Ownership and Individual Preferences

What are the connections between homeownership, housing prices, and social policy preferences? The existing literature examining social policy preferences has two main strands, both built off an analysis of individuals’ labor market status. Class-based theorists argue that labor market income defines preferences over redistribution, with higher income individuals unsupportive of taxes and transfers that cut their net income (Esping-Andersen Reference Esping-Andersen1990; Huber and Stephens Reference Huber and Stephens2001, see also Meltzer and Richard Reference Meltzer and Richard1981). A more recent literature argues that labor market risk, rather than income, produces demand for social insurance, as citizens at risk of employment loss seek social spending to maintain consumption when out of the labor market (Iversen and Soskice, Reference Iversen and Soskice2001; Mares, Reference Mares2003; Rehm, Reference Rehm2011). Rehm, Hacker, and Schlesinger (Reference Rehm, Hacker and Schlesinger2012) argue that these two labor market literatures can be fruitfully combined by examining how the covariation of income and risk produces varied preferences over redistribution and social insurance. The role of asset ownership and wealth, particularly in the form of housing, has, however, been neglected in this focus on the labor market. As I argue below, this neglect is not unimportant: an individual's assets may rise in value even as their labor market income declines or they exit the labor market. If we are to understand individual support for redistribution and social insurance we need to take wealth seriously.

I conceive of wealth as the ownership of assets that can be converted into income either through sale (immediate or as an annuity) or by borrowing against them as collateral (for example, home equity) in order to smooth consumption. As such, it can be considered as a core part of an individual's “permanent income.” In the following, I presume the core asset owned by most citizens is housing, though many of the claims extend to other privately held assets (in the data analysis below I focus on housing but also examine investments where possible).

The “permanent income hypothesis,” dating back to Friedman (Reference Friedman1957) and Modigliani and Brumberg (Reference Modigliani and Brumberg1954), states that an individual's present consumption depends not on their present income, which may be affected by transitory fluctuations, but rather on their permanent income across their life, which includes their wealth. Accordingly, individual consumption should not fluctuate widely with transitory income but should be smoothed over the life cycle. Since housing adds to an individual's permanent income, a rise in the value of their house should enable individuals to raise their consumption. Moreover, this effect should be independent of an individual's transitory labor market income—indeed, citizens can sell their house or borrow against it as collateral in order to sustain consumption during periods of lower income, including retirement. Analyses that neglect the role of housing consequently miss a core element of an individual's permanent income and hence their expected living standards.

Housing serves not only as store of permanent income but also as a form of “self-insurance.” Carroll (Reference Carroll1997) adapts the permanent income hypothesis, noting that citizens may face liquidity constraints when out of the labor market and be unable to easily borrow against future income. This creates an incentive to engage in “precautionary saving” in order to maintain a “buffer-stock” of assets to ensure the ability to maintain consumption during periods of transitory income loss. Whereas assets in the traditional permanent income model exist to smooth consumption over the life cycle, the related buffer-stock argument presumes that housing also helps hedging against uncertainty in the labor market.

How housing affects an individual's permanent income depends not simply on the act of homeownership itself but also on the relative value of the house, which changes across time in real terms. Aggregate rises in house prices may boost an individual's permanent income even as their labor market income stagnates or declines. By contrast, housing busts reduce both expected permanent income and the effectiveness of housing as a buffer stock. Thus, solely examining the effects of being a homeowner without regard to housing prices is misleading. Changes in the value of an individual's housing may be unrelated not only to changes in their individual labor market income but also to broader labor market conditions. For example, Rajan (Reference Rajan2010) notes that median incomes in the USA were stagnant between 2000 and 2007, even as property prices surged. Symmetrically, fluctuations in labor market income or risk may be cushioned by stable housing prices.

