Imputed rent tax

Taxes on imputed rent must be paid by the owner for the dwelling he occupies. This is justified by the fact that homeowners, unlike renters, do not pay any rent and therefore have an additional source of income. Especially, if at the same time the mortgage interest can be deducted from the income tax (see below), a bias in favour of homeowners emerges. The tax on imputed rent is aimed at restoring tax neutrality. In order to evaluate the amount of unpaid rent, fiscal authorities estimate a monetary use value of owner-occupied dwelling. The tax is expected to reduce the formation of homeownership. On the other hand, since the use value of housing can be considered as an additional income, the failure to collect such a tax would mean an unequal treatment of other types of income and, hence, a stimulation of the homeownership. The collection of this tax is often complicated, because the use value is difficult to assess correctly. Moreover, the absence of a tax on imputed rent represents a subsidy, which does not discriminate between the newly built and existing housing. Thus, everyone occupying one’s own dwelling can benefit from it.

Tax deductibility of mortgage payments

The possibility to deduct mortgage interests goes often hand in hand with the tax on imputed rents. It follows from the logic that the cost incurred to obtain an additional income (nonpayment of rent) must be deductible. In some countries, the possibility of mortgage interest relief exists even in the absence of an imputed rent tax. The interest deductibility makes the purchase of a home more attractive. However, this can generate the risk of speculative price bubbles.

Capital gains tax

This tax is imposed in cases when the owner makes profits resulting from the positive difference between the selling price of a dwelling and its purchasing price, provided that this difference cannot be entirely related to the improvements made to the dwelling. The capital gains tax tends to make the purchase of housing less attractive. One of the disadvantages of homeownership compared to renting is its reduced flexibility and mobility. Typically, it takes more time to sell an owner-occupied home than to terminate a rental contract. The absence of a capital gains tax could compensate for such a disadvantage and eventually make it more attractive for renters to become homeowners. On the other hand, it could create incentives for speculating with housing, since the absence of the capital gains tax for housing would make it more attractive than other assets (e.g., shares), which are subject to such a tax. This could stimulate the formation of speculative house price bubbles and, hence, make it more difficult for the low- and middle-income households to purchase homes. Therefore, the capital gains tax on housing is sometimes conceived as a speculation tax from which the owners, who really occupy their dwellings, are exempted. Being a speculative tax, the capital gains tax imposes as a rule a minimum holding period. It means that the real estate must be kept by the owner for a certain time period until it is exempted from the taxation. Table A5 reports different exemptions from the capital gains tax for each country. The capital gains tax is assumed to be applicable regardless of the holding period, save for the cases, where the owner-occupier is explicitly exempted from the tax.

VAT on the new dwellings

The VAT on newly built dwellings is added to the purchasing price of a dwelling offered for sale. As a result, housing becomes more expensive and less attractive to buy. At the same time, exactly as in the case of the imputed rent tax, the VAT for new dwellings allows treating housing similar to other goods, which are subject to VAT. Hence, the absence of the VAT on housing can be considered as a subsidy. Unlike the absence of the tax on imputed rent, the absence of the VAT stimulates the construction of new dwellings.

Leximetric approach to taxation policies

Here, we apply the methodology, which is known as leximetrics, used since at least the early 1990s to measure the intensity of governmental regulations. Leximetrics is employed in a large variety of areas of economics, such as labour markets, finance, shareholder protection, and housing. ^{Footnote 2} There are already several studies examining homeownership taxation (e.g. Wolswijk Reference Wolswijk2009; Figari et al. Reference Figari, Paulus, Sutherland, Tsakloglou, Verbist and Zantomio2012). However, none of them intends to quantify the regulations. The first researcher to quantify the housing ownership policies was Atterhög (Reference Atterhög2005). Based on expert surveys conducted in 18 countries, he built six indices (direct grants for buying, other subsidies, mortgage deduction, grant tax deduction, low property tax, and homeownership allowances) covering the period between 1970 and 2000 at decade frequency. His indices vary between 0 (no support) and 5 (very generous support). Thus, our databases partly overlap (countries, periods, and policies). However, our data have annual frequency, are based on regulation and not expert opinion, and cover a much longer period. In addition, our database and that of Atterhög share only one common policy index – the mortgage deduction.

