May 4, 2016 - Negative equity and party choice in the Netherlands (2006-2012) ... shows, based on the Dutch national elections of 2006, 2010 and 2012,.
04 May 2016
Negative equity and party choice in the Netherlands (2006-2012) Stéfanie André *
HOWCOME
No. 14
Changing Housing Regimes and Trends in Social and Economic Inequality
HOWCOME Working Paper Series
Caroline Dewilde * Ruud Luijkx * Niels Spierings^ * Department of Sociology, Tilburg University ^ Department of Sociology, Radboud University Nijmegen
www.tilburguniversity.edu/howcome
Funded by the European Research Council Grant Agreement No. 283615
1
Negative equity and party choice in the Netherlands (2006-2012)
Abstract Recently, housing and housing wealth have become established as predictors of welfare attitudes like support for redistribution. However, it is unclear if people also change their political behaviour based on housing wealth. This paper shows, based on the Dutch national elections of 2006, 2010 and 2012, that housing wealth is an important predictor of party choice when housing wealth and negative equity become politically salient. We find for the 2012 elections that individuals in households in negative equity are more likely to vote pro-welfare, while respondents with some housing wealth, who are at risk of falling into negative equity, are more likely to vote pro-ownership. This finding is corroborated in our analysis of party choice change. Respondents who see their housing wealth decreased are more likely to change their vote from pro-ownership to centre or pro-welfare, indicating that they indeed respond to changing housing wealth not only in their attitudes but also in their political behaviour.
Keywords: negative equity, housing wealth, party choice, national elections, electoral participation, the Netherlands
1. Introduction Both in sociology and political science, the attention for the impact of housing on social phenomena has increased (Zavisca and Gerber, 2016), especially after the 2008 global economic crisis which affected housing markets and the economy across Europe (Schwartz, 2012). As house prices changed increasing since the 1980s and decreasing steeply after 2008 - the meaning and role of housing for households, in particular homeowners, has changed. This was particularly the case in countries that have experienced housing-finance deregulation, such as liberal welfare states, but also in a number of Scandinavian and Southern-European countries, as well as the Netherlands. For example, in the United Kingdom (UK) house-price increases combined with financial market innovation (e.g. remortgaging) resulted in an attitudinal shift from housing wealth as an asset which is transferred to the next
2
generation, to housing wealth as a resource enabling households to financially manage immediate lifecourse risks and future welfare needs (Lowe et al., 2012). All in all, changes in the value of housing wealth potentially influences a wide range of behaviours and attitudes, besides spending patterns (Ashley and Li, 2014), such as attitudes towards the welfare state or political preferences and behaviour, for example the party people vote for. In (Anglo-Saxon) political science, homeownership has regularly been treated as a determinant of voting behaviour, and homeowners are found to be more likely to vote for right-wing parties (Kelley et al., 1985, Studlar et al., 1990). According to the so-called home-voter thesis, homeownership conservatizes people (Fischel, 2005). For the UK, Hamnett (1999) also showed that gains and losses in housing wealth have political effects. More importantly, he argues that the homeownership effect on political preferences may not be uniform. Although homeowners are on average indeed more likely to vote Conservative, as this party tends to protect homeowners from taxation, homeowners with negative equity were more likely to vote against the Conservative government in the 1992-elections, because they lost trust in the government being able to manage the housing market following the housing bust of the late 1980s. This resulted in a vote that can be interpreted as either a left-wing vote or a protest vote. In the majoritarian Anglo-Saxon systems, which form the context for the largest share of the political science literature focusing on housing, the divide between left and right is quite uniform, and homeowners and other citizens have relatively few choices, also in terms of protest votes. In multiparty systems like the Dutch one, there is no necessary alignment between the economic dimension (left-right or state-market) and the moral-ethical dimension (conservative-progressive). This means that the electorate has to choose on multiple dimensions. The home-voter thesis has only been tested in two- and three-party systems; it therefore is not clear whether it also holds in other countries with different housing markets and party systems. This paper’s first contribution therefore is to empirically test the home-voter thesis in a multiparty system. Moreover, this paper sets out to test under which circumstances housing wealth influences party choice in a proportional multiparty system. We hence further theorize the political consequences of home-owning in a profoundly changed environment of a global housing-induced economic slump. 3
Reasoning from an interest-based voting model, it can be expected that homeowners vote for ‘laissezfair’ economic policies of (economically) right-wing parties, which protect homeowners and homeownership from taxation and are opposed against wealth redistribution through the welfare state. Additionally, housing market financialization has led to an increase in the potential risks as well as the potential benefits of homeownership. Consequently, in financialized regimes, housing wealth has become a central household asset, perceived as a private insurance for life and labour market risks (Ansell, 2014). These increased benefits can be expected to strengthen homeowners’ leaning towards right-wing parties. However, the owner-occupied house cannot always act as a nest-egg or safety net. For instance, insecure labour markets resulted in a substantial number of homeowners needing to depend on social transfers, while at the same time house prices declined. As more people find themselves in or at risk of negative equity, new theorization of the impact of homeownership on voting is called for. There are two most-likely reactions