Figure 1 demonstrates how changes in asset prices—in this case housing—and labor market outcomes operate distinctly. The unbroken black line reflects labor market income across the lifetime—rising when in employment and dropping to zero when out of the labor market, both before (e.g., unemployment) and after retirement. The dashed black straight line reflects permanent income over the lifetime. This income comes from both the labor market and from the housing market. The value of assets such as housing not only reflects lifetime labor earnings (richer citizens can afford more costly housing) but may also come from bequests or from aggregate changes in the value of housing unrelated to the labor market. Hence housing values can be independent of both temporary and average labor market income. Figure 1 emphasizes that an individual's transitory and permanent incomes may diverge substantially at various points, particularly when out of the labor market through unemployment or retirement. The dashed grey line reflects the possibility that an individual's permanent income may also jump (or fall) dramatically as house prices change. If the individual expects house prices to remain high (low) then that produces increased (decreased) expected permanent income and hence consumption.

FIGURE 1.

Permanent Income, Transitory Income, and Asset Price Changes over the Life Cycle

How does variation in housing prices affect preferences over redistribution and social insurance? In terms of redistribution, higher housing prices raise homeowners’ permanent income and thus reduce their expected demand for redistributive tax and transfer schemes. First, if property is taxable, homeowners may be subject to higher taxes, whether as capital gains, inheritance tax, or annual property taxation. Second, higher asset values may lead citizens to lose eligibility to receive means-tested redistributive transfers. Finally, presuming diminishing marginal returns to income, citizens with more valuable property and hence higher permanent income may have lower demand for redistributive transfers themselves. Collapsing house prices should produce greater support for redistribution among homeowners for the reverse reasons.

Housing also affects preferences over social insurance policies that provide income support during periods out of the labor market. The ability to use housing as a means to smooth consumption (through borrowing or sale) while out of the work force and to draw down on housing to fund retirement provides citizens with steady levels of consumption during periods of lower income and thus can substitute for socially provided transfers such as unemployment insurance and pensions. Homeownership thus acts as a form of self-supplied “private insurance” against the welfare losses associated with job loss, especially, per the buffer stock argument, when access to credit may be severely curtailed.Footnote ¹ Higher house prices mean more valuable private insurance and hence should lead to lower demand from homeowners for social insurance as a hedge against such risks. These responses to rising house prices should be mirrored in the case of decreasing house prices, where homeowners should become relatively more supportive of social insurance as the value of their private insurance declines. In summary, house price appreciation reduces the demand for both redistribution and social insurance; house price depreciation increases this demand.

It is worth briefly discussing a number of potential qualifications. First, unless they sell their house, individuals do not see the realized value of their asset. For many individuals the value of their house is a subjective approximation based on information gleaned about the local or national housing market. However, many analyses of demand for social insurance also rely on individuals making judgments about uncertain quantities: specific skills approaches presume that individuals are making judgments about the likelihood of finding another job that rewards their skills like their current occupation (Iversen and Soskice, Reference Iversen and Soskice2001). Thus, even though citizens typically do not know the precise value of their house, local housing conditions will shape their estimate (for example, in UK survey data self-estimates and local housing prices are strongly correlated; see below).

Second, houses are not an especially fungible form of wealth. Selling a house can be a drawn-out process, the speed of which depends on local economic conditions. Given this “specificity” in the value of housing, it is likely that the effectiveness of housing as a cushion against income loss is conditioned by the time frame in which the property must be sold. Whereas retirements are planned over a long horizon, unemployment is a rapid and often unexpected event. Hence it is likely that asset ownership reduces demand for social insurance programs for old age more so than for programs targeted at unemployment.

Third, many homeowners hold mortgages and hence hold relatively little equity in their house. Such individuals may face greater risks in affording their house and have less accumulated wealth in the form of property. Accordingly, it is among individuals with greater equity in their house that we should see the least support for redistribution and social insurance since they face less risk of losing their house and they have already accumulated equity that can substitute for social insurance.