Barrios et al. (Reference Barrios, Denis, Ivaskaite-Tamosiune, Reut and Vazquez Torres2019), on the other hand, consider five homeownership policy indicators (transfer taxes, recurrent property taxes, capital gains taxes, imputed rent taxation, and mortgage interest deduction) to show the distortions for households decisions by computing the cost of owner-occupied housing. Their policy indicators are similar to the ones that we use in this study, except for their additional implicit recurrent property taxes. Another difference is that they measure transfer and capital gains taxation in absolute terms rather than as a dummy variable. Their sample comprises 28 EU countries between 1995 and 2017.

While these approaches examine the existence or magnitude of housing taxation policies, Seelkopf et al. (Reference Seelkopf, Bubek, Eihmanis, Ganderson, Limberg, Mnaili, Zuluaga and Genschel2021) take another dimension of taxation policies into account, which is the year of introduction of the corresponding tax. They achieve this by constructing a new data set containing the year and mode of introduction of six key modern taxes (personal income tax, corporate income tax, social security contributions, inheritance tax, general sales tax, and VAT) in 220 countries between 1750 and 2018. While this is a useful database for the introduction of general tax codes, it is not specific enough for the subdomain of housing and homeownership.

Our approach

In this study, we follow the approach suggested by Kholodilin (Reference Kholodilin2020), who measures the intensity of rental market regulations worldwide over a long period. First, we conduct an overview of the relevant legislation pieces in order to extract information concerning the tax treatment of owner-occupied dwellings. ^{Footnote 3} Second, for each of the four taxation types discussed above, a binary index is constructed that equals one, if regulation is more favourable with respect to homeowners, and zero, otherwise:

(1) $${I_{jt}} = \left\{ {\matrix{ {1,\quad {\rm{if}}\;{\rm{taxation}}\;{\rm{of}}\;{\rm{type}}\;j\;{\rm{is}}\;{\rm{favorable}}\;{\rm{to}}\;{\rm{homeowners}}\;{\rm{in}}\;{\rm{period}}{\mkern 1mu} {\mkern 1mu} t} \hfill \\ {0,\quad {\rm{otherwise}}} \hfill } } \right.$$

Thus, the binary indices for the imputed rent tax, capital gains tax, and VAT are equal to 1, when homeowners are not subject to these taxes, while the binary index for interest deductibility is equal to 1, when such an option is provided to homeowners. When a regulation exists (such as capital gains taxation), but subject to major exemptions (e.g. tax exemptions for certain holding periods), we consider the regulation to not be in place (cf. Table A5 in Supplementary material for a detailed list of exemptions). The resulting binary indices are plotted in Figures 1, 2, 3, and 4 as shaded areas. Each horizontal bar corresponds to an individual country. The darker shades of grey correspond to regulations that are more beneficial for homeowners. Yellow colour denotes missing observations.

Figure 1.

Tax on imputed rent indices, 1901–2020.

Note: Black color stands for the existence of owner-beneficial taxation, and yellow shading indicates missing data.

Figure 2.

Mortgage deductibility indices, 1901–2020.

Note: Black color stands for the existence of owner-beneficial taxation, and yellow shadings indicate missing data.

Figure 3.

Capital gains tax indices, 1901–2020.

Note: Black color stands for the existence of owner-beneficial taxation, and yellow shading indicates missing data.

Figure 4.

VAT on new housing indices, 1901–2020.

Note: Black color stands for the existence of owner-beneficial taxation, and yellow shading indicates missing data.

In addition and to reduce descriptive complexity, we compute a composite homeowneship taxation attractiveness index as a simple average of binary variables:

(2) $$HOT{A_t} = {1 \over J}\sum\limits_{j = 1}^J {{I_{jt}}} $$

where $J = 4$ is the number of individual binary taxation indices. Hence, the index can vary between 0 and 1. The higher its values, the more favourable the housing taxation for homeowners.

The indices of homeownership tax attractiveness cover 37 countries in total which reflects our attempt to cover the economically most important OECD and a dozen of non-OECD countries, where data were available and accessible. Figure 5 shows their geographical distribution in 2020. Again, the shades of grey depict the degree of attractiveness of taxation, while yellow denotes countries for which no such information is available. The composite indices for individual countries are displayed in Figure 6.

