This chapter presents estimates of marginal effective tax rates (METRs) across a range of savings vehicles for 40 OECD and key partner countries to assess the effect of tax systems on the incentives individuals face to save in different forms. The results highlight significant variation in METRs across savings vehicles, with tax systems creating significant incentives to alter portfolio allocation away from that which would be optimal in the absence of taxation. Private pension funds tend to be the most tax-favoured, with owner-occupied residential property also significantly tax-favoured. In contrast to owner-occupied residential property, rental property is often subject to relatively high METRs due to the application of progressive marginal personal income tax rates, capital gains taxes and significant property taxes. Bank accounts and corporate bonds also tend to be relatively heavily taxed in many countries.
Taxation of Household Savings
Chapter 3. Marginal effective tax rates on household savings
Abstract
3.1. Introduction
An individual can choose to save in a wide range of different forms. Typical options include putting money in a bank account, pension scheme or investment fund, and purchasing bonds, equities or residential property. However, these different savings vehicles may be taxed in very different ways, potentially distorting the savings portfolio choices that individuals and households make. Given the strong empirical evidence, highlighted in Chapter 1, that taxes do affect household portfolio allocation decisions, the differential taxation of savings vehicles is likely to lead to significant welfare costs. There may be a specific policy rationale for the provision of tax concessions to certain savings vehicles, while tax differentials may also have arisen over time through piecemeal reform processes. In either case, it is important to be able to quantify the tax differentials in place so that the incentive effects of the tax system are clear, and the merits of the underlying tax provisions can be properly assessed.
Any assessment of the impact of tax systems on the incentives individuals face to save in different forms is complicated by the range of different taxes and tax designs that are applied to different savings vehicles. A focus on statutory tax rates, for example, will not capture the impact of multiple taxes on a particular savings vehicle, of deductions and variations in the tax base, of different holding periods and the potential build-up of untaxed or tax-deferred returns, and of variation in the type of return generated (e.g. interest, dividends, or capital gains). In addition, statutory tax rates do not take account of inflation which can impose a substantial additional tax on the return to savings. In contrast, marginal effective tax rate (METR) modelling enables the impact of a wide range of taxes and tax design features to be summarised in the one indicator.
This chapter presents estimates of METRs across a range of savings vehicles for 40 OECD and key partner countries to assess the effect of tax systems on the incentives individuals face to save in different forms. METR calculations are based on rules in place as of 1 July 2016 as described in country responses to a questionnaire sent to country delegates to Working Party No. 2 on Tax Policy Analysis and Tax Statistics of the OECD’s Committee on Fiscal Affairs.
The results for the 40 countries highlight significant variation in METRs across savings vehicles, with tax systems creating significant incentives to alter portfolio allocation away from that which would be optimal in a no-tax world. Some assets tend to be particularly tax-favoured as compared to others. Pension funds tend to be the most tax-favoured, with owner-occupied residential property also significantly tax-favoured. In contrast to owner-occupied residential property, rental property is often subject to relatively high METRs due to the application of progressive marginal personal income tax (PIT) rates, capital gains taxes and significant property taxes. Bank accounts and corporate bonds also tend to be relatively heavily taxed in many countries. As noted above, while there may be a clear policy rationale for the provision of some concessions, the merits of these concessions need to be weighed against the resulting welfare costs. In this regard, country-specific circumstances are likely to be particularly important.
The chapter proceeds as follows: 3.2 outlines the METR methodology and key assumptions made, before 3.3 presents the METR results.
3.2. Methodology
The METR methodology in this report follows broadly the approach of the OECD’s 1994 Taxation and Household Savings study (OECD, 1994), which itself drew on the methods used by King and Fullerton (1984).1 This section describes this methodology, including the key assumptions that have been made as well as the limitations of the approach. The underlying METR equations derived under this methodology and used to calculate the METRs for each savings vehicle are presented in ANNEX A.
Marginal Investment
As emphasised in the OECD (1994) study, the appropriate way to analyse the effect of tax on savings decisions is to examine the incentives faced by the taxpayer at the margin. The analysis therefore focuses on a saver who is contemplating investing an additional currency unit in one of a range of potential savings vehicles. The investment is a marginal investment, both in terms of being an incremental purchase of the asset, and in terms of generating a net return just sufficient to make the investment worthwhile (as compared to the next best savings opportunity).
The approach assumes a fixed pre-tax real rate of return and calculates the minimum post-tax real rate of return that will for that asset, at the margin, make the investment worthwhile. The METR can then be calculated as the difference between the pre- and post-tax rates of return (the savings income tax wedge) divided by the pre-tax rate of return.
The post-tax rate of return is determined by explicitly modelling the stream of returns and taxes associated with a marginal investment over time. The modelling incorporates the impact of a wide range of taxes, including income taxes (at the personal level as well as at the level of the savings intermediary such as an investment or pension fund), taxes on realised capital gains (with or without indexation), taxes on gross asset purchases and/or sales, transaction taxes and taxes on the stock of an asset (i.e. wealth taxes). The tax gain as a result of the deductibility of savings from taxable income as well as of interest payments that have to be paid if the investment is partly or fully financed with borrowed funds are modelled as well.
