New partisan swings and the role of homeownership: Introduction

The last few decades have been characterized by a series of important changes in the partisan configurations of many European countries. The rise of right‐wing populism, the appearance of a new type of Left with social democratic parties endorsing reforms in favour of the liberalization and flexibilization of the labour force and the emergence of left‐wing populism are recent trends that have profoundly modified the balance of power between political forces in several countries. The main explanations for these new partisan trends rely on cultural shifts related to the revival of traditional values for right‐wing populism, linked to the search for parties embodying true left‐wing ideology for left‐wing populism (Albertazzi & Mcdonnell, Reference Albertazzi and Mcdonnell2007)Footnote ¹, or, for the new type of Left, associated with the appearance of a pragmatic shift among left‐wing voters convinced of the effectiveness of liberal economic recipes. The main issue with all these explanations is that the change of partisan preference appears out of the blue, the underlying trigger(s) remaining unexplained or only tentatively spelled out. In particular, the traditional values behind right‐wing populism are often linked to the increased presence of immigrant populations as a trigger, seen as a threat to local and national identities, while other researchers show that in reality those that have the closest interactions with immigrants have the lowest likelihood of developing conservative tendencies (e.g., Pettigrew & Tropp, Reference Pettigrew and Tropp2013).

Nevertheless, the emergence of these shifts is an empirical fact confirmed by multiple surveys. These findings, however, are almost exclusively based on cross‐sectional studies. Therefore, the dynamic changes of partisan preferences are measured only at the aggregate level without providing an analysis of how and when individuals experienced a shift within their life course. By relying on representative panel data allowing us to observe the same individuals over time, we aim to fill this gap in the literature on the causes behind such changes. Since a change in partisan preferences, especially if based on cultural explanations, implies a change in the fundamental beliefs of an individual, it should require substantial time to develop. The switch to a new partisan preference should be the final outcome of a gradual process of attitudinal change. Individual‐level panel data are required in order to measure the impact of determinants whose influence may become visible only in the long run.

Our take is that certain economic and objective factors may create a rational basis for becoming receptive to the arguments of specific political forces. In the long run, these interests may become so deeply internalized that the distinction between their original rational motivations and value driven reasons linked to traditional values may become blurry. We argue that homeownership may be one of these factors. In this paper, we examine the dynamic relationship between homeownership and partisan preferences in Germany, Switzerland and the United Kingdom. We expect homeownership to be a relevant factor affecting partisan preferences for two main reasons. First, becoming a homeowner represents a crucial transition in the life of an individual, with important social and economic ramifications. The purchase of a home is a sign of the stabilization of an individual's residential trajectory and in most cases also of familial, professional and friendship ties. Moreover, homeowners live in their (successive) owner‐occupied units for decades. For most, it constitutes their largest asset (and debt) item and housing‐related expenses are their biggest budget item. The nearly universal trend towards homeownership with rising mortgage debt and house prices has put housing back on the agenda of politics and political science (Ansell, Reference Ansell2014; Beckmann, Reference Beckmann2020). Until 2007, homeownership had been growing since the interwar period in all OECD countries, leaving only Germany and Switzerland with slight tenant majorities (Kohl, Reference Kohl2017). Since 1990, most countries saw an unprecedented boom in house prices and mortgage indebtedness, resulting in housing debt being larger than business debt and housing wealth worth more than financial wealth (Jordà et al., Reference Jordà, Schularick and Taylor2016). With housing figuring big in households’ budgets and portfolios, these trends also contributed to the rising inequality in terms of wealth‐to‐income (Bonnet et al., Reference Bonnet, Bono, Chapelle and Wasmer2014), post‐housing‐expenses Gini coefficients (Dustmann et al., Reference Dustmann, Fitzenberger and Zimmermann2018) and between tenants and homeowners (Baldenius et al., Reference Baldenius, Kohl and Schularick2020). In short, homeownership is a crucial social vector in modern societies.