of homeowners with negative housing wealth. First, to vote for a party that aims to stabilize the housing market, increase sales and house prices, in order to move out of negative equity. Second, to vote for a party that is supportive of decent welfare benefits in case of negative life-events. This leads to the question which option is chosen under which circumstances by homeowners, and whether the homeownership-effect is moderated by the amount of wealth in the owned house and by housing market and life-course risks. Theorizing the impact of negative equity on voting behaviour is the second major contribution of this paper. Overall, we focus on homeowners’ party choice in a European multiparty system. Our research question reads: how and under which conditions does housing wealth influence party choice in the Netherlands in the national elections of 2006, 2010 and 2012? These three elections form a particularly interesting case. First, they took place prior, during, and at the end of the Dutch and international housing crisis, meaning that the contextual conditions vary as the number of households with negative equity increased from 14% (2006) to 34% (2014) (CBS, 2015). Due to this variation, the effect of housing wealth can be tested more robustly. Second, the political context varies because the financial benefits of homeownership in the form of mortgage interest deduction (MID) became a major electoral issue since 2010. Third, the multiparty system gives voters more options to change 4
their party preference than two- or three-party systems, which allows us to test whether previous conclusions from studies in majoritarian systems can be generalized to other systems. Below, we first provide a historical background on housing and housing policy in the Netherlands to set the scene for our theory and hypotheses which are discussed in the third section. This is followed by data and methods in the fourth section and by results in the fifth. The concluding section reflects on the results and their larger significance.
2. Setting the scene Before turning to the theory, we provide a background to the Dutch case. Homeownership has been the minority tenure in the Netherlands until the 2000s. Social housing was built after WWII to relieve housing stress, and until today there is a relatively large social rental sector (Pittini and Laino, 2011). The Dutch government has, however, given substantial support to homeownership for decades. Homeownership support was aimed at encouraging middle- and high-income tenants in the social rental sector to move to the owner-occupied sector. At the same time, the government has been selling social rental housing to sitting tenants (Boelhouwer & Neuteboom, 2003). This decreased the total social rental sector from 46% of all housing in 1986 to 29% in 2012 (Blijie et al., 2013). Currently, homeownership is the largest tenure in the Netherlands (59% in 2012), but only 10% of these households are mortgage-free. House prices increased between 1984 and 2008, partly caused by increasing credit availability through the inclusion of the second income and new mortgage products (Boelhouwer, 2002), which also increased the mortgage debt as a percentage of GDP from 40% in 1992 to 109% in 2012. However, like in many other countries, the economic crisis caused, and was reinforced, by a concomitant housing crisis (Schwartz, 2012). Unemployment increased, house prices fell, and forced house sales increased between 2008 and 2014. The number of households in negative equity increased from 14% in 2006 to 34% in 2014 (de Vries, 2014). A particular aspect of the Dutch housing market is the National Mortgage Guarantee (NHG), an upfront premium-based insurance established in 1993, which covers losses coming from non-
5
voluntary sales of the primary residence. This means that homeowners who bought their residence with NHG-coverage and are in arrears because of unemployment, disability, widowhood or relationship dissolution are eligible for cancellation of their debt in case of a non-voluntary sale. Between 2005 and 2014, the number of applications for remission of the remaining debt increased from 592 to 4799; the number of granted applications varied between 71% and 81%.i Although NHG covers losses from non-voluntary sales, this only affects a small part of the homeowners. Half of the households in negative equity are not eligible for NHG debt remission (Kennisportal, 2015), for example because the purchase price was larger than the NHG-limit, the house was bought for more than 50% with an interest-only mortgage, or because owners sell the house for other reasons than the eligibility conditions. All in all, negative equity is clearly an issue in the Netherlands, as also stated by the Dutch Central Bank (Bank, 2014). We therefore expect it to impact on party choice. Until 2006 (the Balkenende II government, a coalition of Christian-democrats (CDA), conservative liberals (VVD), and liberal-democrats (D66)), housing policy was left to social housing corporations and banks. This changed when the economic and financial crisis hit: in 2009 the first policy measures by the Balkenende IV government (CDA, with social-democrats (PvdA), and rightwing Christians (CU)) aimed at temporary increasing the guarantee limit of NHG from 265,000 to 350,000 euros, and at stimulating housing construction (Priemus, 2010, Priemus, 2013). The housing crisis also increased the saliency of housing policies. Especially MID became an electoral issue in the 2010 campaigns, when the liberals declared it ‘untouchable’, although their position became slightly more flexible in the 2012 elections. In 2010, in the eye of the economic crisis, the new minority government (Rutte I: VVD, CDA, receiving parliamentary support from the populist radical right PVV) presented major plans to revitalize the owner-occupied housing market. The transfer tax was (temporary) lowered and it became possible to deduct mortgage interest on two homes in case of double housing costs. In the rental sector, a new right-to-buy scheme was enacted, as well as an income-based raise of rents in order to motivate higher incomes to move to the owner-occupied sector. Finally, the 2012 government (Rutte II: VVD and PvdA) was the first one to address the reduction of
6
the MID. Since 2014, MID for new mortgagees is only applicable when the mortgage is paid off in 30 years through an annuities or linear mortgage.