Finally, citizens’ short-term evaluations of the price of their home may be driven, at least partly, by “irrational exuberance” rather than the long-term “fundamentals” of the housing market (Shiller, Reference Shiller2007). Indeed, there is substantial evidence that homeowners in the 2004–2006 peak of the recent housing bubble believed their homes would continue to rise in value at a substantial rate over the coming decade (Case, Shiller, and Thompson, Reference Case, Shiller and Thompson2012). Thus even if homeowners had cashed out much of the appreciated value of their home through home equality loans (HELOCs), their expectation was that they would continue to have substantial equity remaining as a nest egg. That these homeowners’ expectations proved false was unfortunate but since policy preferences will be driven by subjective price expectations, we should expect rising house prices to produce less support for social spending even when equity has been drawn down through HELOCs.

From Voters to Parties

How do these arguments about the microfoundations of ownership aggregate upwards into actual policy outcomes? To make this micro- to macroconnection we first need to look at how changing house prices filter into the electorate as a whole and then secondly theorize how parties might respond to changing preferences among the public.

Above I argued that rising house prices shift the preferences of homeowners towards reduced redistribution and social insurance spending. In the aggregate, since homeowners are a majority in most countries this should reduce overall support for social spending. However, it is also worth considering which particular homeowners are most likely to be affected and how the role of partisan ideology might impact the connection between an individual's economic fortunes and their policy preferences. In particular, I argue that the effects of house price appreciation on individual policy preferences are likely to be strongest for right-wing voters.

Recent work has argued that the way citizens interpret economic data—both aggregate and individual—may be influenced by their ideological preferences, perhaps driven by elite behavior and party priming (Bartels, Reference Bartels2002; Carsey and Layman, Reference Carsey and Layman2006). For example, in a recent study of the credit crisis, Margalit (Reference Margalit2013) finds that the effects of experiencing job loss or insecurity on preferences over unemployment insurance are much stronger among right-wing voters than among left-wing voters. The implication of this work is that left-wing voters may be primed to resist cuts to social spending regardless of other individual characteristics, ideology thereby operating as a filter on economically derived preferences. Conversely, right-wing voters, lacking these ideological commitments, are likely to be more responsive to changes in asset prices. This “ideological filter” effect implies differential responsiveness to house price changes among partisans.

Putting these expectations together, for two reasons I expect that at the macrolevel right-wing parties’ policymaking should be especially responsive to changes in house prices. First, several studies show that right-wing parties disproportionately represent homeowners (Kingston, Thompson, and Eichar Reference Kingston, Thompson and Eichar1984; Verberg, Reference Verberg2000). Hence, during house price booms I expect right-wing parties to channel the preferences of these constituents for lower social insurance spending. Left-wing parties, with constituent bases that more heavily represent nonowners, should not respond in this fashion. Second, because of the “ideological filter” effect, right-wing voters are more likely than left-wing voters to alter their social policy preferences in response to changes in house prices. This implies that rising house prices create an opportunity for right-wing parties to cut social spending as their electoral base becomes more inclined towards that ideological end. In summary, right-wing parties should be more responsive to house prices than left-wing parties since they contain more homeowners in general and right-identifying homeowners are more responsive to changes in house prices in terms of their policy preferences.Footnote ² Finally, where homeownership is higher, homeowners form a larger potential constituency of votes, including key swing voters. During house price booms this creates a greater electoral opportunity for right-wing parties to curtail social insurance programs. Hence in high homeownership countries the effect of house price booms on right-wing parties reducing support for social insurance should be amplified.

To summarize the theory developed in this section, I have argued that thinking about housing as a key element of individual's permanent income provides novel insights into both social policy preference formation and policy outcomes themselves. Citizens will become less supportive of both redistribution and social insurance the greater the value of their house. Political parties will respond to changes in support for social policy brought about by changes in house prices. In particular, right wing parties, with a support base disproportionately composed of homeowners, will be especially likely to cut back social programs during housing booms.