Second, a rich existing literature has examined the link between homeownership and partisan preferences, conservatism in particular. With the long historical relationship between property and democracies, it should hardly be surprising that these significant and ongoing property changes have political consequences, the question being of what kind? A host of recent contributions have taken up a long‐standing presumption shared by colourful figures as diverse as Friedrich Engels (Reference Engels1872/3) or Frédérick Le Play (Reference Le Play1864) and argued that homeownership has a conservatizing influence: it makes homeowners politically more active, “better citizens” (Glaeser & DiPasquale, Reference Glaeser and DiPasquale1998), vote for conservative (and incumbent) parties (Larsen et al., Reference Larsen, Hjorth, Dinesen and Sønderskov2019) and hold anti‐welfare preferences due to financial independence gained through higher prices (Ansell, Reference Ansell2014) and the discipline of mortgage debt (Kemeny, Reference Kemeny1992). If excluded from homeownership or price gains, by contrast, people are likely to turn to populist protest parties (Adler & Ansell, Reference Adler and Ansell2019). If these studies were correct, the rise of mortgage‐encumbered, owner‐occupied but unequally distributed housing wealth could explain some of the macro‐political trends such as the persistence of conservative parties, the decline of social democracy, or the rise of populism. These explanations remain nonetheless highly speculative since they are largely based on cross‐sectional studies. Only recently have some authors gone beyond the cross‐sectional approach by modelling changes in political preferences through changes in housing variables (Ansell, Reference Ansell2014; Larsen et al., Reference Larsen, Hjorth, Dinesen and Sønderskov2019; Yoder, Reference Yoder2020). While in this paper we also use panel data to control for individual fixed effects, we go one step further by using panel data to empirically measure, to our knowledge for the first time, the link between homeownership and political attitudes as a dynamic relationship characterized by continuous attitudinal changes before and after having become a homeowner. A longitudinal approach allows us to revisit the results of existing research and to empirically test dynamic mechanisms that so far have been only theoretically postulated in quantitative studies.

Regarding country selection, although our focus on Germany, Switzerland and the United Kingdom is mainly for pragmatic reasons since they are the only countries with long‐run panel data on homeownership and political preferences, these three national contexts are also representative of different housing regimes: they represent both low‐ and high‐homeownership countries, low‐ and high‐indebtedness regimes with dualist and unitary rental markets and different house price trajectories. Based on the survey data we describe below, homeownership rates have risen from 47 to 49 per cent in Germany between 1984 and 2018, from 46 to 51 per cent between 1999 and 2018 in Switzerland and from 70 to 79 per cent between 1991 and 2017 in the United Kingdom, whereas total outstanding residential loans to income were at 71 and 101 per cent in 2019, respectively (EMF, 2020). The three countries are also all different cases of the aforementioned partisan trends.

Our findings, in short, reverse most received views in the predominantly cross‐sectional literature: throughout all countries and models, the transition into homeownership – irrespective of the levels of house‐price gains or mortgage debt – is never a game‐changing or shocking event, but is rather part of the general gradual tendencies in an individual's biography. These tendencies start years before the home purchase, in anticipation, and go on for years afterwards, through socialization in homeownership environments, without the property itself being predominant in the story. What is more, these tendencies mostly contradict existing findings in the literature. While political activity does indeed increase, in the long course towards homeownership individuals are less and less likely to vote for conservative parties. We do, however, find that in all countries the transition to homeownership lies on the path towards more support for Labour and the social democrats. As New Labour moved towards the centre in our period of study, we interpret this surprising finding as part of the new Brahmin Left (Piketty, Reference Piketty2019). Homeownership – and not just education – has been part of the embourgeoisement package which replaced the traditional working‐class constituency with higher‐income and homeowning Brahmins. If relevant at all, encumbrance with mortgage debt and house price gains, in turn, do not strengthen conservatism any further. In sum, cross‐sectionally, homeowners are a special, more conservative group, politically distinct from tenants beyond class, gender or religion. Yet, becoming a homeowner is not a political shock in one's biography, but only gradually moves an individual away from many conservatisms.

The paper is organized as follows: We first situate our research in the broader literature on recent macro‐political trends and outline the key expectations the existing housing literature has about political consequences. We then introduce our panel data and methodology before presenting the results. Our discussion elaborates on how homeownership, through anticipation and socialization, acts as a long‐stretched process and as part of a more complex social bundle of living arrangements. We conclude by highlighting the main future research avenues the paper opens.