3. Theoretical background Historically, the Dutch electorate was one of the most predictable in Western Europe, due to pillarization in socio-religious groups with corresponding political parties (Lijphart, 1968). The socialization model of voting was dominant in explaining party choice: people would vote for the party connected to their pillar. However, following secularization, this model became less important as an explanation, whilst interest-based rational choice models gained in importance and new voting issues (e.g. environmental decline) were introduced (Smits and Spierings, 2012). In this section, we argue why housing wealth and housing market risks might help to explain party choice. Moreover, we theorize how the relationship between housing wealth and party choice might differ across households and contexts.
Homeownership and the ballot box Compared to other tenures, homeownership has “ballot box consequences, because, among other things, it affects voters’ preferences for the level of public spending, taxation and interest rates.” (Schwartz and Seabrooke, 2008, p. 239). These ballot box consequences are related to several issues: pensions, housing wealth, and welfare policy. First, Castles (1998) argued that homeownership compensates for low public pensions, because it leads to reduced housing costs in old age. But also in asset-based welfare approaches housing wealth is propagated for alleviating later welfare needs, for a critical discussion see Malpass (2008). Second, housing wealth is the largest asset of most household, while mortgages are the largest debt in most OECD countries (Schwartz, 2012). This large mortgage debt ties the local economy, as well as individual mortgage-holders, to the global economy (Schwartz and Seabrooke, 2008). Consequently, economic shocks and increasing unemployment affect homeowners more than tenants, because not only their labour market position is at stake, but also their housing investment. Third,
7
changes in the housing market lead political parties to respond. For example, following the economic crisis, the Conservatives in the UK were anxious “to re-establish their credentials as the ‘property owning democracy’ party” (Lund, 2015). In conclusion, homeownership likely matters for party choice, but the homeownership-effect cannot be expected to be uniform, as will become clear from the discussion below.
Housing wealth and party choice Interest-based voting assumes that individuals are risk-averse, which means that when they face uncertainty or lower future revenue streams, they become more supportive of social welfare programs (Iversen and Soskice, 2001, Margalit, 2013). So far, the focus has mainly been on labour market risks, which determine individuals’ preferences for social policy that would protect them from income shocks (Iversen and Soskice, 2001, Blekesaune and Quadagno, 2003). However, an increasing amount of research focused on the effects of homeownership on policy preferences, as the owned home increasingly serves as a store of permanent income and as a form of ‘self-insurance’ against negative life events (Ansell, 2014). Several studies have shown that renters and homeowners have different policy preferences, controlling for their socio-economic background (André and Dewilde, 2016, Brunner et al., 2015, Scheve and Slaughter, 2001). For instance, André & Dewilde (2016) found that across European countries, owners prefer less redistribution by government compared to renters. Moreover, in more financialized housing regimes (with potentially higher levels of expected housing wealth accumulation), the difference between owners and renters is larger, with owners preferring significantly less redistribution. In general, it is thus expected that front-payments, mortgages and expected housing-wealth gains crowd out taxes, and thus make homeowners a natural constituency for more liberal parties that restrict redistribution and welfare state spending (Studlar et al., 1990, Kemeny, 1981). While these mechanisms focus on housing wealth, homeownership figures as a proxy for housing wealth and the concomitant feeling of security in many studies. Hardly any study has assessed the actual effect of housing wealth on redistributive preferences. A notable exception is Ansell (2014), who found homeowners to reduce their support for redistributive welfare state policies when their 8
house price increased, while homeowners who lost equity increased their support (Ansell, 2014). We expect these tendencies to become apparent in party choice too, not just in policy preferences. Furthermore, if the mechanism behind the homeownership-effect is ‘housing wealth’-based security, this effect can be expected to differ for those experiencing negative equity. Housing insecurity has much in common with fears of unemployment, causing stress, anxiety, and physical health problems (Nettleton and Burrows, 1998). Homeowners in negative equity are more at risk of losing their house when they also face labour market insecurity, increasing their preference for more and better welfare transfers, and this may change their party choice accordingly. Concretely, we expect two reactions of homeowners with negative equity with regard to party choice to be most-likely. First, they are expected to vote for a party that aims to stabilize the housing market, increase sales, and in the long run increase house prices in order to get out of negative equity and decrease their insecurity. These policies consist, among others, of mortgage interest deduction (MID), which lowers monthly housing costs. The dismantling of MID was one of the major issues in the Dutch national elections of 2010 and 2012: if the MID would be cut back, this would lead to lower lending capabilities, and in the long run lower house prices, which is not in the interest of people who already own a property. However, when confronted with negative equity, homeowners might alternatively vote for parties that are supportive of the welfare state, as these promise more generous social transfers in case of income losses caused by negative life events. These pro-welfare parties are in general also the parties that would support a large (social) rental sector, which people could use to fall back on should they fall out of the homeownership market. All in all, we expect stronger effects of housing wealth on party choice for homeowners in negative equity and homeowners with limited housing wealth - as they are at risk of falling into negative equity - than for homeowners with high housing wealth. Put differently, we argue that even though security through housing wealth matters, insecurity through housing market risks is more consequential for party choice. In a multiparty system, we thus expect that particularly homeowners in negative equity or with limited housing wealth (ceteris paribus) vote for either a party that is clearly pro-welfare or a party that is clearly pro-ownership, instead of voting for a party that is neither (a centre party). 9
Hypothesis 1: homeowners in negative equity or with limited housing wealth are more likely to vote for a party that is pro-welfare or pro-ownership, than homeowners with high housing wealth.