Housing and Policy Preferences in the United States

I begin by examining preferences over the funding of the Social Security retirement program in the United States using the ANES 2000–2004 panel survey. The main dependent variable is a three-point scale asking whether individuals would like to see federal spending on Social Security decrease, remain the same, or increase. To examine change over time I convert this variable into a change measure, which is accordingly a five-point scale (ranging from someone who initially wanted to increase spending in 2000 but to decrease it in 2004 to the reverse scenario).

In terms of independent variables, I examine a variety of indicators of house price appreciation, with the expectation that homeowners who have experienced rising local house prices will become less supportive of Social Security over the 2000–2004 period. I draw on data at the Metropolitan Statistical Area (MSA) and U.S. state level provided by the Federal Housing Finance Agency (FHFA). These data come in two types: Purchase Only data which is calculated from same-dwelling sales and All Transactions data which includes both purchases and refinancing.Footnote ³ I adjust these data for consumer price inflation so that they measure real appreciation. As my core measure of house price appreciation, I create a variable that combines information on whether the respondent is a homeowner across the full 2000–2004 period multiplied by the regional increase in house prices. I examine (a) state-level appreciation (Δ St. House Price) and (b) state/MSA appreciation (Δ St./MSA House Price), where I employ the state-level variable for those respondents without MSA-level data. These variables equal zero for renters and the percent change in state (or state/MSA) house prices for homeowners. Thus this measure represents a respondent's expected gain in wealth due to regional house price appreciation across the panel.

Table 1 examines how changes in regional house prices for American homeowners between 2000 and 2004 affected their support for Social Security. As control variables I include changes in income, homeownership status, party ID, and retirement status, and measures for the unchanging variables of gender, race, and age in 2000. Models 1 and 2 use the Purchase Only state-level data, models 3 and 4 use the Purchase Only state or MSA data, models 5 and 6 use the All Transactions state-level data, and models 7 and 8 use the All Transactions state or MSA data. Even-numbered models add dummies for state or MSAs and are thus fixed effects model netting out differences between regions driven by economic, political, or social conditions.

TABLE 1.

Panel Analysis of Change in Social Security Preferences: ANES 2000–4

Clustered standard errors in parentheses: *p < 0.1; **p < 0.05; ***p < 0.01.

Across all specifications I find a negative effect of estimated house price appreciation between 2000 and 2004 on changes in preferences over Social Security spending. To demonstrate the substantive magnitude of these effects, using model 4, I find that homeowners who experienced house price appreciation of 10% between 2000 and 2004 (the 10th percentile among homeowners) had an estimated probability of 14.2 percent of desiring less Social Security spending in 2004 than they did in 2000. By contrast, homeowners with house price appreciation of 50% (the 80th percentile) had a 22.5% probability of desiring less spending in 2004 than they did in 2000. Since stability of preferences is extremely common (around 85% of the sample) these predicted effects of house price appreciation are quite dramatic.

If the level of Social Security preferences in 2004 is used as the dependent variable (see supplementary material), the same difference in house price appreciation is associated with a decrease from 66% to 53% in the probability of supporting more spending in 2004. Thus examining both changes and levels of preferences, we find house price appreciation has a negative effect on support for social insurance. This negative effect of house price appreciation is limited to homeowners—in a sample restricted to homeowners, regional price changes have a negative impact on changes in social security preferences, whereas among renters the effect is positive, albeit not statistically significant at conventional levels (see supplementary material).Footnote ⁴

In the supplementary material, I also examine several other measures of asset ownership: stock ownership in 2000, the change in ownership between 2000 and 2004, and whether individuals owned a personal retirement account in 2004. These measures are binary and thus do not tap how much investment citizens have nor their likely gains or losses, and hence are considerably less satisfactory than the regional house price index as a measure of asset wealth (analysis of the British Household Panel Survey below rectifies this omission). Nonetheless I do find that individuals with stock market assets in 2000, or retirement accounts in 2004, appear to become less supportive of Social Security between 2000 and 2004, consistent with the argument developed in the theory section as applied to financial assets. House price appreciation remains negatively related to changes in social security preferences at a statistically significant level.Footnote ⁵ In all, the ANES data strongly support the conjecture that citizens become less supportive of social security as their wealth, especially in housing, rises.