Partisan preferences and homeownership: Theoretical framework

Operationalizing the longitudinal perspective: Data and methodology

Leads and lags analysis: Empirical strategy

In order to examine the link between homeownership and political preferences in a longitudinal perspective, we rely on a leads and lags analysis. Instead of considering a binary treatment variable, we partition both the non‐homeownership and the homeownership phase into several dummies, each one identifying a specific moment in the pre‐ and post‐homeownership trajectory. Formally, we consider the following functional form:

(1) $\begin{array}{c}{Y}_{\mathrm{it}}=\alpha +{\beta }_{-L}{H}_{i{t}_{-L}}+{\beta }_{-\left(L-1\right)}{H}_{i{t}_{-\left(L-1\right)}}+\cdots +{\beta }_{-2}{H}_{i{t}_{-2}}\\ +{\beta }_{-1}{H}_{i{t}_{-1}}+{\beta }_1{H}_{i{t}_1}+{\beta }_2{H}_{i{t}_2}+\cdots +{\beta }_{L-1}{H}_{i{t}_{L-1}}\\ +{\beta }_L{H}_{i{t}_L}+C_{\mathrm{it}}^{{}^{\prime }}\gamma +{\nu }_i+{\mu }_{\mathrm{it}},\\ \mathrm{for}i=1,2,\text{…},N\mathrm{and}t=1,2,\text{…},T\end{array}$\begin{equation} \def\eqcellsep{&}\begin{array}{l} {Y_{{\rm{it}}}} = \alpha + {\rm{\;}}{\beta _{ - L}}{H_{{\rm{i}}{{\rm{t}}_{ - L}}}} + {\beta _{ - \left( {L - 1} \right)}}{H_{{\rm{i}}{{\rm{t}}_{ - \left( {L - 1} \right)}}}} + \cdots {\rm{\;}} + {\beta _{ - 2}}{H_{{\rm{i}}{{\rm{t}}_{ - 2}}}}\\ \quad \quad \,\, + \,{\rm{\;}}{\beta _{ - 1}}{H_{{\rm{i}}{{\rm{t}}_{ - 1}}}} + {\beta _1}{{\rm{H}}_{{\rm{i}}{{\rm{t}}_1}}} + {\beta _2}{H_{{\rm{i}}{{\rm{t}}_2}}} + \cdots {\rm{\;}} + {\beta _{L - 1}}{H_{{\rm{i}}{{\rm{t}}_{L - 1}}}}\\ \quad \quad \,\, + \,{\beta _{\rm{L}}}{H_{{\rm{i}}{{\rm{t}}_L}}} + C_{{\rm{it}}}^{\rm{^{\prime}}}\gamma + {\nu _{\rm{i}}} + {\mu _{{\rm{it}}}},\\ {\rm{for}}\,i = 1,2, \ldots ,N\,{\rm{and}}\,t = 1,2, \ldots ,T \end{array} \end{equation}

the subscripts i and t represent individuals and time‐periods, respectively; Y _it the dependent variable; α an intercept term constant across individuals and time periods; L is the time‐interval we focus on before and after the homeownership transition; β_‐L, β_{‐(L ‐1)},…, β₋₂, β₋₁ the coefficients associated with the status L years, L – 1 years,…, 2 years, 1 year before becoming a homeowner, respectively; H _it‐L, H _{it‐(L ‐1)},…, H _it‐2, H _it‐1 dummy variables coded as 1 if an observation represents the status L years, L‐1 years,…, 2 years, 1 year before becoming a homeowner, respectively, and 0 otherwise; symmetrically, β₁, β₂,…, β_{L ‐1}, β_L, represent the coefficients associated with the status 1 year, 2 years,…, L‐1 years, L years after having become a homeowner; H _it1, H _it2,…, H _{itL ‐1}, H _itL dummy variables coded as 1 if an observation represents the status 1 years, 2 years,…, L‐1 years, L years after having become a homeowner; γ is the coefficient associated with the control variables described above; C _it a vector including the observed control variables described above; ν_i corresponds to all variables that affect the dependent variable and vary across individuals but not over time; μ _it represents all variables not included in the model that affect the dependent variable and that vary across individuals and over time.

We estimate this model through fixed effects, controlling hence for all time‐invariant heterogeneity between the individuals contributing to the different moments of the homeownership trajectory included in the error term ν_i and that could bias the estimation. Since it is the point furthest away from the transition of interest, we consider ‐L as a reference category against which the estimates of all other dummies are benchmarked. Since the three surveys we rely on have varying lengths, L has a different value for every survey: 25 years or more with SOEP data; 15 years or more with SHP data; 19 years or more with BHPS/UKHLS data.

Since we want to consider both the statistical and substantive significance of the effects we detect (Bernardi et al., Reference Bernardi, Chakhaia and Leopold2017), it is important to highlight that the linear approach has the advantage of providing estimates that are to be directly interpreted as average marginal effects. This is obvious when the dependent variable is numeric, since fixed effects is a linear model, but also when the dependent variable is binary, since we are implicitly using a linear probability model (LPM). Additional advantages provided by the LPM approach over non‐linear models, in particular when working with panel data, are described in online Appendix D.