Who votes pro-welfare and who votes pro-ownership in negative equity? The reasoning above evidently raises the question which homeowners decide to vote pro-ownership and which decide to vote pro-welfare, instead of voting for a centre party. We argue that differential party choice might be explained by income risks, financial literacy and timing of buying. People are selected into different types of homeownership based on their socio-economic position, and although all socio-economic groups are affected by negative equity, the overall risk is larger for low-income homeowners, because they are hit harder by income losses and are less likely to benefit from homeownership (Engelhardt et al., 2010). In addition, unemployment and insecure labour market positions (Green and Hendershott, 2001) and relationship dissolution (Dewilde, 2008) form a higher risk for low-income groups. Also, these risks might cause homeowners with negative equity to vote for pro-welfare parties, because the welfare state acts as an insurance to help pay the mortgage when unemployed or disabled, and because there are programs to assist single-parent and low-income families. Therefore we expect homeowners in negative equity with high-income risks to choose for a pro-welfare party instead of a pro-ownership party. A second explanation for differential party choice is financial literacy. As the less-educated have less financial literacy skills and generally less understanding of how housing markets work, they might be more likely to vote for welfare state benefits than to vote for a pro-ownership party, in order to provide a safety-net for themselves. Welfare benefits are easier to understand than the intricacies of the housing market. Third, differential party choice might be related to the timing of purchase. Homeowners who recently purchased their home apparently had enough trust in the government and housing market to move on or up the housing ladder, despite the housing crisis. While particularly first-time homeowners are expected to get into or be in negative equity because of debt-financed transaction costs (this is possible in the Netherlands), the recent buyers can nevertheless be expected to have faith in the housing market, and thus to support the pro-ownership parties. We can thus formulate three interaction hypotheses: 10
Hypothesis 2a: the positive relationship between negative equity/limited housing wealth and prowelfare voting is stronger for homeowners who belong to a risk group. Hypothesis 2b: the positive relationship between negative equity/limited housing and pro-welfare voting is stronger for homeowners who are lower educated. Hypothesis 2c: the positive relationship between negative equity/limited housing and pro-ownership voting is stronger for homeowners who are recent buyers.
Towards a dynamic model While in the previous sections we focused on the effect of housing wealth on party choice, here we advance our analysis by looking at changes in housing markets and wealth, and how these impact on party choice and party choice change. First, it is important to realize that housing markets and economic circumstances change, which might influence housing wealth and the role of housing wealth. The onset of the economic crisis and concomitant housing crisis in 2008 increased the salience of housing in policy and of housing for people’s welfare position. Consequently, housing-related concerns might more easily outweigh other considerations of voters. In the Dutch situation, the future of the MID - previously perceived of as ‘untouchable’ - and other measures to increase house prices and sales became major issues in the 2010 and 2012 elections, while they were not in 2006. This political saliency, in combination with (the fear of) decreasing housing wealth leads us to expect that housing wealth has the largest influence on party choice in 2012. Second, also changes in housing wealth itself are important. Recent research on electoral participation showed that homeowners have a higher propensity to vote than tenants (André et al., 2015, McCabe, 2013). If indeed homeowners adapt their party choice based on housing wealth, or if decreasing housing wealth helped to mobilize homeowners to vote or to vote differently, policies will change through homeowners. This would potentially increase the inequality between homeowners and tenants, resulting in a (new) socio-economic political cleavage, as was found in the US (Barreto et al., 2007). Based on our cross-sectional hypotheses above, we can thus expect homeowners to vote proownership or pro-welfare instead of voting for a centre party in the case of entrance into negative 11
equity or housing wealth decrease, in order to reduce the risk of eviction and future welfare losses. Such a longitudinal analysis can be seen as an additional test of the theoretical mechanisms outlined in the previous sections.