Figure 5.

Map of the composite homeownership tax attractiveness index, 2020.

Note: Darker shading stands for more owner-beneficial taxation, and yellow shading indicates missing data.

Figure 6.

Composite tenure neutrality tax indices, 1901–2020.

Note: Darker shading stands for more owner-beneficial taxation, and yellow shading indicates missing data.

Finally, we also compute the degree of neutrality of homeownership taxation with respect to the housing tenure. If the tax treatment is more favourable towards homeowners, then ceteris paribus it can create an additional incentive for people to choose owning over renting. The taxation neutrality is defined through the following two cases: either the imputed rent tax is absent and mortgage payments are not deductible or the imputed rent tax is levied and mortgage deductions are allowed.

(3) $$TN{I_t} = {{I\,_t^{\rm imputed\ rent\ tax} + I\, _t^{\rm mortgage\ deductibility} - 1} \over 2}$$

Thus, the value of this index corresponding to the taxation neutrality will be equal 0. When it is below zero, taxation is biased towards renters, while when it is positive, it is biased towards homeowners.

Individual tax indices

We first assess the tax indices descriptively. In many countries, the imputed rent tax existed at earlier stages, but was lifted in the 1960–1970s, as seen in Figure 1. Some countries have never introduced it. Only three countries (Chile, India, and Turkey) have had imputed rent tax throughout the whole observation period. The picture for mortgage deductibility is much more heterogeneous. In fact, countries are divided into three groups: those that had this option over the whole period; those that never had it; and those that had it for some period (see Figure 2). In addition, several countries (Australia, Austria, Chile, France, and Russia) eliminated this benefit but later re-introduced it. In Russia, it was related to radical changes in the economic and political system that occurred in 1917 and then in 1991. A similarly heterogeneous picture is observed for the capital gains tax (see Figure 3). Finally, the VAT on new housing appears to be a relatively new development, as Figure 4 shows. In part, it has to do with relative novelty of the VAT as such. However, even under the predecessor of the VAT – sales tax – exceptions for housing often existed. In most cases, the VAT for the new housing was introduced between 1960 and 2000.

The composite index of homeownership taxation attractiveness as defined in Equation 2 is displayed in Figure 5. It shows that the attractiveness varied considerably between 1901 and 2020. Therefore, it is difficult to draw generalised conclusions. In many countries, the tax treatment of homeownership has become less attractive, for example, in Brazil, Japan, and Turkey. Other countries made their taxation more favourable for their homeowners: e.g., Finland and the US. In fact, the US taxation policy has been always biased towards homeownership, but a particularly strong increase in this bias occurred in the mid-1990s. All in all, the Anglophone countries appear to give more fiscal preference to homeowners. By contrast, in the Scandinavian countries, during the long period under inspection, the taxation has been much less favourable to homeowners. This is, in parts, due to countries like Sweden having followed an explicit tenure neutrality policy (Bengtsson Reference Bengtsson2006). In parts, countries have simply used homeownership subsidies other than fiscal ones. Moreover, since the 1990s, the situation changed in the opposite direction.

Table 1 shows the long-run averages of the four individual indices and of the composite tax neutrality index, which are calculated over the period between 1901 and 2020. It allows comparing the degree of the long-run attractiveness and neutrality of taxation of owner-occupied housing.

Table 1.

Long-term averages of taxation attractiveness and neutrality indices, 1901–2020

In most Anglophone countries (Australia, Ireland, United Kingdom, and the US), the long-run average of the attractiveness index is much higher than the long-run average for the whole sample. Portugal belongs to the countries with a very high attractiveness index. Thus, during the past 100 years, these countries have conducted taxation policies that fostered the formation of homeownership. By contrast, the lowest levels of the composite attractiveness index are observed in the Scandinavian countries (Denmark, Norway, and Sweden). There is a broad correspondence between countries with early, high homeownership and fiscal homeownership support and between countries with later, lower homeownership and the absence of such support.