The analyses are based only on the first spell of homeownership we observe in the time windows covered by the participation in the surveys of every respondent that experienced a switch from non‐homeowner to homeowner. Only these individuals contribute to the estimates of the leads and lags dummy variables and the number of observations decreases as we move away from the years close to the transition we are interested in (cf. online Appendix B, Tables B5–7). In addition, in the presence of gaps in participation, we tried to save as many observations as possible by inferring the exact moment in the homeownership trajectory whenever possible. The biggest methodological issue we face is the presence of strong multicollinearity between dummies representing consecutive moments in the attitudinal trajectory. We deal with the issue by considering sequences of pairs of years instead of single years. This decreases the number of multicollinear dummies and increases the statistical power of tests since it is based on data points composed of a larger number of observations. In order to take into account the panel structure of the data, we use cluster robust standard errors, with the individual as cluster unit.

Main and heterogeneous attitudinal trajectories: Empirical findings

We deliver the empirical findings in two parts. We start by presenting the results of the leads and lags analysis for the main sample. We then delve into the heterogeneous effects by mortgage burden in Germany and by housing value change in the United Kingdom.

Meaningful longitudinal trends: Dynamic effects analysis

The regression tables estimated through Equation (1) are available in online Appendix C, Tables C1–C10. Instead of focusing on them in the main text, we plot the estimates we are interested in, that is, the dummy variables identifying attitudinal variations before and after the homeownership transition, and examine the attitudinal trends visible for every dependent variable. The x‐axis represents pairs of years in the membership trajectory. In order to be able to visually compare the magnitude of the trends across dependent variables, we re‐scaled all non‐binary variables to a 0–10 scale, set the limits of the y‐axis to [−3; 3] for 0–10 dependent variables and to [−0.3; 0.3] for binary outcomes. On every plot, there are hence two y‐axes, one on the left, comprised between [−0.3; 0.3] referring to binary outcomes, and one on the right, having [−3; 3] boundaries for 0–10 dependent variables. Instead of adding confidence intervals that would make most of our plots no longer comprehensible and cannot be used to formally test the statistical significance of the difference between two coefficients, we provide explicit tests of the difference between the coefficients identifying the starting and the end points of the trend, whenever we identify a meaningful trend. For example, a 0.5 increase between 10 years before and 10 years after homeownership significant at the 5 per cent level is indicated as (−10 → 10: 0.5, p < 5 per cent). We comment only on ‘meaningful trends’, that is, trends showing an increasing or decreasing evolution confirmed by multiple consecutive dummies.

We start with Germany and examine the dynamic link between homeownership and the likelihood of having a partisan preference, interest in politics and the party an individual feels closer to (Figure 1). The gradual transition into homeownership leads to a continuous increase in the likelihood of having a partisan preference, starting well before homeownership and continuing after having become a homeowner (−25 → 25: 0.23, p < 5 per cent)Footnote ⁸. Homeowners also tend to become more interested in politics well before buying a home and continue on the same path after the transition (−18/−17 → 25: 0.94, p < 5 per cent). The propensity to favour the Social Democratic Party (SPD) also increases continuously (−24/−23 → 25: 0.19, p < 5 per cent), while the preference for the Christian Democratic Union (CDU) remains unaffected. The vote for right‐wing populist parties exhibits a very consistent increasing trend throughout the attitudinal trajectory that reaches a high level of statistical significance despite its low magnitude (−25 → 25: 0.079, p < 1 per cent). On the contrary, the tendency to feel close to Die Linke is not associated with any significant trend.

Figure 1. Dynamic effects of homeownership on political preferences in Germany.

Turning our attention to Switzerland, we consider the propensity to take part in federal polls, the level of interest in politics and the preferences for the two main parties (Figure 2). As in Germany, we observe an increase in the behavioural dimension of political participation through a positive impact on participation in federal polls that peaks some years after having become a homeowner (−10/−9 → 3/4: 0.50, p = 0.12). The positive influence is more continuous and stronger on the attitudinal dimension, represented by interest in politics (−15 → 15: 1.39, p < 5 per cent). In terms of partisan preferences, the continuous increase in the likelihood of voting for the SP (−14/−13 → 15: 0.17, p < 10 per cent) is mirrored by a negative trend concerning the propensity to support the SPP (‐10/‐9 → 15: −0.11, p < 10 per cent).