Hypothesis 3: there is no effect of housing wealth on party choice in 2006, this effect is small in the 2010 election and largest in the 2012 election. Hypothesis 4: homeowners who enter negative equity or lose housing wealth, are more likely to change their vote to a party that is pro-welfare or pro-ownership, while homeowners with increasing housing wealth are more likely to vote for a centre party.
Data and methods To test our hypotheses we use the Dutch LISS-panel (Long-term Internet Study for the Social Sciences), which covers three national elections. The 2006 election predates the crisis, while the 2010 election took place during the crisis, and the 2012 election at the end of the crisis. Each respondent is asked in every wave for which party he/she has voted in the last national election. In our final sample, after excluding non-voters and missing party choice (on average 18%), we have 1,541 (2006), 1,805 (2010) and 1,760 (2012) homeowners.ii The panel was invited to answer a module on different topics (e.g. politics, income, housing) every month, and therefore the respondents with missing values are different in each wave. We used other waves to impute missing values. Furthermore, we cleaned the data on house prices and mortgage debtiii using information from all waves. For the longitudinal sample we have 427 respondents. The large attrition compared to the cross-sectional sample is caused by missing values, which accumulate with each monthly wave.
Dependent variable: party choice We classified Dutch political parties based on both homeownership (own analysis of party programs) and general welfare state policy (Manifesto Project and Chapel Hill data). Based on the party programs, three general patterns can be discerned. First, homeownership and housing have become major issues in the 2010 and 2012 elections and were hardly important in the 2006 election, when 12
housing was not even mentioned by multiple parties. Second, the major issue with regard to homeownership is the future of the MID, as all parties discuss this issue in 2012, while other issues tend to be discussed by only a few parties. Third, the oher major housing-related issue is the future of the social rental sector: right-to-buy regulation and income-based rents in social housing. Taken all information together, we could classify the parties based on their homeownership and welfare state position in three clearly distinct groups: 1) pro-welfare parties, which want to limit or abolish MID, focus on government-initiated expansion of the social rental sector and are in favour of large welfare state benefits; 2) centre parties that want to cut back MID, have attention for the social rental sector and are in favour of an average welfare state; 3) pro-ownership parties, which want to make little to no changes to MID and prefer low welfare state benefits and thus a low tax-burden. The classification of parties is shown in Table 1.iv In the first group, we find the (small) left-wing parties the greens (GL), animal party (PvdD), socialist party (SP) – and the elderly party (50plus). These parties want to drastically limit or abolish MID, want to build more social housing and are in favour of large welfare state benefits. In the second group we find the social-democrats (PvdA), liberaldemocrats (D66), small right-wing and orthodox Christians (CU and SGP), the Christian-Democrats (CDA) and the populist radical right (PVV). The first four parties want to maintain MID but think there should be an upper limit, besides restricting MID to linear and annuity-based mortgages. The last two parties are more in favour of homeownership, but at the same time offer a much larger welfarestate safety-net compared with the next group. The conservative-liberal party (VVD) is the only party in Group 3: it proposed only minimal changes to MID in the 2012 election (and declared it a nonnegotiable issue in earlier elections) and favours a small welfare state. While only one party is found in Group 3, this groups still represent a large share of the votes as the VVD was the 4th party in 2006 with 15% of the votes (the largest party got less than twice that with 27%), and it was the largest party in 2010 (20%) and 2012 (27%). Our classification differs from the classic left-right division, especially for the labour party, which is generally or historically considered left-wing. However, our specific focus on their homeownership position places them in the centre category. As we want to evaluate the impact of
13
housing wealth, this focus on housing policies is more accurate compared to a focus on general economic positions.v
[Table 1 about here]
Independent variables Housing wealth is the difference between the remaining mortgage debt and the WOZ-value (WOZ= Real Estate Appraisal Act). For 28% of the sample, we did not have this information and we classify them as ‘missing’. Because buying a house in the Netherlands often means entrance into negative equity, as the buyer’s costs are debt-financed, we calculated a threshold indicating ‘limited negative equity’, including buyer’s costs for most houses. The 20.000€ cut-off point was calculated as the average transaction costs (8%) of the average sale price (250.000€) in 2008. For savings mortgages, we estimated housing wealth based on length of residence: the amount that is theoretically in the savings account is deducted from the remaining mortgage debt. The categories are: serious negative equity (larger than -20.000€), limited negative equity (-20.000 through 0€), limited housing wealth (0€ through 20.000€), high housing wealth (>20.000€) and missing. We control for other structural and socio-economic explanations of party choice. We use the natural logarithm of the households’ financial wealth, which consists of savings and investments (missing values are mean-imputed and a missing-dummy was created). We did the same for gross monthly household income. Gender is included with male as reference category. Marital status is measured as: married or living together, divorced or separated while living alone, widowed while living alone and single. Place of residence is dummified into: very urban, moderately urban, limited urban, not very urban and rural. Unemployed/disabled is coded (1) when the bread-winner and/or the partner is unemployed and/or disabled. Education is measured in three categories (low, middle and high). The mean political trust in institutions on an 11-point scale for government, parliament, politicians and parties (Cronbach’s alpha= 0.905) is used. Political efficacy is the mean on six items measuring perceived political competence (Cronbach’s alpha=0.610). We also measure if the household head and/or partner has a temporary contract instead of both having a permanent contract. 14
We further made a total vulnerability index of being unemployed/disabled, divorced, on a temporary contract, or on a low income. Descriptive statistics are presented in Table 2.