Housing and Policy Preferences in the United Kingdom

To check on the generalizability of these findings about policy preferences in the American context, I now turn to survey evidence from the United Kingdom, which not only provides data on individuals across time but also provides individuals’ estimates of the value of their house. The British Household Panel Survey (BHPS) includes over 50,000 citizens in a panel study for which I have data from 1991, 1993, 1995, 1997, 2000, 2004, and 2006, during which time the British housing market saw first a crash and then a long rise in house prices. During these particular surveys respondents were asked questions about their housing status (did they own or rent) and if they owned to evaluate the value of their house (referred to as House Price). Using this information I create an indicator of the change in the estimated value of their house (in £10,000 units), adjusted for consumer price inflation, from survey to survey.Footnote ⁶ I also include a second measure of assets: annual income from investments and dividends, also in £10,000 units, referred to as Investment Income. I control for homeownership, annual labor market income (in £10,000 units), gender, age, employment status, education, number of children (not shown), a time trend, and for voters of the main three political parties, partisanship.

Models 1–4 in Table 2 examine the closest question on the BHPS to tapping social insurance preferences—asking people whether they agree that the government should see to it that everyone looking for work can find work—that is, whether the government should implement a de facto “full employment” policy. This is not an ideal measure of support for social insurance but it does address government support for people at risk of unemployment or low incomes. I use as the dependent variable change from period to period in support for the prompt, using changes in the independent variables (and levels of age and gender) as explanatory variables. Since the prompt has just five values, the possible level of change is limited and I use an ordered logit model, clustering standard errors by individual respondent. Models 1 and 2 examine the effects of changes in House Price alone, whereas models 3 and 4 add Investment Income, which reduces the sample by around 15%. Models 1 and 3 exclude change in partisanship and regional dummies whereas models 2 and 4 include these, at the loss of a further 40% of the observations.

TABLE 2.

British Household Panel Survey Analysis

Panel clustered standard errors in parentheses: *p < 0.1; **p < 0.05; ***p < 0.01.

Notes: Models 1–4 are ordered logit estimations. Models 5–8 are linear estimations with panel random effects.

Across all the models I find a statistically significant negative effect of self-evaluated house price appreciation on changes in support for employment policies.Footnote ⁷ In the supplementary material I show that these effects are consistent regardless of whether the sample is limited to increases in house prices or to decreases: house price increases lead to reduced support, house price decreases lead to increased support. The substantial effect of rather rapid house price appreciation—say 50% increase over three years—is to reduce support for full employment policies (either strongly or somewhat support) by around five percentage points. Changes in Investment Income and labor market income, by contrast, do not appear related to changes in support for employment policy.

In models 5–8 of Table 2, I extend the analysis to a broader dependent variable—a compound ideological index composed of the full employment question plus similar questions about whether there is one rule for the rich and one for the poor, whether industries should be nationalized, whether private enterprise solves economic problems, whether strong trade unions protect employees, and whether ordinary people are sharing in the nation's wealth. Combining these into a 30-point compound index—increasing in left-wing ideology—permits the use of linear models and the inclusion of individual-level random effects to account for variance across respondents. Models 5–8 use an error correction model with the respondent's score on the ideological index as the dependent variable and including a lagged dependent variable and changes and lagged levels of house prices and the other independent variables.

The results indicate that both changes in house prices and their level have strong negative effects on the compound ideological index and hence are associated with more right-wing ideological views. This implies both within-individual (changes) and between-individual (levels) variation in house prices matter for policy preferences.Footnote ⁸ The long-run effects of house price changes, taking into account the lagged dependent variable, are large: a doubling in house prices is associated with a long-run decrease of two-thirds of a within-person standard deviation in ideological preferences. Unlike the case of full employment preferences, Investment Income is also strongly and negatively related to ideological preferences. An increase of £10,000 in investment income is associated with a long-run decrease of almost half a within-person standard deviation in ideological preferences.