Figure 2. Dynamic effects of homeownership on political preferences in Switzerland.

Moving to the United Kingdom (Figure 3), we do not detect any significant trend regarding the likelihood of having a partisan preference, although interest in politics exhibits an increase especially in the pre‐homeownership phase (−18/−17 → −8/−7: 1.06, p < 5 per cent). Regarding the preference for the Labour Party, although there is a general positive trend throughout the attitudinal trajectory (−19 → 9/10: 0.17, p = 0.12), the strongest spike is observable during the homeownership transition (−2/−1 → 1/2: 0.019, p < 5 per cent). On the contrary, a continuous disaffection with the Conservative Party is visible on the graph (−10/−9 → 15/16: −0.15, p < 5 per cent). Finally, as in Germany, the right‐wing populist Party UKIP gains a continuous share of voters among homeowners (−19 → 13/14: 0.065, p < 5 per cent).

Figure 3. Dynamic effects of homeownership on political preferences in the United Kingdom.

The weight of mortgage burden and the unbearable lightness of wealth: Heterogeneous longitudinal trends

Looking at the attitudinal trajectories by mortgage burden in Germany (Figure 4), they logically concern only the post‐homeownership phase. Note that Germany is internationally rather a case of lower indebtedness with a declining trend since 2000.

Figure 4. Dynamic effects of homeownership on political preferences by mortgage burden in Germany.

Focusing on the individuals that have no mortgage burden, they experience an increased likelihood of having a partisan preference only some years after having purchased a home (7/8 → 21/22: 0.084, p = 0.19), while their level of interest in politics remains unaffected. In terms of partisan preferences, the only significant trends we detect concern the positive effect regarding a right‐wing populist vote (1/2 → 21/22: 0.056, p < 5 per cent), mirrored by a negative one regarding a left‐wing populist vote (1/2 → 23/24: −0.038, p = 0.10). Focusing on those homeowners that do have a mortgage to repay, the first tertile does not exhibit any significant trend, either in the two dimensions of political involvement or in the two outcome variables linked to SPD (although an important positive but not statistically significant trend is clearly visible) or CDU preferences. They do however experience small but consistent opposite trends in terms of their propensity to support right‐wing (1/2 → 5/6: −0.013, p < 10 per cent) or left‐wing (1/2 → 17/18: 0.046, p = 0.13) populist parties. Focusing on the second tertile, while the likelihood of having a partisan preference is not significantly affected, the level of interest in politics grows continuously after having bought a home (1/2 → 25: 1.55, p < 1 per cent). In terms of actual partisan preferences, the positive impact on the support for the SPD (3/4 → 15/16: 0.069, p = 0.19) is mirrored by a decreasing tendency to feel close to the CDU (3/4 → 25: −0.16, p < 5 per cent). These individuals also experience an increasing attraction towards right‐wing (1/2 → 15/16: 0.021, p = 0.16), but not left‐wing populist parties. The category of individuals with the highest housing debt burden relative to their income does not exhibit any significant variation in the two outcome variables related to political involvement. Looking at their partisan preferences, they only experience a continuous tendency to become attracted to right‐wing populist parties (1/2 → 25: 0.060, p = 0.17).

Finally, we focus on the heterogeneous effects by self‐evaluated changes in housing value in the United Kingdom (Figure 5). The individuals in the first tertile do not exhibit any obvious trend related to the two dimensions of political involvement (having a preferred party and interest in politics) we consider. In terms of partisan preferences, they tend to become less likely to feel close both to Labour (3/4 → 13/14: −0.19, p = 0.10) and to the Conservative Party (although not significantly in this case), while they increase their propensity to support UKIP (3/4 → 17/18: 0.071, p = 0.12). Looking at the second tertile, no political involvement effects are noticeable. In terms of preferences for the two mainstream parties, these individuals become slightly more likely to support the Labour Party (7/8 → 11/12: 0.057, p = 0.13) and strongly less likely to feel close to the Conservative Party (1/2 → 19: −0.22, p < 5 per cent). No effect on their UKIP preferences is noticeable. In the third tertile, in terms of political involvement, no meaningful trends are visible. These individuals exhibit a decreasing tendency to support the Conservative Party (1/2 → 7/8: −0.057, p = 0.15), while they become more attracted by UKIP (1/2 → 13/14: 0.052, p = 0.14).

Figure 5. Dynamic effects of homeownership on political preferences by change in housing value in the United Kingdom.