[Table 2 about here]
Method We analyse party choice with multinomial regression analysis. We present the logit effects for each variable. Because these effects are hard to interpret, depend on the reference category, and are not strictly comparable across samples (Mood, 2010), we also report Average Marginal Effects (AME) for each category of housing wealth in the text and figures. AMEs reflect the change in party choice given a change in the level of housing wealth, holding all other variables constant at their sample means (Mood, 2010). AMEs show the likelihood in percentage points of voting for a certain party. Our interaction hypotheses are tested with Marginal Effects at Representative values (MER). Furthermore, we perform logit and multinominal logit panel fixed-effects regressions to analyse party choice change (Allison, 2009, Pforr, 2014).
Results We start with the descriptive statistics of who are in negative equity in the Netherlands in 2012, which is the year in which it was most prominent. We conclude that the young, employed, and recent buyers, with high income and high education are most likely to be in negative equity. This makes intuitive sense, since these are the people that were able to finance a large mortgage and entered the housing market at its high. For instance, respondents with a household income of 2 times modal or more (13% of the respondents) are three times more likely to be in negative equity than respondents in households with a modal income or less (4%). Respondents in negative equity are also more likely to be recent buyers (0-5 years = 19%) and not so recent buyers (5-10 years = 14%) than people who have lived in their house for 16 years or more (1%). Hardly anyone over 60 years of age is in negative equity.
[Tables 3 & 4 about here] 15
Party choice In Table 3, we present the logits for housing wealth on party choice for each year. The reference categories are high housing wealth >20.000€ on the core explanatory variable, and voting for a centre party on the categorical dependent variable. We also present the total effect of housing wealth with a χ²-test. We hypothesized that homeowners in negative equity and with limited housing wealth are more likely to either vote pro-welfare or pro-ownership. We find no controlled overall effect in the 2006 election and in the 2010 election. For 2012, however, we do find an overall effect of housing wealth on party choice. Figure 1 shows the predicted probability of voting for a certain party (in percentages) for each category of housing wealth (which sums up to 100% for each category of wealth). First, we find that people with high housing wealth are more likely to vote for a centre party (53%) than homeowners with limited wealth or in negative equity (32-41%). Second, compared to homeowners with high housing wealth, homeowners with serious negative equity are 6 percentage points more likely to vote pro-ownership, while this is 13 percentage points for homeowners with limited negative equity and 17 percentage points for homeowners with limited housing wealth. Especially respondents with limited housing wealth vote pro-ownership and are least likely to vote pro-welfare, while the difference between these two outcomes is smaller for homeowners in negative equity, indicating that they are, at least partially, more likely to vote pro-welfare and thus for a statesupported safety-net. As Table 4 shows more elaborately for 2012, we find the effects of housing wealth to change only marginally when control variables are included. In the first model, we only include the variables that are theoretically prior to housing wealth, in the second model we also include variables that could co-vary with housing wealth. Overall the results suggest that people in negative equity have higher likelihoods to divert from the centre to pro-welfare or pro-ownership parties in 2012 (Table 3 Model 1, being Table 4 Model 2) and those in risk of falling in negative equity are also more likely to vote proownership than centre compared to owners with high wealth. These findings support hypothesis 1, although only at the moment that housing wealth and negative equity became politically salient for a large part of the population, in 2012.vi 16
[Figure 1 about here]
Our interaction hypotheses are aimed at testing whether certain characteristic of respondents in negative equity (income risks, financial literacy and recent buyers) predict either pro-welfare of proownership voting, compared to voting for a centre party. To this end, we present Marginal Effects at Representative values (MER) in Figures 2, 3 and 4 (the underlying models are not shown). These show party choice probabilities for groups at risk and groups who are not (or less); the dotted lines are for the households who are at risk. For each group there are theoretically three lines in each figure, one for voting pro-welfare, one for pro-ownership, and one for a centre party, which sum up to 100%. For clarity of presentation, we only present the likelihood for pro-welfare and pro-ownership. We hypothesized that those in negative equity which are in a risk group would have a higher likelihood to vote pro-welfare instead of pro-ownership. If this is true, we would see that the lines in the figure would converge when housing wealth increases. Figure 2 shows the result regarding Hypothesis 2a, with vulnerability