Table 3 alters the specification by introducing measures that are intended to directly pick up permanent versus transitory housing, investment, and labor market income. Here I follow Stegmueller (Reference Stegmueller2013) with an individual's average house price across the panel and their period deviation from this average as proxies for permanent and transitory housing wealth, along with controls for permanent and transitory investment and labor market income. I find that both permanent and transitory housing wealth have negative effects on both full employment preferences and ideology as in the error correction model. Similarly, across all specifications there is a strong negative effect of permanent investment income as well as a transitory effect for ideological preferences (though not full employment). Notably these are distinct from the effects of labor market income over the panel, implying that assets have a long-run effect independent from expected labor market income.

TABLE 3.

Permanent and Transitory Income and Assets in the BHPS

Panel clustered standard errors in parentheses: *p < 0.1; **p < 0.05; ***p < 0.01.

It is revealing to consider the relative magnitude of the effects of housing wealth, investment income, and labor market income. Each variable is measured such that a one-unit increase equals £10,000. In terms of raw coefficients, the estimated effect of permanent labor market income is six to eight times as large as that of house prices for full employment preferences and four times as large for ideological preferences. The estimated effects of investment income are higher: around 14 times as high as the effect of house prices across the models. However, these ratios need to be interpreted carefully since for homeowners in the sample, the average house price is £140,040, the average income is £14,773, and the average investment income is £1,263. Adjusted by these starting points, a £10,000 increase represents very different percentage increases. Indeed, adjusted in this way, the effect of housing is substantially larger than that of the other measures: 1.5 to 2.25 times larger than labor market income and eight times larger than investment income. Moreover, whereas 75 percent of the sample have £1,000 or less in annual investment income (25 percent have essentially zero investment income), over 75 percent own houses. This suggests that the aggregate political ramifications of shocks to housing are likely to be substantially larger than those to investment income, though both push in the same direction.

The BHPS provides a further range of variables that permit analysis of different measures of house equity and their effect on preferences.Footnote ⁹ Table 4 begins in models 1 and 2 by examining a measure House Gain produced by subtracting the initial purchase cost of the house from the respondent's current valuation (these models follow the specification of models 3 and 7 in Table 2). These models produce similar effects to the baseline House Price measure, albeit of somewhat smaller magnitude. Since these gains to house prices are potentially independent of the initial affordability of the house (and thus represent an asset “windfall”)—these results are strongly supportive of an effect of housing independent of labor market income.

TABLE 4.

Alternative Measures of Housing Equity in the BHPS

Note: Demographic variables not reported (see supplementary material).

Panel clustered standard errors in parentheses: *p < 0.1; **p < 0.05; ***p < 0.01.

Models 3 and 4 use data from the UK Land Registry on average house prices (adjusted for inflation and in £10,000 units) for 136 local education authority (LEA) districts in England dating back to 1996 (and thus excluding the first few waves of the survey). I interact the LEA mean price variable with homeownership to produce a measure similar to that used in the analysis of the ANES panel. This variable correlates at 0.52 with the respondent's estimates of individual house prices and picks up market characteristics for the local housing market as well as avoiding the potential problem of misestimation or biased estimation of house prices. It does, however, fail to pick up individual characteristics of the respondent's house not influenced by local average prices. In both models, the LEA homeowner appreciation variable has an estimated negative effect on policy preferences very similar in magnitude to that seen in models 3 and 7 of Table 2.Footnote ¹⁰

Models 5–8 examine issues related to repaying mortgages (hence I limit the analysis to homeowners). I argued above that changes in total equity should drive policy preferences, however, many homeowners still have substantial mortgages, reducing their share of equity in the house. Moreover, mortgage holders face the risk of losing the house should they miss payments. Models 5 and 6 examine a question in the BHPS asking homeowners whether they have faced difficulties paying for their house in the past two years. Respondents answering yes appear more likely to support employment policies and to have more left-wing ideological preferences but the negative effect of house price appreciation remains stable. A similar story can be seen in models 7 and 8 that add measures for the number of years left on the mortgage and the percentage of the initial purchase cost funded by borrowing. Though these measures do not appear to impact employment policy preferences, citizens with more years left to pay and who borrowed initially more tend to have more redistributive ideological preferences, even as rising house prices push them away from such attitudes. These results indicate that it is owning equity in a house, rather than the title to the house, that produces decreased support for redistribution.

