M. (2008) Homeownership, job duration, and wages, Journal of Urban Economics 63(1), 130-145. T. Tohmo and J. Viinikainen Home-ownership and unemployment 349 12(4), 342-355, 2023 Official Statistics of Finland (OSF). 2012. Population structure [e-publication]. ISSN=1797-5395. 2011. Helsinki: Statistics Finland [referred: 31.10.2022]. Oswald, A.J. (1996) A Conjecture on the Explanation for High Unemployment in the Industrialized Nations: Part I, University of Warwick, Working Paper No. 475. Oswald, A.J. (1999) The housing market and Europe's unemployment: A non-technical paper. Unpublished. University of Warwick. https://web.archive.org/web/20170329042914id_/http://www.andrewoswald.com:80/docs/homesnt.pdf Oswald, A.J. (2009) The housing market and Europe's unemployment: a non-technical paper. In van Ewijk, C. and van Leuvensteijn, M. (Eds): Homeownership and the Labour Market in Europe. Oxford University Press: Oxford, 43-51. Pehkonen, J. (1999) Unemployment and home–ownership, Applied Economics Letters, 6(5), 263-265. Van Leuvensteijn, M., and Koning, P. (2004) The effect of home-ownership on labor mobility in the Netherlands, Journal of Urban Economics 55 (3), 580-596. Appendix: Tables not included in the text Table A1. Home-ownership rate and unemployment by dwelling size in university municipalities, 1998-2011 Dependent variable: ln unemployment (municipal unemployment rate in year t) (1) (2) (3) (4) Panel A: First lag Ln home-ownershipt-1 7-29 m2 -0.122 Ln home-ownershipt-1 30-59 m2 0.098 Ln home-ownershipt-1 60-89 m2 -0.278 Ln home-ownershipt-1 90-199 m2 0.326 Ln home-ownershipt-1 ≥ 120 m2 -0.359** Panel B: Second lag Ln home-ownershipt-2 7-29 m2 -0.039 Ln home-ownershipt-2 30-59 m2 -0.025 Ln home-ownershipt-2 .60-89 m2 -0.104 Ln home-ownershipt-2 .90-199 m2 0.073 Ln home-ownershipt-2 ≥ 120 m2 -0.346** Panel C: Third lag Ln home-ownershipt-3 7-29 m2 -0.172 Ln home-ownershipt-3 30-59 m2 0.220 Ln home-ownershipt-3 60-89 m2 0.025 Ln home-ownershipt-3 90-199 m2 -0.079 Ln home-ownershipt-3 ≥ 120 m2 -0.320*** Panel D: Fourth lag Ln home-ownershipt-4 7-29 m2 -0.102 Ln home-ownershipt-4 30-59 m2 0.515 Ln home-ownershipt-4 60-89 m2 -0.225 Ln home-ownershipt-4 90-199 m2 -0.340 Ln home-ownershipt-4 ≥ 120 m2 -0.276 Observations 130 120 110 100 R2 (within) 0.951 0.938 0.935 0.930 Note. All models control for the first lag of unemployment rate, education and age structure, total net costs of public services, year fixed effects and municipality fixed effects. * Significant at 10%; ** significant at 5%; *** significant at 1%.