as indicator of risk. It shows that there are no differences between vulnerable and non-vulnerable households (based on divorce, low income, unemployment, disability or a temporary contract), in their likelihood of voting pro-welfare, based on housing wealth. Vulnerable households are less likely to vote pro-ownership, but this is the case for each category of housing wealth (no interaction). We thus have to reject hypothesis 2a. Based on the financial-literacy argument, we expected low-educated homeowners to have less understanding of how the housing market works and therefore to have a higher likelihood to vote prowelfare (Hypothesis 2b). Figure 3 indeed shows that low-educated respondents are 3 percentage points more likely to vote pro-welfare than high-educated respondents when they are in negative equity, while this reduces to 1 percentage points for homeowners with high housing wealth (i.e. an interaction effect). We do not find an interaction effect for pro-ownership voting, low-educated homeowners are, in each category of housing wealth, 4 percentage points less likely to vote pro-ownership compared to high-educated homeowners. Financial literacy - proxied by education - apparently does modestly
17
explain (part) of the pro-welfare effect but not the pro-ownership effect of negative equity. This confirms our hypothesis. Figure 4 shows that recent buyers, who were expected to be in negative equity because of debt-financed transaction costs (Hypothesis 2c), although this might have worsened due to decreasing house prices, are more likely to vote for a pro-ownership party in each category of housing wealth. There are, however, no significant differences between recent buyers and the other homeowners with respect to the association between negative equity and pro-welfare/pro-ownership voting (no interaction). We thus have to reject the hypothesis.
[Figures 2, 3 and 4 about here]
Party choice over time Last, we also study the changes in the housing market and housing wealth, and how these impact on party choice. Regarding the changing housing market and concomitant political saliency of housing policies - as illustrated by our analysis of the manifestos - we hypothesized that there is no influence of housing wealth on party choice in the 2006 election, while the effect should be largest in the 2012 election. Since we only found a significant overall effect of housing wealth on party choice in 2012, we can conclude that this is indeed the case, confirming Hypothesis 3. Regarding changes in housing wealth and party choice, we hypothesized that entry into negative equity and/or decreasing housing wealth would lead to voting pro-welfare or pro-ownership instead of voting for a centre party (Hypothesis 4). In 92% of the cases we indeed see that people who change their party choice, change between pro-welfare or pro-ownership to or from the centre, whereby the shares moving to and from the centre are roughly equal from 2006 to 2010, whereas the people changing party between 2010 and 2012 mainly moved from the centre to the pro-welfare or pro-ownership parties. Tables 5 and 6 present the logit and multinomial logit panel fixed-effects regressions with which we tested the causes of switching between parties. This method controls for time-invariant unobserved variables. In both analyses, the technique only includes the respondents who have altered 18
their party choice between elections. In Table 6, the fixed-effects logit models are presented. We find that when housing wealth increases, the odds of voting pro-ownership compared to voting pro-welfare or centre increase considerably, given that the respondent did not vote pro-ownership in the last election. In other words, when housing wealth decreased people who changed their vote were more likely to leave the pro-ownership party. Housing wealth is not an explanation for why people change their vote to a pro-welfare party (the effect is not significant), given that they did not vote pro-welfare in the last election. These results are a first indication that in case of housing wealth decreases, people who used to vote pro-ownership change their vote to a pro-welfare or centre party; while people who used to vote pro-welfare do not change their party choice based on housing wealth changes. Entry into negative equity, which was included as a separate dummy variable, does not significantly influence party choice changes. In Table 6, the results of the fixed-effects multinominal regressions are presented. We report the model with pro-ownership as the reference category, because the significant differences are found compared to this choice. We find that when housing wealth increases, the likelihood of voting prowelfare decreases compared to voting pro-ownership and the same is true for centre voting (compared to pro-ownership). Overall, we can thus conclude that, although on a limited sample, party choice change towards pro-welfare parties (from pro-ownership parties) is indeed induced by decreasing housing wealth, while party-choice change towards pro-ownership parties is induced by increasing housing wealth. In neither of these models we find an effect of falling into negative equity, however, this is based on only eight homeowners that changed their negative equity status between waves in this small sample. We can therefore only partly accept hypothesis 4.