Types of savings vehicles modelled
While it is not practicable to cover all possible types of savings vehicles present in all countries, this report does extend considerably beyond the five-asset scope of the original OECD (1994) study. METRs are calculated for the following different savings vehicles, which can be expected to cover the majority of household savings in most countries:
Bank deposits
Corporate bonds
Government bonds
Equities (purchase of corporate shares)
Investment fund assets (marketable collective investment vehicles)
Private pensions
Deposits in individual tax-favoured savings accounts
Equity-financed owner-occupied residential property
Equity-financed rented residential property
Debt-financed owner-occupied residential property
Debt-financed rented residential property
Fixed pre-tax rate of return approach
The adoption of a fixed pre-tax rate of return follows the approach taken in the OECD (1994) study. This approach is conventionally denoted the fixed-p (or fixed-r) approach and focuses on investments that would yield the same pre-tax return at the margin and thus be equally attractive in the absence of tax. As emphasised in the OECD (1994) study, the fixed-p approach effectively assumes that arbitrage between different assets in international capital markets results in all assets generating the same rates of return before taxes. Individual savers then arrange their portfolios in light of the individual-level tax rules.
This approach is consistent with the highly integrated nature of modern international capital markets and the predominantly residence-based taxation of international portfolio savings. The main alternative approach would have been to fix the post-tax rate of return (and then calculate the required pre-tax rate of return to generate this; i.e. a fixed-s approach). However, such an approach would be more appropriate in the case where international portfolio savings were predominantly taxed on a source basis – as arbitrage would then equate post-tax returns across assets. Meanwhile, in closed economies, any tax would change the interest rate in that economy, and so unless the elasticities of saving and investment with respect to the interest rate were known, any assumption about fixing either the pre- or post-tax rates of return would necessarily be only approximate.
The pre-tax real rate of return adopted in calculating the preliminary METRs presented in the next section is three percent. Results were also calculated for a pre-tax real rate of return of two and four percent to examine the sensitivity of the results to the pre-tax real rate of return. For completeness, these additional results are presented in ANNEX B.
Inflation rate
Results are presented for the different savings vehicles first allowing inflation to vary across countries (reflecting the actual inflation rates in each country). To aid cross-country comparison, results are also shown for a common inflation rate (the OECD average). In both cases, inflation rates are based on the five-year average from Q3 2011 to Q2 2016.2 Results were also calculated for a zero rate of inflation to examine the sensitivity of the results to inflation. For completeness, these additional results are also presented in ANNEX B.
Tax rates
METR results are presented for three different taxpayer types to account for potential variation in statutory marginal tax rates depending on an individual’s situation. In the majority of cases, simply specifying an income level is sufficient to determine the appropriate marginal tax rate to apply for each taxpayer type. However, in some cases the appropriate statutory marginal tax rate, or even the applicability of a tax, may also depend on the amount and type of capital income or wealth held by the taxpayer.3 As such, each taxpayer type is defined by both an income level and a wealth level held at the time their investment decision is being made (with wealth further split between housing and non-housing wealth). Capital income levels, when necessary, are then determined by applying a 3% real return on wealth (under the assumption that all non-property income is from wealth held in the particular asset being modelled).4 The three taxpayer types are defined as follows:
Low-rate taxpayer: a single individual earning annual combined (labour plus capital) income equal to 67% of the average wage, with no or minimal net wealth such that their marginal currency unit of savings always benefits from a tax free allowance or tax credit, if one is provided.
Average-rate taxpayer: a single individual earning annual combined (labour plus capital) income equal to 100% of the average wage, with net wealth equal to six times the average wage, of which three-quarters is held in residential property.
High-rate taxpayer: a single individual earning annual combined (labour plus capital) income equal to 500% of the average wage, with net wealth equal to twenty times the average wage, of which half is held in residential property.
Holding period
The length of time that an asset is held can be crucial in determining the taxes paid on the investment. The approach taken here again follows broadly the approach of OECD (1994) in applying a fixed probability of sale to each period. The expected return is then calculated for each asset, on a risk-neutral basis. The expected holding period of the asset is equal to the reciprocal of the fixed probability of sale in each period.
This “endogenous asset holding period” approach has a number of advantages. First, it allows the modelling of different holding periods while avoiding the presence of time-related variables in the analytical solution of the derived effective tax rates. Second, it allows for the impact of deferral of taxes that only have to be paid when the asset is sold. Third, where tax rates vary with the holding period, the method could be potentially extended to implicitly weight the different tax rates according to the different holding periods. The latter extension is not pursued in this report to allow for the possibility of showing METRs for different holding periods.
The expected holding period adopted in calculating the METRs presented in the next section is five years for all savings vehicles with the exception of pension funds and housing where the expected holding period is 20 years. The probability of sale applied to each period is therefore 0.05 for pension funds and housing, and 0.2 for all other assets. Results for additional expected holding periods are presented in ANNEX B.