The last column of Table 1 reports the long-term averages of the neutrality index as defined in Equation 3. It varies between −0.5 in Turkey (biased towards rental tenure) and 0.5 in the US (biased towards homeownership). Only six countries have negative neutrality indices (Chile, France, Luxembourg, Peru, Sweden, and Turkey) pointing to a taxation bias towards the rental tenure most of the time. In 10 countries, taxation is neutral in the long run: Canada, Germany, India, Israel, Japan, Latvia, The Netherlands, New Zealand, South Africa, and Switzerland.

Correlation with other regulation indices

Homeownership taxation does not exist in a regulatory vacuum, as governments have a wide range of regulations in place, for example, those concerning finance, labour, and product markets. Many of them are approximated by indices constructed in existing research. The country and year coverage is very different, but in order to see what kind of public policy the homeownership tax treatment is, we inductively explore the correlations between homeownership tax attractiveness and each of the 50 other indices. These indices have varying dimensions: from 5 to 200 countries and from 1 to 112 years. Among the 50 indices, 19 represent cross-sections, while the rest are panel data. Figure 7 displays only the statistically significant coefficients of correlation between our composite homeownership tax attractiveness index and various indices representing other types of governmental regulations (see the Appendix B in Supplementary material for all others). The correlation coefficients are computed for all available values of each pair of indices. Thus, if both indices have longitudinal structure, the country-specific indices are correlated for the maximum of overlapping country-years.

Figure 7.

Correlations between homeownership tax attractiveness and other regulation indices

Sources: Abiad et al. (Reference Abiad, Detragiache and Tressel2008); Armour et al. (Reference Armour, Lele, Mollica, Siems and Deakin2006); Armour et al. (Reference Armour, Deakin and Siems2016); Botero and Ponce (Reference Botero and Ponce2011); Botero et al. (Reference Botero, Djankov, La Porta, Lopez de Silanes and Shleifer2004); Deakin et al. (Reference Deakin, Lele and Siems2007); Djankov et al. (Reference Djankov, McLiesh and Shleifer2007); Freedom House (2020); Howell (Reference Howell2005); Gwartney and Lawson (Reference Gwartney and Lawson2003); Kholodilin (Reference Kholodilin2020); Lele and Siems (Reference Lele and Siems2007); Nicoletti et al. (Reference Nicoletti, Scarpetta and Boylaud1999); Pistor et al. (Reference Pistor, Raiser and Gelfer2000); Seelkopf et al. (Reference Seelkopf, Bubek, Eihmanis, Ganderson, Limberg, Mnaili, Zuluaga and Genschel2021), OECD, and own calculations.

The selected indicators come from various sources and cover different areas of governmental regulations. OECD indicators, ^{Footnote 4} concerning the individual/collective dismissals and temporary contracts, measure the strictness of employment protection legislation (EPL). Higher values signify more comprehensive EPL within a country in a certain year. Similarly, Nicoletti et al. (Reference Nicoletti, Scarpetta and Boylaud1999) use data from the OECD Employment Outlook 1999 in order to gauge EPL. Botero et al. (Reference Botero, Djankov, La Porta, Lopez de Silanes and Shleifer2004) examine employment law, industrial/collective relations law, and social security laws in 85 countries for the year 1997. Higher values indicate a higher degree of regulation and, thus, of protection of the employees. The project “Law, Finance and Development” at the Centre for Business Research at the University of Cambridge (Armour et al. Reference Armour, Deakin and Siems2016) built several indicators to measure shareholder, creditor, and worker protection (labour regulation) that take higher values for more protection of the corresponding market participants through legislation. Djankov et al. (Reference Djankov, McLiesh and Shleifer2007) also measure creditor protection in 129 countries following the approach of La Porta et al. (Reference La Porta, Lopez-de Silanes, Shleifer and Vishny1999). Again, higher values of the indices imply stronger protection of the creditor rights. Pistor et al. (Reference Pistor, Raiser and Gelfer2000) use the same approach for several transition economies. Abiad et al. (Reference Abiad, Detragiache and Tressel2008) created an index that measures financial reform, with higher values indicating more liberalisation concerning financial laws. Finally, Botero and Ponce (Reference Botero and Ponce2011) create the Rule of the Law index and Freedom House ^{Footnote 5} measures freedom in the world, again higher values for both indices indicating more rule of the law/freedom.