[Tables 5 and 6 about here]
Robustness checks We performed several robustness checks. First, we estimated models including more control variables (number of children, religion, watching politics on television, political interest, and government satisfaction). This did not change the main effects of housing wealth. However, controlling for length19
of-residence and related variables (age, pensioned) changed the results because housing wealth and length of residence/age are strongly associated: on average people who owned their house longer and older people had more time to pay off the mortgage and bought in times when house prices were lower. While housing wealth is the theoretical reason to expect an effect on party choice instead of age, we decided to exclude age from the analyses. Since we also found the effects to hold in the fixedeffects panel regressions, this decision seems justified. Second, when testing for the influence of house price change instead of housing wealth change we found even stronger results then those described above. When house prices decreased (irrespective of mortgage payments), the likelihood of voting pro-welfare and the likelihood of voting pro-ownership increased compared with voting for a centre party, which gives some support for Hypothesis 4.vii
Conclusion and discussion The impact of housing, homeownership and housing wealth on social and political attitudes has received growing attention in social science (Zavisca and Gerber, 2016). In this paper, we theorized and tested the effect of housing wealth on party choice in a multi-party setting: the Dutch national elections of 2006, 2010 and 2012. Rather than testing whether people’s attitudes become more redistributive when they are in negative equity (Ansell, 2014), we analyse whether these changed attitudes are expressed in their political behaviour, i.e. party choice. We argued that homeownership does not automatically imply that people would vote for a pro-ownership anti-welfare party, as is often assumed and found in two-party systems (Studlar et al., 1990, Fischel, 2005), but depends on the amount of housing wealth. Especially in case of negative equity, when housing wealth is more politically salient, we expect effects of housing wealth on party choice, urging respondents to either vote for a party that wants to increase house prices and decrease the tax burden (pro-ownership) or to vote for a party that aims to increase welfare state benefits and thus provide a safety net (pro-welfare). We indeed find that housing wealth, when it became politically salient in 2012, affected party choice, while it did not affect party choice in 2006 and 2010 when it was less prominent on the political agenda. Most importantly, the reactions of homeowners are not uniform. People in negative equity are more likely to vote pro-welfare and pro-ownership than people with housing wealth. 20
Negative equity thus pushes them to choose between either a larger welfare state or homeownership benefits. At the same time, people with limited housing wealth (who are at risk of falling into negative equity) are most likely to vote pro-ownership, and thus place their eggs in the ‘private insurance basket’, while housing wealth is much less important for those with high housing wealth. These households with high housing wealth might feel safe enough and also tend to have higher financial wealth, which makes the issue less salient for them. Besides testing the home-voter thesis on a non-majoritarian non-Anglo-Saxon country, we also set out to refine our understanding of how and why people are influenced in their party choice by housing wealth. However, the differentiation between pro-welfare and pro-ownership voting (vis-à-vis centre voting) cannot be explained with being at risk of income losses or being a recent buyer as we hypothesized. Nonetheless, people with low financial literacy (i.e. education) are more likely to vote pro-welfare when in negative equity compared to people with high financial literacy, while they did not differ in their party choice at higher levels of housing wealth. Our additional longitudinal analyses corroborate the cross-sectional findings. When housing wealth or house prices decreased and people change their vote between elections, they become more likely to vote pro-welfare or centre, than to vote pro-ownership. In sum, this provides further indications that in case of wealth decreases, people change their vote towards a more redistributive party choice and not only change their attitude (Ansell, 2014). At the same time, house price increases induced a shift to pro-ownership voting. This means that changing house prices and housing wealth indeed are of influence on party choice and thus on election outcomes, particularly when the issue is politically salient. Our study focused on the Netherlands, before and during the housing crisis. This allowed for a more in-depth understanding of the dynamics of party choice, but it implied data limitations. Especially for the longitudinal analysis we had large panel attrition and a small number of changes in the panel and were therefore limited in explaining the ‘double reaction’ of pro-welfare and proownership voting. Further research on would be very desirable to study party choice change in more detail.
21
The effect of negative equity on party choice seems quite large, given the low objective risks in the extensive Dutch welfare state. However, unemployment benefits and welfare benefits, especially for young people, have been dramatically reduced in the last years while unemployment increased, resulting in more insecurity for a large part of the electorate. We would thus expect effects to be even larger in more liberal welfare states with fewer benefits like the UK.
22
Tables and figures Table 1. Classification of parties based on housing and welfare state policy Pro-welfare Centre Pro-ownership Group 1: Limit or abolish Group 2: Cut back on MID, Group 3: Maintain MID or MID, large social rental average social rental sector, limited changes, focus on sector, large welfare state average welfare state benefits homeownership, small welfare benefits state GroenLinks (greens) SGP (orthodox Christians) VVD (liberal-conservatives) PvdD (animal party) PvdA (labor) SP (socialist party) D66 (social liberals) 50plus (elderly party) CU (moderate Christians) CDA (Christian-democrats) PVV (freedom party)
23
Table 2. Descriptive statistics for 2006, 2010 and 2012 2006 (N=1541) Housing wealth (euro) Min Max Serious negative equity >-20.000 0 1 Limited negative equity -20.000 0.65* 0.07 0.34 0.41 Limited NE -20.000 Limited NE -20.000 Limited NE