Cross-National Survey Data

I now turn to cross-national survey data that allow me to address two important questions: to what extent do these survey results hold up outside of the USA and the UK; and what is the effect of a substantial decrease in property prices that leaves citizens with negative equity, as many citizens experienced during the recent housing downturn? I use the International Social Survey Program (ISSP) from 2009 to examine the effects of home equity on preferences over redistribution for over 15,000 citizens in 29 industrialized countries, many (but not all) of which had experienced significant house price declines following the peak of the housing boom. Unlike the previous surveys I do not have house price data matched to citizens nor is the ISSP a dynamic panel survey. It does, however, contain a question that taps into citizens’ equity in their house.

Specifically the ISSP asks “How much money would be left if the home you and your family live in was sold?” and presents a series of categories: just debts (negative equity); do not own; and a ten-point equity scale normalized by national currency (e.g., in the USA this runs from under $30,000 to over $900,000 in ten (differently sized) increments). Accordingly, this measure helps in examining whether the effects of housing wealth on policy preferences are symmetric in both directions: do homeowners with negative equity support redistribution more than renters; do homeowners with positive equity support redistribution less than renters (or homeowners with less/negative equity)?

As a dependent variable I use the question that best taps redistributive preferences: “Do you think the government should reduce differences in income” (a question used by Rehm (Reference Rehm2011), among others). I control for homeownership, income (logged relative to the mean), sex, age, party identification (rising in right-wing support), work status (dummies not reported), education level (not reported), number of members of the family, and religiosity. I run a series of ordered logit models with survey weights and country-clustered standard errors.Footnote ¹¹ The models then divide citizens into subsamples by house equity status (models 2–4).

Model 1 of Table 5 begins by examining all available respondents and includes both the house equity variable and a homeownership dummy. The equity variable has the expected negative sign and is statistically significant at the 1% level. The predicted substantive impact is large—whereas a homeowner with negative equity is expected to strongly support government redistribution with a probability of 34%, this probability declines to 30% for renters, 27% for homeowners with a moderate level of positive equity, and to just 21% for homeowners with a high level of positive equity. Model 2 restricts the sample to renters and citizens with negative equity, finding again that renters are substantially less supportive of government redistribution than are citizens with negative equity. Models 3 and 4 show that when we limit the sample to (a) only homeowners and (b) only those with positive equity, the negative effect on support for redistribution of rising home equity continues to hold. Accordingly, it appears that the effects of asset values operate in both directions—when assets are highly valued they discourage support of redistribution but when their value turns negative they produce greater support than the baseline of nonownership.

TABLE 5.

International Social Survey 2009: Should Government Reduce Income Gaps?

Country clustered standard errors in parentheses: *p < 0.1; **p < 0.05; ***p < 0.01.

Note: Dummies for education and employment status not reported.

Partisanship, Housing Prices, and Support for Social Insurance

I argued in the theory section that the effects of property prices on redistribution and social insurance preferences are likely to be stronger among right-wing voters than left-wing voters since, for ideological reasons, the latter group is likely to be less responsive to changes in personal economic conditions. I re-examine models from the ANES, BHPS, and ISSP, dividing voters by partisanship.