Non‐shocker: Discussion

The longitudinal perspective on the different timing effects of the transition to homeownership gives a novel view of the relationship between housing markets and political behaviour. In this discussion, we want to hold the wealth of new findings against the predominant expectations the cross‐sectional literature has formed. We first expand on our main finding regarding the very general expectation of how housing and politics are linked. Our longitudinal approach reveals, throughout, models and countries that the move into homeownership generally does come with a decade‐lasting effect on political outcomes. Housing tenure thus contains an important political dimension. Yet, it never comes as an explosive shock, but is itself part of a decade‐lasting trajectory of individuals’ political shifts. We then turn to the more substantive expectations. In a nutshell, with the exception of general political involvement, the longitudinal view generally does not confirm existing expectations as to partisan outcomes or as to the relevance of mortgage burdens and house price gains. In these cases, the story we get by looking at the dynamic link between homeownership and political stances is clearly different from the one existing cross‐sectional research describes through conditional comparisons of homeowners and tenants. This most notably holds for one great reversal we find: homeownership moves individuals in their lifecourse away from the major conservative parties and, inversely, brings them closer to (New) Labour. If we can speak of a conservatizing effect at all, then it is through an embourgeoisement process.

First of all, homeownership is a non‐shocker within people's biography. The diverse empirical results should not make us miss the forest for the trees, that is, in the majority of models and throughout countries, the transition into homeownership is an important ingredient in the long‐term shifts of respondents’ political views. Housing is therefore an important dimension to consider in general and specifically in relation to the recent rise of conservatism and populist parties. Yet, contrary to existing thinking, homeownership never is a game‐changer in an individual's lifecourse of political views and activities. It rather has decade‐long anticipation and socialization effects. In the trajectory of one's life, the purchase of a house is part of a continuous process rather than a discrete event. This holds for virtually all models, where sudden spikes before and after the homeownership transition are the great exception.

Theoretically, this empirical finding sheds light on the long‐term causal structure of housing decisions and their intricate entanglement with other lifecourse domains. We find long anticipation and socialization effects that range over years or even decades before and after the actual purchase of one's first home. Anticipation, in practical terms, can start as early as birth or early childhood, when German parents, for instance, open savings accounts and specialized building‐society accounts whose government‐subsidies specifically earmark funds for the later purchase or construction of a home. Building societies have actively marketed their products to (prospective) parents. The concrete anticipation through ex ante savings in the purview of one's first home‐purchase may continue when individuals start earning their own salary, with special savings vehicles and government subsidies geared towards young future homeowners (Börsch‐Supan & Stahl, Reference Börsch‐Supan and Stahl1991). A second anticipation effect builds up through one's early socialization as studies find that parental homeownership and the living environment one grows up in shape later tenure preferences (Lersch & Luijk, Reference Lersch and Luijkx2015). The anticipation effect is crucial in order to investigate the potential consequences of an event such as homeownership since the first purchase occurs about midway in one's life trajectory, with the United Kingdom's median first‐buy age of 31 years leading and Germany (36) and Switzerland (38) lagging. In light of the many studies which point to the early years until adolescence as crucial in forming one's worldviews, the anticipation effect should have non‐negligible weight (Alwin & Krosnick, Reference Alwin and Krosnick1991). Yet, our results also show that this is not at the cost of post‐purchase socialization effects. Socialization effects in the long durée themselves have been frequently alluded to in the literature and act primarily through the living environment and neighbourhood the new home situates one in. Neighbourhoods act as part of adults’ secondary socialization, come with a package of social resources and opportunities and are tied to consumption and mobility patterns (Kemeny, Reference Kemeny1992). All this needs to be taken into account to explain why mature individuals can still be under the spell of their home years after the initial purchase.

In addition, the presence of dynamic changes in political attitudes associated with homeownership resonates not only with the idea of homeownership as a long‐term process but also with the inertial character that existing literature has underlined for political attitudes. Both the level of political involvement (Prior, Reference Prior2010) and partisan preferences (Sears et al., Reference Sears, Lau, Tyler and Allen1980) are related to early‐life experiences and symbolic attachments that can be modified through gradual effects that materialize only through long‐term processes such as the one associated with homeownership. This makes the adoption of a longitudinal approach very relevant in any analysis linked to changing political views.