Despite the heterogeneous nature of these data, the resulting picture situates fiscal homeownership support in the regulatory landscape quite clearly (Figure 7). The strongest correlations are observed between the homeownership taxation index, on the one hand, and regulations concerning financial markets, labour markets, and the rule of law, on the other hand. The correlation is negative and relatively strong for labour market indices. In particular, the correlation is significant with the strength of labour market regulation (Botero et al. Reference Botero, Djankov, La Porta, Lopez de Silanes and Shleifer2004; Armour et al. Reference Armour, Deakin and Siems2016) as well as with the strictness of employment protection. The correlation is much weaker and positive for financial market indices and for rule-of-law and freedom indices.

Thus, a more generous support of homeownership is accompanied by more liberal financial regulations, stronger creditor protection, and weaker regulation of labour. This purely bivariate result is broadly in line with the asset-based welfare hypotheses which consider homeownership support as a financial-market kind of welfare that is at odds with labour-supportive kinds of welfare arrangements. The indices also capture a certain left-right dimension, where homeownership support is rather associated with regulation situated on the political right, while the associations with welfare state indicators remain insignificant. The evidence for the above welfare hypothesis is therefore mixed: homeownership tax welfare is not correlated with the conventional welfare understood as social security but rather with characteristics of asset-based welfare.

When it comes to the other housing-specific regulation, we can draw on existing regulation data measuring the intensity of rent control and tenant protection through indices of similar country and even time coverage (Kholodilin Reference Kholodilin2020). Figure 8 shows the indices of homeownership taxation attractiveness and neutrality as well as rent control. All these indices are computed as simple averages for the 37 countries under inspection (except for South Korea for which no rent control index is available) and then normalised to render the picture more readable.

Figure 8.

Homeownership tax attractiveness index vs. rent control index.

Source: Kholodilin (Reference Kholodilin2020) and own representation.

Note: The values of indices are normalised by subtracting averages and dividing by the standard deviation so that the indices have the same scale and become easier to compare. Higher values of indices correspond either to stricter rent control or to more state support of the homeownership.

The relationship between the tax attractiveness and rent control indices varies strongly over time. Until 1920, both indices moved in the same direction: stricter rent controls were accompanied by more tax benefits for homeowners. In the interwar period, the relationship changed: a certain retreat of rent controls went hand in hand with more favourable taxation of homeowners. After World War II, the state lifted rent controls and cut benefits to homeowners. However, from the mid-1990s onward, governments have shown renewed interest in fiscal support for homeowners, while rent controls kept decreasing. In 2020, the emergency measures related to the COVID-19 pandemic led to a spike in rent control regulations. Interestingly, the relationship between homeownership taxation neutrality and rent control is much more clear cut: over most of the period, they are negatively correlated. Between 1965 and 2009, the aggregate neutrality index was on an upward trend. However, after the Great Recession of 2008–2009, it declined, as a reaction to the burst of the huge speculative bubble in the housing market in many countries. Supporting homeowners or tenants has thus stood in a certain historical trade-off relationship.

Regression analysis

After these binary correlations and general trends, we turn to a multivariate analysis of the factors determining the attractiveness and neutrality of the homeownership tax treatment. Following the underlying first expectation that it follows welfare state logic, we first introduce fundamentals such as log per capita GDP (LGDP_PC), population (Lpop), state information capacity (infcap_pca), and the number of taxes (TID) introduced. This broadly corresponds to the expectation that fiscal homeownership welfare is mainly driven by economic or demographic fundamentals and the secular trend towards state and fiscal capacity. The data are retrieved from the Maddison Project Database, ^{Footnote 6} the Information Capacity Dataset (Brambor et al. Reference Brambor, Goenaga, Lindvall and Teorell2020) – an index varying between 0 and 1, where higher values imply a high information capacity ^{Footnote 7} – and Seelkopf et al. (Reference Seelkopf, Bubek, Eihmanis, Ganderson, Limberg, Mnaili, Zuluaga and Genschel2021); see Table 2 for descriptive statistics. To speak to the polity expectation, we introduce a time-invariant legal-origin variable (La Porta et al., Reference La Porta, Lopez-de Silanes, Shleifer and Vishny1998) which groups countries into a categorical variable describing their legal system (e.g., Common law, French Civil Law, etc.). We use this variable because our taxation law outcome variables are closely associated with aspects of countries’ legal system and it has been used previously to explain tax-system features (Beck et al. Reference Beck, Demirgüç-Kunt and Levine2003). The legal-origin variable itself correlates strongly with welfare typologies and even the varieties of capitalism, but its individual country classifications are less contested than these alternative typologies. Still, the variable just allows to reveal associations, not explanatory substance, and requires further interpretation below.