I begin in Table 6a with U.S. data from the ANES 2000-4 panel. I use two different techniques to split the sample, building off model 3 in Table 1. Models 1 and 2 split voters by whether they identified more with the Republican or Democratic Party in 2004. Only the former group shows the negative effect of house price appreciation on changes in social security preferences between 2000 and 2004. Models 3 and 4 split by changes in partisanship between 2000 and 2004. For voters who moved to the right we see a strong negative impact of house price appreciation on changes in Social Security support, whereas we do not see a similar impact among those who moved to the left.Footnote ¹² In the supplementary material, following Margalit (Reference Margalit2013), I also examine restricting the panel sample to exclude individuals whose previous policy preference was in an “extreme” category. I find that that right-wing voters are always negatively affected by house appreciation, but among left voters only those who did not want more spending in 2000 are negatively affected. Left-wing voters who wanted more spending in 2000 are unaffected by house prices—evidence for the “ideological filter” effect: these voters are unwilling to countenance cuts regardless of individual economic conditions.

TABLE 6a.

House Price Changes for Varying Partisans: USA

Table 6b undertakes a similar exercise for British voters, dividing them into those who identify with the Conservative Party and those who identify with the Labour Party. In the case of full employment preferences, there is little appreciable difference in the effects of house price appreciation between Conservative and Labour identifiers (building off model 1 of Table 2) but in terms of ideology (building off model 5 of Table 2) we again see right-wing (Conservative) identifiers having stronger effects of house price appreciation than left-wing (Labour) identifiers. In the supplementary material I again examine restricting the sample by previous support for the dependent variable. In the case of the ideology index, I find that right-wing voters are always affected by house prices regardless of their previous ideological preferences but more ideological left voters are unaffected by house prices. I also find that right-wing voters are consistently more strongly affected by house prices regardless of whether the Conservative or Labour party was in power.

TABLE 6b.

House Price Changes for Varying Partisans: UK

Table 6c replicates the partisan distinction analysis for the ISSP sample. Again right-wing voters appear more strongly affected by changes in their housing wealth in terms of their support for redistribution. This difference holds up regardless of whether we examine the whole sample (models 1 and 2), renters versus citizens with negative equity (models 3 and 4), all owners (models 5 and 6) or all owners with positive equity (models 7 and 8). In the supplementary material I show that right-wing voters are more responsive to house equity changes regardless of the current partisanship of the government in power—accordingly, this is not driven by satisfaction with the current regime producing more optimistic views of equity positions. While party types vary greatly across the 29 countries in the ISSP, this pattern suggests that the results found in the UK and USA can be extended more broadly. Indeed, excluding the case of full employment policies in the BHPS, across the remaining seven comparisons in the ANES, BHPS, and ISSP the estimated coefficient on house prices/equity is consistently 1.5 to 3 times larger for right-wing voters.

TABLE 6c.

House Price Changes for Varying Partisans: ISSP

Panel clustered standard errors in parentheses: *p < 0.1; **p < 0.05; ***p < 0.01.

Figure 2 displays the differential effects of housing equity for left and right voters in the ISSP.Footnote ¹³ It shows that the overall effect of house equity is substantially stronger for right-wing voters who drop from a 32% probability of supporting redistribution when in negative equity to 12% when having high equity. By contrast, left-wing voters only decline from 42% to 33% probability of support. Being in negative equity is associated with right-wing voters having similar preferences to left-wing voters, supporting the finding in Margalit (Reference Margalit2013) that negative economic shocks push conservative voters towards greater support of redistribution. There is substantial separation of right and left voters with high equity and the effects of moving into this category are stronger for right-wing versus left-wing voters. This supports an ideological filter effect of partisanship: right-wing voters’ redistributive preferences are strongly affected by both negative and positive house price changes whereas left-wing voters do not become substantially less supportive of redistribution when they have high levels of equity.

FIGURE 2.

Attitudes Towards Redistribution from the ISSP 2009

To summarize, the analyses in this section of the effects of homeownership and house prices on individual social policy preferences show a powerful negative relationship that previous studies of preferences have not uncovered. However, we are left with the question as to whether these preferences matter substantively—do political parties respond to the shifting views of voters?