More substantively, we concur with existing cross‐sectional literature when it expects that homeownership is associated with more political involvement, both in terms of the likelihood of having a partisan preference and being interested in the political world more generally. While in Germany and in Switzerland we find statistically and substantively significant longitudinal effects on both dimensions of political involvement, in the United Kingdom, only the level of interest in politics is positively affected by the attitudinal process associated with homeownership. The absence of a significant effect on the likelihood of having a partisan preference may be explained by the presence of a binary political system making it easier to feel closer to at least one of the two parties. Indeed, the descriptive statistics in Table B4 show that most respondents in the United Kingdom have a partisan preference (84 per cent), hence decreasing the leeway for homeownership to have an additional effect.

Our longitudinal findings for homeownership's partisan effects definitely provide the biggest challenge to the existing literature, which in almost all studies highlights cross‐sectional associations of homeownership with conservative party choice and attitudes, particularly when amplified by homeowners’ mortgage burdens and wealth‐producing house price gains. Here the longitudinal view presents the completely opposite picture and does so consistently through all countries: the first home purchase is not located on a trajectory of increasing likelihood of voting for conservative parties. It even has a statistically and substantively significant negative effect for the SPP and British Conservatives and is very small and not significantly different from zero for the CDU. On the contrary, homeownership makes voting for social democratic parties more likely. This effect is positive in all three countries and of similar magnitude (around 0.2).

This is a great reversal of the predominant expectation that has always linked homeownership to conservatism. The Labour movement, ever since Engel's famous ‘Housing Question’, had feared that the gradual embourgeoisement of the working class through homeownership might move union members, social democrats and left‐wing voters to the right. This fear was perhaps strongest during the Trente Glorieuses when the rise in income, education levels, but also homeownership was particularly pronounced (Groux & Lévy, Reference Groux and Lévy1993) and its consequences hotly debated (Goldthorpe et al., Reference Goldthorpe, Lockwood, Bechhofer and Platt1967). For the three countries examined here, at least since the 1980s this initial fear of the left has not necessarily been warranted. The transition towards homeownership has turned individuals away from conservative parties and has even made the turn to socialist parties and Labour more likely. But rather than being a source of rejoicing for the traditional Left, we think that this finding adds a housing‐aspect to the central finding of Thomas Piketty's (Reference Piketty2019) recent oeuvre that reveals how the constituency of left‐wing parties has persistently changed from low‐income, low‐education working class to high‐income and high‐education. The transition of the average left‐wing voter into homeownership could be seen as the housing‐complement of the turn to the Brahmin left, which is not only educated and relatively wealthy but also no longer rents. The changing Labour constituency can also be linked to the general turn of New Labour to the right in the period of our study (Häusermann, Reference Häusermann, Manow, Palier and Schwander2018), with left‐wing parties increasingly promoting homeownership themselves (Kohl, Reference Kohl2020).

Looking at social democratic embourgeoisement from another angle, sociopsychological approaches highlight the need individuals have to find a balance between rational beliefs and normative convictions (Heider, [1958] Reference Heider2013). It is hence possible to interpret the social democratic embourgeoisement we detect as a solution the New Left provides to homeowners to find a consistent way to solve the contradictions arising from their left‐wing ideals and economic interests. On the one hand, the new liberalizing policies offered by the New Left match the economic needs of homeowners who are wealthier than the average citizen. On the other hand, these parties are still considered ideologically left‐wing oriented, which allows new homeowners to soothe their conscience by still feeling at least formally left‐wing.

Thus, the classical right‐left associations with homeownership found in existing cross‐sectional studies are longitudinally completely reversed for the centre‐left parties, but these effects do not travel any further to the more radical parties of the political spectrum. For the German radical left‐wing Die Linke, for instance, there is no discernible dynamic link with homeownership. As to markedly right‐wing parties such as the German AfD and British UKIP, homeownership has a persistently small (which is linked to the small share of voters they target), but significantly positive effect – contrary to the non‐effect or negative effect we found for the dominant centre‐right parties. The classical conservatizing effect of homeownership hence only plays out for extreme right‐wing parties, where homeownership acts perhaps less as a bulwark against an invasive welfare state than one against the perceived dangers of economic globalization or migration.

A final expectation of the existing cross‐sectional literature is that higher mortgage burdens and house price gains are associated with even more conservative voting and house price losses with populist voting. Data restrictions limit our findings here to Germany and the United Kingdom, respectively, but for these countries our longitudinal findings do not highlight any clear empirical pattern. While some barely significant trends appear, we do not observe any increasing or decreasing effects that can be systematically linked to a higher or lower mortgage burden or to an increasing or decreasing housing value. Therefore, we are not able to dynamically confirm the results of the existing literature, but we are also not able to definitely disprove them either.