For capital gains taxes that vary with the holding period, a simplifying assumption has been made in the modelling. Rather than modelling for each year the particular capital gains tax rate that would apply if the asset was sold in that particular year, the calculations use for each year the tax rate that matches the expected holding period. As pointed out, this modelling simplification allows results to be presented for assets held for different expected holding periods.
Type of returns generated
The return to investment in equities may be in the form of dividends, capital gains or a combination of the two. The OECD (1994) study assumed a fixed proportion of returns were derived as dividends and as capital gains. A different approach is taken here, with METRs presented separately for three different scenarios: 100% of the return as dividends; 100% of the return as capital gains; and 50% of the return as each.
Similarly, the return on bonds can be in the form of interest or, when sold below par, as capital gains. A similar approach to that applied to equities has been adopted, with results presented for bonds issued at par (where 100% of the return is in the form of interest) and for bonds issued below par (where 75% of the return is in the form of interest and 25% is capital gains).
Investment funds and pension funds
A simplified approach has been taken regarding the taxation of investment funds. These are modelled under the assumption that all income earned in the fund is retained and reinvested each year until final disposition. Many countries do not impose any tax at the level of the investment fund. However, for countries that impose a fixed rate annual tax at the level of the investment fund, this is modelled with the remaining funds then assumed to be immediately reinvested. If different tax rates are applied at the investment fund level depending on the type of income earned, all income is assumed to be from dividends. In countries where distribution is legally required each year, all income is again assumed to be from dividends, with the remaining funds then being immediately reinvested as a new investment within in the same type of fund.
Income received within a pension fund is also currently assumed to be taxed at a single rate (if it is taxed at all), with the remaining funds then immediately reinvested. On final distribution, the most concessionary tax treatment available is modelled. For example, lower rates (or exemptions) may be provided if a private pension is paid out as an annuity rather than a lump sum, or if payments are not made until a specific retirement age is reached. In such cases, an annuity is assumed to have been chosen and the required retirement age is assumed to have been reached.
For countries that tax distributions from private pension funds at progressive marginal rates, we model two scenarios: first, income in retirement is assumed to be the same as when contributions are made (so marginal tax rates will remain the same); and second, income in retirement is assumed to be lower than when making contributions (potentially resulting in a lower marginal tax rate at retirement). The assumed income reductions are as follows:
A high-rate taxpayer’s income is assumed to fall in retirement from 500% of the average wage to 250% of the average wage.
An average-rate taxpayer’s income is assumed to fall in retirement from 100% of the average wage to 67% of the average wage.
A low-rate taxpayer’s income is assumed to fall in retirement from 67% of the average wage to 50% of the average wage.
Social security contributions
In addition to income taxes, social security contributions (SSCs) may in some cases be payable on income from capital. Following the approach of the OECD (1994) study, the current study focuses on personal taxes on household savings, and hence SSCs are beyond its scope. As such, SSCs are not taken account of in the modelling, either in terms of their imposition on income from capital, or on whether certain payments (e.g. pension contributions, mortgage interest deductions) reduce the base on which contributions are due. There are two exceptions to this general approach: France and the Netherlands5 considered SSCs applicable on capital income to be sufficiently close in nature to pure taxes as to warrant their inclusion in the modelling.
Recurrent property taxes
In most countries, recurrent property taxes are implemented at the sub-central level and rates and bases can vary across the country. Consequently one of two simplified modelling approaches has been adopted on a country-by-country basis. Countries have either applied an estimated average tax rate for the entire country, or a representative municipality/region has been chosen to be adopted in the modelling.
In most countries, the tax base is related closely to the property value. However, the base is often lower than current market value (which is effectively the tax base in the METR calculations – see the methodological annex). Where available, an estimate has consequently been made of the degree to which the actual tax base is undervalued as compared to current market value. In countries where an estimate of undervaluation was not available (Bulgaria, Germany, Iceland, Spain, Turkey) recurrent property tax figures are likely to be overstated. In some countries the tax base depends predominantly on house or property size rather than value, in which case an assumption has been made regarding a “representative” property size, again on a country-by-country basis.
Given the variation in approaches taken, the recurrent property tax modelling should be considered indicative and approximate only.
Limitations of the methodology
While there is great analytical utility in being able to summarise complex tax systems into a single comparable parameter, the METR methodology is not without limitations. In particular, the METRs are scenario and assumption driven, with results sensitive to those scenarios (e.g. the income level chosen) and assumptions (e.g. on inflation, the real rate of return, expected holding period, split between income and capital gains).
For simplicity, the modelling also effectively assumes certainty as to the pre-tax rate of return, the tax system and other parameters that in reality may be subject to change over time. While there is uncertainty over the holding period of an asset, savers are assumed to be risk neutral to this uncertainty – which may not be the case in reality. More fundamentally, the fixed-p approach ignores differences in risk among alternative investments (i.e. higher risk assets are likely to require higher pre-tax rates of return) and that rates of return may vary with household wealth (e.g. because of better quality information of investment opportunities).
The study is also limited to examining the impact of taxes directly imposed at the personal level on savings. It therefore does not account for the possible impact of other taxes, such as taxes directly on labour and consumption, or estate and gift taxes, on savings behaviour at the margin. That said, such effects (if any) are likely to be small.