Table 2.

Descriptive statistics of explanatory variables

Note: N stands for the number of observations; St. dev. denotes standard deviation, Q1 and Q3 stand for the 1st and 3rd quartiles, respectively.

To further address our theoretical partisan and welfare expectations, we introduce the left-right orientation of the government, the social expenditure to GDP, and rental market regulation into the model. This is to test whether homeownership policies follow a partisan, welfare, or housing-regulation logic and limits our main analysis to advanced capitalist democracies only and excludes developing countries and countries with a socialist history during our observation period. Beyond reasons of data availability, this subsample reasonably includes only countries which are broadly comparable in terms of political and economic background conditions. We also report a model with less variables, but including a maximum of 30 countries in Appendix. The Left_right index – ranging from 0 (very left) to 10 (very right) – is computed as the average of a cabinet right/left index weighted by the share of elections seats based on the country-specific index from the ParlGov database (Döring and Manow Reference Döring and Manow2020). For robustness check, we use the leftgov index of Scheve and Stasavage (Reference Scheve and Stasavage2009): the index equals 1, if the country’s prime minister and/or president is from a left party, and 0, otherwise. Welfare expenditure is from the OECD ^{Footnote 8} and rental housing market regulation indices are from Kholodilin (Reference Kholodilin2020).

We first estimate the following fixed-effects panel data model:

(4) $${y_{it}} = \mu + \sum\limits_{k = 1}^K {{\beta _k}} x_{i,t - 1}^k + {u_i} + {v_t} + {\varepsilon _{it}}$$

where ${y_{it}}$ is one of the four homeownership tax treatments in country i in year t; μ is the intercept; $x_{it}^k$ is an explanatory variable k; ${u_i}$ are the country fixed effects; ${v_t}$ are the year fixed effects; and ${\varepsilon _{it}}$ is the random disturbance. The potential endogeneity problem (e.g., homeownership taxation and rent control can be driven by common factors) is dealt with by using the lags of the explanatory variables. Moreover, the use of lags allows accounting for the lengthy process the policymakers and legislators need to go through before the policies reacting to some shock will be adopted (Blanchard and Perotti Reference Blanchard and Perotti2002, p. 1334). Second, in addition to the fixed-effects model, we also estimate a random-effects model in order to include the time-invariant variable legal origin into the analysis.

We use ordinary least squares (OLS) regressions to estimate all models and report robust standard errors. In case of the four individual binary indices, we have discrete-choice models. These models can be estimated using binary regression, including both logit/probit or linear probability model (PLM). The latter linearly regresses a binary-dependent variable on the explanatory variables (Wooldridge Reference Wooldridge2015, p. 248 ff). The PLM is both computationally much easier to estimate and more stable than the panel logit or probit models with random effects. ^{Footnote 9}

We report the fixed- and random-effects models in Tables 3 and 4, respectively, each time for the four individual taxation variables. The models using fixed effects differentiate between country- and country-year fixed effects but do not display any qualitative differences. The “fundamentals hypothesis” suggested that basic socioeconomic and state-infrastructural factors could explain homeownership tax welfare. The coefficients of the fundamental variables generally point in the unexpected directions, but at low significance levels: the association with GDP is at times even negative, suggesting that homeownership support through fiscal exemption follows more a stimulus than a luxury-tax-gift logic. One persistent significant effect throughout models is that small states tend to make more use of fiscal homeownership measures than large ones. This could be for various reasons: small (Scandinavian) countries had their modern tax system in place earlier, they can implement redistribution measures easier given small, homogeneous populations and fiscal homeownership subsidies may face less race-to-the-bottom pressures from small countries’ stronger exposure to international trade.