Overall, the longue‐durée influence of homeownership in consolidating long‐held views can thus not be nailed down to one singular moment. It is also hard to disentangle from other long‐term biographical trends with which it is bundled. After all, working towards the first home purchase coincides with the formation of one's family and professional career. After the purchase, the housing environment locks one into a specific geographical setting that structures mobility patterns, children's education, memberships in organizations and participation in social activities all the while family and professional trajectories equally take new stable turns. This confirms some known results in lifecourse research, showing the strong interdependence between professional and parental trajectories (Widmer & Ritschard, Reference Widmer and Ritschard2009). Our findings imply that homeownership is the outcome of general processes concerning multiple dimensions of an individual's life that together gradually influence the purchase of a home and partisan preferences.

It takes time: Conclusion

Social processes take time to materialize. This platitude is often forgotten under the predominance of cross‐sectional studies based on snapshot results. The great advantage of the longitudinal view is to highlight the lagging and leading timing effects of major biographical events which, after all, can come less as a shock than as a gradual process. It is this longue‐durée perspective of individuals over time which has allowed us to trace long‐term processes such as the embourgeoisement of the left – and not the expected conservatizing effects of traditional conservative social classes. It also allowed us to detect some potential material underlying conditions of what years later can surface as the cultural shift explaining populism. In that, our paper does not only make a methodological point, but substantively suggests using long‐term material changes as a predictor for the seemingly sudden appearance of cultural and attitudinal shifts more generally.

The conservative founders of the social sciences, foremost the truly reactionary French reformer under Napoleon III, Frédérick Le Play, considered private ownership as only one of various drivers which would stabilize societies in unrest. We also saw that the transition into homeownership represents only a specific moment of a gradual trajectory of individuals’ political views. While this paper has picked homeownership as a potentially conservatizing force, other drivers might as well be part of individuals’ trajectories and Le Play notably mentioned family, labour and religion. In modern times, the formation of one's own family and the entry into the labour force are both similarly life‐marking events and part of the potentially conservatizing bundle. Along one's lifecourse, people also climb up the housing ladder and job promotion ladder, which could suggest more changes beyond the momentum of the first buy examined here. Further research could not only easily apply the approach used here to these other crucial spheres of life but could also make use of these multiple parallel sequences to figure out the complex bundles of long‐term changes in people's Weltanschauung.

In addition to being linked to other inextricable events, all of our analyses estimate average effects at the country level. Because of statistical power issues and space limitations, in this paper we provided only the general national‐level picture, which certainly hides important heterogeneous effects between specific segments of the population. In particular, the urban‐rural divide may be one of the main sources of such heterogeneity. Unfortunately, only SHP data provide us with a variable that consistently differentiates rural from urban regions. In addition, statistical power issues make it difficult to obtain reliable results for both types of areas. Nevertheless, in all three countries we see an extremely high correlation between homeownership and single‐family houses. The latter are concentrated in residential areas outside of the city centre and predominantly owner‐occupied. This means that leaving the city centre is one of the dimensions associated with the bundle of life changes that go together with homeownership. Housing tenure could hence be part of the recent renaissance of the urban‐rural divide.

Another salient source of heterogeneity is the divide between East and West Germany. Exploratory analyses show that the empirical patterns between the two regions are quite different. Nevertheless, since East German respondents represent 17.73 per cent of our observations, which are not enough to obtain reliable results, we opted to provide only average marginal effects representative of Germany as a whole. Finally, generational and social class heterogeneous effects would also deserve further investigation. In future research, similar dynamic analyses could also be used to look at housing effects on the level of local politics with more georeferenced data than ours (Yoder, Reference Yoder2020).

This strong attachment to location, family formation and labour market choice also make housing special among total household assets and should be taken into account in a more general theory of asset‐based voting and attitudes. Rather than taking overall wealth as a starting point, such a theory should decompose households’ asset portfolios and growth over time to see by how much tangible and financial, illiquid and liquid, risky and non‐risky or high‐ and low‐return assets determine voters’ choices and attitudes. The individual home rather is a material, illiquid, non‐risky and low‐return asset and entering such an asset class as described in this article may have very different political effects when compared to entering a life insurance contract or stock portfolio. With rising wealth‐to‐income ratios everywhere, of which housing assets form a dominant part, asset‐based voting will arguably gain in importance for studying future democratic processes in times of financial capitalism.