The omission of the corporate income tax (CIT) from the modelling means that the METRs may not present a full picture of all the taxes imposed on equity investment. The degree to which the METRs capture the total tax burden on equity investment will vary across countries, depending on the mix of taxes imposed at the corporate and personal level and the country’s approach to integration of corporate and personal level taxation. Using the METRs to compare the total tax imposed on equity with tax imposed on other assets, or to compare across countries, should therefore be undertaken with great caution.
Table 3.1 highlights the degree of variation in approaches to the integration of corporate and personal level taxation. A number of countries apply pure classical systems, while many apply a modified classical system where dividends are taxed at a lower rate than other forms of capital income (thereby providing partial relief from double taxation). Many countries apply fixed withholding rates that may be lower than the tax that would have been due if they were subject to marginal PIT rates. Partial tax relief is also applied in various other ways at either corporate level (Allowance for Corporate Equity) or individual level (partial inclusion, dividend exemption, or via the provision of a rate of return allowance). Partial imputation systems may give full relief from corporate level tax to some taxpayers but not to all, while full imputation gives full relief from corporate level tax at the personal level.
Table 3.1. Approaches to integration of corporate and personal level taxation
Note: For more detail on different approaches to the integration of corporate and personal level taxes, see Harding and Marten (2018).
Source: Harding and Marten (2018), OECD Tax Database, IBFD Tax Research Platform.
3.3. Results
This section presents detailed METR results for all countries and all asset types. Table 3.2 Table 3.4 present results using the actual country inflation rate for the three different taxpayer types (low-rate, average-rate and high-rate taxpayers). These results are used for within-country comparison of METRs across asset types, but not for cross-country comparison. Table 3.5 Table 3.7 instead present results using a common (OECD average) inflation rate to aid cross-country comparison of METRs for a particular asset type (by removing variations in inflation as a cause of METR differences). Overall results from Table 3.2 Table 3.4 are first summarised below in box-and-whisker plot form, while results are also presented separately for each country in graphical form at the end of the section.
Table 3.2. Marginal effective tax rates by asset, 2016 – Personal tax rate: 67% of average wage case; Inflation: country actual
Table 3.2. Marginal effective tax rates by asset, 2016 – Personal tax rate: 67% of average wage case; Inflation: country actual (cont.)
Table 3.3. Marginal effective tax rates by asset, 2016 – Personal tax rate: 100% of average wage case; Inflation: country actual
Table 3.3. Marginal effective tax rates by asset, 2016 – Personal tax rate: 100% of average wage case; Inflation: country actual (cont.)
Table 3.4. Marginal effective tax rates by asset, 2016 – Personal tax rate: 500% of average wage case; Inflation: country actual
Table 3.4. Marginal effective tax rates by asset, 2016 – Personal tax rate: 500% of average wage case; Inflation: country actual (cont.)
Table 3.5. Marginal effective tax rates by asset, 2016 – Personal tax rate: 67% of average wage case; Inflation: OECD average
Table 3.5. Marginal effective tax rates by asset, 2016 – Personal tax rate: 67% of average wage case; Inflation: OECD average (cont.)
Table 3.6. Marginal effective tax rates by asset, 2016 – Personal tax rate: 100% of average wage case; Inflation: OECD average
Table 3.6. Marginal effective tax rates by asset, 2016 – Personal tax rate: 100% of average wage case; Inflation: OECD average (cont.)
Table 3.7. Marginal effective tax rates by asset, 2016 – Personal tax rate: 500% of average wage case; Inflation: OECD average
Table 3.7. Marginal effective tax rates by asset, 2016 – Personal tax rate: 500% of average wage case; Inflation: OECD average (cont.)
Overall, the results show significant variation in METRs across asset types, with tax systems creating significant incentives to alter portfolio allocation away from that which would be optimal in a no-tax world. This variation can be seen in Figure 3.1, which summarises the distribution of METR results across countries in box-and-whisker plots for each asset type. The bold horizontal line within each “box” represents the median METR for that asset type, while the box itself reflects the inter-quartile range (the distance between the first and third quartile). The extreme points of each “whisker” show the minimum and maximum METR values for each asset type (excluding outliers6 ).
Figure 3.1. Distribution of marginal effective tax rates across countries for each asset type, 2016
Explanatory note: The bold horizontal line within each “box” represents the median METR for that asset type, while the box itself reflects the inter-quartile range. The extreme points of each “whisker” show the minimum and maximum METR values for each asset type, excluding outliers. An outlier is defined as a result more than 1.5 times the inter-quartile range below the first quartile or above the third quartile.
Figure 3.1 highlights that some asset types tend to be particularly tax-favoured as compared to others. Pension funds tend to be the most tax-favoured, with owner-occupied residential property, and tax-favoured savings accounts (when present) also tending to be significantly tax-favoured. The tax advantage towards pension funds tends to increase with income, though there is considerable variation across countries. In contrast to owner-occupied residential property, rental property is often subject to relatively high METRs due to the application of progressive marginal personal income tax (PIT) rates, capital gains taxes and significant property taxes. Bank accounts and corporate bonds tend to be relatively heavily taxed in many countries.