Table 3.

Estimation results of fixed-effects panel data model: individual components

***p < 0.001; **p < 0.01; *p < 0.05.

Note: The estimation results are based on the data covering 19 advanced economies (Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, and United Kingdom) between 1901 and 2016.

Table 4.

Estimation results of random-effects panel data model: individual components

***p < 0.001; **p < 0.01; *p < 0.05.

Note: The estimation results are based on the data covering 19 advanced economies (Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, and United Kingdom) between 1901 and 2016.

This is also in line with the significant positive effects of information and taxation capacity on various homeownership taxation items: information capacity tends to increase the likelihood of being lenient on taxing homeowners’ capital gains. The more developed a country’s public registers and surveying apparatus, the more likely it can also use more fine-grained tax exemption policies. The development of a modern tax state, in turn, increases the likelihood of not exempting new construction from VAT and of using mortgage deductions. To use the instrument of tax exemptions, taxes and a modern tax administration have to be in place.

When it comes to partisan effects, the first finding to be highlighted is the absence of many significant effects which is surprising given the strong (centre-left) partisan expectations about traditional social security welfare. The one significant coefficient even shows the opposite sign, linking mortgage deduction policies to centre-right parties, which holds for parties’ share in parliament but not for the political position of the executive (cf. Appendix A in Supplementary material). As a result, social expenditure (but also social housing in Appendix A in Supplementary material) is rather positively correlated with items such as imputed rent which are favourable to homeowners. This is a further evidence against a simple welfare hypothesis: homeownership welfare has, if at all, a reversed partisan logic and does not necessarily work like conventional social security welfare. The negative association of imputed rent exemptions with rental regulation confirms the bivariate findings that there is a certain homeowner–tenant trade-off baked into homeownership welfare policies. The different sign of the coefficients for VAT exemptions shows that this policy is more of a construction stimulus policy, switched on in times of rent control, and less one in favour of homeowner welfare.

This becomes further visible in the strongest and most persistently significant effects which are related to the “polity hypotheses”, proxied through countries’ legal origin: relative to Common-law countries, all other jurisdictions show less fiscal homeownership support, with Scandinavian countries being the least supportive, French and German jurisdictions in between. For instance, being a Scandinavian country reduces the probability of not charging a tax on imputed rent or a capital gain tax by roughly 75–80%. The countries with German legal origin are likewise less likely to free their homeowners from the tax on imputed rent by about 30% and from capital gain tax by slightly more than 50%. With respect to the English legal origin, all other legal origins provide by 10–20% less attractive taxation treatment to the homeowners. The one indicator for which the relationship is reversed is mortgage deductions which have had a long history in Scandinavian countries and also explains their very high levels of household indebtedness, exceeding Anglophone countries’ ones (Anderson and Kurzer Reference Anderson and Kurzer2019).

Appendix A in Supplementary material shows a number of different specifications for robustness. In particular, we used alternative measures of social expenditure (such as the more housing-specific proportion of the social housing stock) which uses a subsample of countries and shows as well a complementary relationship with the neutrality index (Table A1 in Supplementary material). Using the political position of the executive (such as the extended left government index of Scheve and Stasavage Reference Scheve and Stasavage2009), however, does not reproduce the above partisan finding based on the party seat shares (Table A2 in Supplementary material). These tables also include findings on the level of the two composite neutrality and attractiveness indices which reflect the average effects of their individual components and generally point in the direction of the results based on individual components. The time of introduction and logic of index components, e.g., exempting capital gains or mortgage payments, can be very different: the former requires a capital gains tax, the latter requires a pre-existing income tax, for instance. For this reason, we prefer to present and interpret the regression effects on individual homeowner taxation items rather than the aggregate indices which sometimes tend to average out different effect sizes. Finally, the above analysis focused only on comparable advanced democracies, not least for reasons of data availability. Using less variables, we can extend both fixed- and random-effects analyses to 30 countries, including more developing countries. The results in Tables A4 and A3 in Supplementary material confirm our main findings even in this large sample.