METRs are often significantly higher than statutory tax rates because they combine the impact of multiple taxes – as well as the impact of inflation – in a single indicator. (Box 3.1 discusses further the differences between METRs and statutory tax rates.)
Box 3.1. Comparing effective and statutory marginal tax rates
The marginal effective tax rates (METRs) presented in Table 3.2 Table 3.7 are often significantly higher than marginal statutory tax rates applicable to income from saving in a particular asset. As presented in detail in the methodology section and ANNEX A, this is because METRs combine various factors – including the impact of multiple taxes, deductions and variations in tax bases, and the impact of inflation – in a single tax burden indicator. As such, METRs provide a far more complete picture of the tax-induced incentives individuals face to save in different assets.
Factors that increase METRs include:
The presence of multiple taxes on saving in a particular asset, such as income taxes, capital gains taxes, transaction taxes, and wealth or property taxes (if applicable in a particular country).
Where savings occurs through an intermediary, the imposition of taxes on the same income at both the intermediate and personal level.
Inflation, where taxes are imposed on nominal rather than real returns.
Factors that reduce METRs include:
Tax deferral effects – where statutory tax rates are imposed only once income is realised as opposed to when it accrues (e.g. capital gains); or where lower statutory tax rates apply beyond a specified holding period.
Reductions in the tax base – such as basic allowances or exemptions for an amount of savings income, tax deductions or tax credits (e.g. for contributions to private pension schemes or for mortgage interest payments).
The impact of these different factors may also vary with the underlying assumptions of the specific METR being calculated – e.g. the taxpayer type (and corresponding income and wealth levels), how long the asset is owned, and the type of return generated (e.g. interest, dividends or capital gains).
The METR calculations presented in this chapter show that, in general, factors increasing METRs above statutory tax rates tend to outweigh factors decreasing them below statutory rates, particularly for higher income taxpayers. The most significant exception to this is tax relief for private pension contributions which often more than outweighs the tax paid when the pension is received in retirement, resulting in negative METRs.
The rest of this section examines in detail the results for each asset type, in turn, before discussing the sensitivity of the results to the underlying modelling assumptions.
Results for different savings vehicles
Bank accounts
Income earned from savings in bank accounts often faces relatively high METRs as compared to other asset types. Nominal interest is taxed, either at marginal PIT rates (e.g. Australia, the United Kingdom and the United States), or at flat withholding rates (e.g. Germany, Italy and Korea). Countries with progressive PIT rate systems unsurprisingly tend to impose higher METRs on higher tax rate taxpayers than withholding rate countries or countries with flat-rate PIT systems (e.g. the Czech Republic and Hungary).7
The lowest METRs for bank account savings across all taxpayer types are observed in Argentina and Estonia, where interest income is exempt from any taxation, and in Colombia which imposes a flat 4% tax rate on all interest income.8 Uniquely amongst the 40 countries modelled, Colombia taxes real rather than nominal interest income.
A number of other countries impose zero METRs on low-rate and (in some cases) average-rate taxpayers, but positive METRs on high-rate taxpayers (Chile, Germany, Iceland, Lithuania, the Netherlands, South Africa and the United Kingdom). This is generally due to the exemption of a fixed amount of interest income from taxation.9 For example, in Iceland, the first ISK 125 000 (EUR 1 000) of interest income is tax exempt. In Lithuania, the exemption relates to interest accruing from a fixed annual deposit amount (EUR 500 in 2016). In South Africa, while a zero METR applies to low-rate and average-rate taxpayers (as they are assumed to earn less interest income than the ZAR 28 300 threshold), the combination of a high withholding tax rate and significant inflation results in a very high METR for high-rate taxpayers. In the Netherlands, a deemed return on asset value is taxed, rather than the actual income. The first EUR 24 437 (in 2016) of total assets excluding pensions and owner-occupied housing is exempt. For assets above this amount, a flat 30% tax is imposed on a deemed 4% return.10
Corporate bonds
The return on corporate (or government) bonds can be in the form of interest or, when issued below par, as capital gains. As such, METR results are presented for two scenarios: bonds issued at par (where 100% of the return is in the form of interest); and bonds issued below par (where 75% of the return is in the form of interest and 25% is capital gains).
Thirty-two of the forty countries tax interest from bank accounts identically to interest from corporate bonds, so that METRs are equivalent when bonds are issued at par. This is illustrated in Figure 3.2 which presents METRs for the average rate taxpayer case for bank accounts and corporate bonds (as detailed in Table 3.3).
Figure 3.2. Marginal effective tax rates for bank accounts and corporate bonds, 2016
Personal tax rate: 100% of average wage case. Inflation: country actual
Argentina, Austria, Belgium, Colombia, Estonia, and Switzerland impose higher METRs on interest income from corporate bonds than from bank accounts. Austria, for example, applies a higher withholding tax rate to interest from corporate bonds than to bank interest. In contrast, Bulgaria taxes interest on bank accounts, but exempts interest on corporate bonds as long as they are sold on a regulated exchange. Ireland, meanwhile, imposes a fixed withholding tax on bank interest, but imposes progressive marginal PIT rates on interest from corporate bonds. This results in a higher statutory tax rate (and METR) on bank interest than on corporate bonds for low-rate taxpayers, but the reverse is the case for average and high-rate taxpayers.
In almost all countries, corporate bonds issued below par face a lower tax rate than those issued at par. This, again, can be clearly seen in Figure 3.2 for the average rate taxpayer. At a minimum this is due to the deferral of tax payments until realisation of the capital gain. In addition, some countries tax capital gains at either a lower statutory tax rate than interest income (e.g. Hungary), or tax only part of the gain (e.g. Canada), which further reduces METRs. Meanwhile, some countries (e.g. the Czech Republic – if the bonds are held for three years or more) exempt capital gains entirely. In contrast, Latvia imposes a higher tax rate on capital gains than on interest from bonds, leading to a higher METR on bonds issued below par.
In Ireland, while interest income from corporate bonds is taxed at progressive marginal PIT rates, capital gains are taxed at a flat rate. As a result, bonds issued at a discount have a higher METR for low-rate taxpayers, but a lower METR for average and high-rate taxpayers. In Finland, while capital gains for high-rate taxpayers are taxed at a slightly higher statutory rate than interest is, this effect is outweighed by the tax deferral effect so that METRs remain slightly lower for bonds issued at a discount.
While not captured in the modelling results presented in this report, New Zealand’s “financial arrangement” rules treat capital gains as interest income and tax this on an accrual basis. This leads to identical METRs for bonds issued at par and at a discount – as there is no deferral of tax liability. However, these rules only apply to high-income taxpayers meeting certain requirements.11 Nevertheless, the financial arrangement rules do ensure that capital gains are treated as taxable income for all taxpayers, whereas they would otherwise be untaxed (as New Zealand does not tax capital gains on a comprehensive basis).
Government bonds
Only four countries (Chile, Greece, Ireland and Italy) tax government bonds differently to corporate bonds. In each case government bonds receive concessionary treatment. Chile and Ireland tax interest from corporate and government bonds identically, but exempt capital gains from government bonds. As Chile exempts a minimum amount of interest and capital gains income, this difference is only seen in the results for the high-rate taxpayer. Greece exempts government bonds from taxation on both interest and capital gains, while Italy imposes a lower withholding tax on interest and capital gains from government bonds.
Equities (purchase of corporate shares)
The return on an equity investment may take the form of either dividend income or capital gains. To take account of this, METR results for equity investments are presented for three different scenarios: 100%, 50% and 0% distribution of income as dividends, with the non-distributed income being taxed as a capital gain on realisation.
As noted earlier, the results do not take account of taxation at the corporate level, and therefore will not fully reflect the total tax imposed on an equity investment. Using the METRs to compare the total tax imposed on equity with tax imposed on other assets, or to compare across countries, should be undertaken with great caution. In particular, a country’s approach to integrating corporate and personal level taxation needs to be considered when interpreting the results. For example, Colombia imposes zero METRs irrespective of the distribution policy for all three taxpayer types, while Germany and the United Kingdom impose zero METRs for low-rate and average-rate taxpayers. Nevertheless, tax is still paid at the corporate level in these countries.12
Overall, METRs for equity investments are generally lower than those on bank accounts and bonds. Furthermore, rates typically fall as capital gains make up a greater part of the overall return. This is illustrated in Figure 3.3, which presents METRs for the average rate taxpayer case for the three distribution scenarios (as detailed in Table 3.3).
Figure 3.3. Marginal effective tax rates for shares, 2016
Personal tax rate: 100% of average wage case. Inflation: country actual
The fall in METRs highlights the fact that capital gains are typically taxed more favourably than dividend income. At the very least, as with bonds issued below par, this is due to the deferral of taxation of capital gains until realisation. More often, though, capital gains on equity investments are also taxed at lower rates than dividend income. These reduced rates typically apply only for “long term” capital gains held for longer than a specified minimum period. This period is typically between one and five years meaning that the reduced rates apply for all countries presented (as the results presented are based on a five-year expected holding period). In several countries (Bulgaria, the Czech Republic, Luxembourg, New Zealand and Turkey), capital gains are exempt, resulting in METRs of zero when capital gains make up the entire return. This is also the case for low and average rate taxpayers in Chile and Lithuania.
In several other countries (Estonia, Finland, Greece, Japan, Latvia, the Slovak Republic) METRs actually increase when capital gains make up a greater part of the overall return – as capital gains are taxed at higher rates than dividends (generally due to reductions in dividend taxation to reduce the extent of double taxation on corporate income). At the extreme, the METRs for Estonia and the Slovak Republic are zero when dividends make up the entire return, due to the non-taxation of dividend income at the personal level.
In Australia, only 50% of capital gains are taxable. However, this effect is outweighed by the impact of imputation (“franking”) tax credits on dividend income, so that METRs increase slightly when capital gains make up a greater part of the overall return (as shown in Figure 3.3). Canada also applies a 50% capital gains tax reduction and an imputation system.13 For low- and average-rate taxpayers, imputation credits also outweigh the impact of the capital gains tax reduction. However, the opposite occurs for high-rate taxpayers, so that METRs fall when capital gains make up a greater part of the overall return.
In Chile, Mexico and New Zealand, the impact of imputation credits outweighs the personal income tax due on dividend income for some taxpayer types. This results in negative METRs when income is fully distributed as dividends in all three countries for low-rate taxpayers, in Chile and Mexico for average-rate taxpayers, and also in Chile for high-rate taxpayers. METRs then increase as capital gains make up a greater part of the overall return. However, for average- and high-rate taxpayers in New Zealand, METRs fall when capital gains make up a greater part of the overall return because capital gains are untaxed. It should again be borne in mind that these results do not take account of taxation at the corporate level, and the combined corporate plus personal level tax burden would be positive in each country.
In Japan, the combination of taxation of dividend income at progressive PIT rates and flat-rate taxation of capital gains results in METRs decreasing as capital gains make up a greater part of the overall return for high-rate taxpayers, but increasing for low-rate and average-rate taxpayers (as shown in Figure 3.3). Similarly mixed results occur in Ireland.
METRs are constant irrespective of distribution policy in Argentina, the Netherlands and Norway. In Argentina, this is because only the wealth tax is imposed on shares. In Norway, no tax is applied directly on dividends or capital gains in the modelling due to Norway’s risk-free return allowance and the assumption that, on the marginal euro of savings, the return generated is simply the risk-free return. (A non-marginal unit of savings with a return above the risk-free rate would be taxed). As such, only Norway’s wealth tax is applied. In the Netherlands, shares are subject to the deemed return system, and hence the nature of the income is irrelevant.
Investment funds
Rather than investing directly in assets such as bonds and equities, a taxpayer may put their money into an investment fund that pools all the individual investors’ money together and invests it “collectively” in a portfolio of assets, including bonds and equities. Countries may tax the income from such investment funds at two levels – each year as it is earned within the fund, and on distribution from the fund. In a small number of countries (Argentina, Colombia, France, Norway, Spain and Switzerland), a wealth tax is also applied to wealth held in investment funds above a threshold amount.14
Tax treatment of investment funds varies widely across the 40 countries considered in this report, with METRs consequently also varying significantly. Some countries tax investment funds annually in the hands of the individual investor. This is either achieved by treating the fund as a pass-through entity (as in Australia, Canada, the United Kingdom and the United States); by deeming a distribution to have occurred each year if income has not been actually distributed (as in Germany); or by requiring distribution each year (as in Korea). In most of these cases investors are taxed at marginal PIT rates, with METRs consequently varying depending on the applicable marginal PIT rate.
A larger number of countries only tax investment fund income on final distribution to the individual investor. These distributions (assumed in the modelling to be capital gains on the sale of shares/units in the investment fund) are then taxed at flat rates in the majority of cases so that METRs tend to be constant across the three taxpayer types. In a small number of cases, the distributed capital gains are untaxed which, together with the lack of taxation at the fund level, leads to zero METRs. This is the case in Belgium, Bulgaria and Colombia, and in Luxembourg as long as the shares in the investment fund are held for more than three years. In Lithuania, a tax-free threshold means that capital gains are only taxed in the high-rate taxpayer case. In the Slovak Republic, capital gains on investment funds are exempt if they are held for 15 years or more. However, this is not captured in Table 3.2 Table 3.7 which assume a five-year holding period.
France and Norway do not tax returns at the fund level either. However, in addition to taxing distributions to the investor, they impose a wealth tax on the investment fund wealth held by top-rate taxpayers. Colombia also imposes a wealth tax – but with a minimum wealth threshold that is so high as to not affect the three taxpayer types presented in Table 3.2 Table 3.7.
Meanwhile, some countries tax investment funds at both the fund level and separately on distribution. While in some cases this can result in particularly high METRs, in other cases taxation on distribution may be limited. For example, base investment funds in the Czech Republic are taxed at 5% at the fund level, and are taxed again on distribution, but if the return generated is a capital gain – as is modelled – then this will be tax free.
Argentina is the only country not to impose tax at either the fund level or on any type of distribution (i.e. not just capital gains). Argentina’s wealth tax does apply to investment funds, although its minimum wealth threshold means that only the high-rate taxpayer type is subject to it.
Private pensions
Private pension schemes are highly tax-favoured in the majority of countries considered in this report. Most countries allow a deduction or tax credit for contributions made to an approved private pension scheme, though all countries have some limitation on the amount of eligible contributions. The majority of countries also do not tax the income accrued within a pension fund, and many countries provide concessionary treatment on distribution. The combined effect of these provisions is to produce zero or negative METRs for all three taxpayer types in 33 out of 40 countries, as shown in the “deductible contributions” pension fund columns in Table 3.2 Table 3.7.
A zero METR will occur when a country simply does not tax any income from a private pension. This is the case in Argentina and Turkey, where no tax relief is provided for pension contributions, but no tax is imposed on accumulation or on distribution (a “TEE” system). Zero rates will also occur when the tax rate at which deductions on contributions are provided matches the tax rate paid on distribution and there is no taxation on accumulation within the fund. The upfront tax saving in this “EET” system effectively funds the tax due on final distribution. However, where final taxation is at a rate lower than the initial deductions, the upfront tax saving more than compensates the taxpayer for their future tax liability – and hence a subsidy arises.15
The modelling results presented in Table 3.2 Table 3.7 assume that the taxpayer earns the same level of income when they make their contributions as when their pension is distributed, and hence face the same marginal PIT rate when making contributions and receiving distributions (when distributions are taxed at marginal PIT rates). Under this assumption, the results show that ten countries impose zero METRs for all three taxpayer types (Argentina, Canada, Chile, the Czech Republic, Finland, Germany16 , Greece, Iceland, Spain and the United States).17
In 22 countries, concessionary tax rates are applied on distributions (an “EEt” system, or “EEE” in the case of Austria, Bulgaria, Hungary and Lithuania), and hence METRs are negative for all three taxpayer types.18 Even if accumulations within the fund are taxed, a lower tax rate on distribution can still result in a negative METR. For example, Italy taxes accumulations within a pension fund, but the concessionary rate on distribution is sufficient to outweigh this effect and still produce a negative METR (this can be categorised as an “Ett” system). In contrast, Denmark also taxes accumulations but provides no concessionary treatment on distribution, so overall METRs are positive despite the deductibility of contributions.
Examining the “deductible contributions” results across the three different taxpayer types, shows that in some countries the tax subsidy provided by pension schemes increases with income. This is the case in nine countries (Colombia, France, Italy, Japan, Luxembourg, Poland, Slovenia, Switzerland and the United Kingdom). The reason is that contributions are deducted at the taxpayer’s marginal tax rate – so higher tax rate taxpayers get a larger upfront tax reduction, but face the same tax rate on distribution as lower-rate taxpayers. In Mexico, the METR is lowest for the average-rate taxpayer rather than the high-rate taxpayer as the high-rate taxpayer is subject to taxation on distribution whereas lower rate taxpayers are not.
To prevent such regressive effects, a number of countries (e.g. Austria, Belgium, Hungary, Korea and Israel) provide a tax credit rather than deduction for pension contributions. The Korean system goes even further, providing a larger tax credit (15%) for lower income taxpayers than higher income taxpayers (12%); METRs are consequently lower for the low-rate taxpayer case than the average-rate and high-rate taxpayer cases. A similar effect is achieved in Norway by allowing all taxpayers to claim a deduction for a surcharge, however, the surcharge is only applied to distributions to high-rate taxpayers. Meanwhile, Bulgaria, Estonia, Hungary, Latvia and Lithuania allow deductions at marginal PIT rates, but their flat rate PIT systems mean that all taxpayers benefit proportionately.
Several countries do not allow deductions for contributions, and consequently apply positive, and often progressive, METRs. For example, New Zealand applies a TTE system – where contributions are not deductible, income accumulating in the pension fund is taxed (at progressive rates), but distributions are not taxed. Meanwhile, the Slovak Republic and Sweden exempt accumulations, but tax distributions at flat and progressive rates, respectively (a “TET” system).
To take account of the varying limits in place on deductibility in almost all countries, results are also presented in Table 3.2 Table 3.7 for a marginal investment in a pension fund where the contribution is not deductible (in the “non-deductible contributions” column). Such an approach obviously limits the tax cost of EEt systems or EET systems where taxpayers can be subject to significantly lower marginal tax rates on distribution. The results highlight the importance of deductibility in providing concessional treatment. All METRs are now either zero or positive, with many rates increasing considerably. The main exceptions are the countries that apply TTE and TET systems, and therefore remain unchanged (New Zealand, the Slovak Republic, Sweden, Turkey), and the Netherlands – which simply swaps an EET system for a TEE system on contributions in excess of a contribution threshold (EUR 100 000), so the METR remains at zero. Greece is the only country that has no limitation on deductibility.
The results presented so far highlight the significantly concessionary nature of EET systems. However, they may still understate the extent of the concession in countries that apply marginal PIT rates as it is very likely that distributions would be taxed at lower rates than deductions were granted at. This is because a taxpayer receiving pension payments is likely to be earning less income than when they were making contributions, if for no other reason than that they are likely to now be retired from the labour force.
Table 3.8 Table 3.10 present results where the taxpayer is assumed to earn less in retirement than when making contributions, and therefore potentially faces lower marginal PIT rates on distributions (although this depends on the interaction with the particular rate schedule). METRs fall in 4, 14 and 11 countries for low-rate, average-rate and high-rate taxpayers, respectively.
Table 3.8. Marginal effective tax rates on private pensions, 2016 – Personal tax rate: 67% of average wage case; Inflation: country actual
Table 3.9. Marginal effective tax rates on private pensions, 2016 – Personal tax rate: 100% of average wage case; Inflation: country actual
Table 3.10. Marginal effective tax rates on private pensions, 2016 – Personal tax rate: 500% of average wage case; Inflation: country actual
The reductions in METRs are also highlighted for the average-rate taxpayer case (for deductible contributions) in Figure 3.4. Six of the ten countries with EET systems previously producing zero METRs now have negative METRs. For example, the marginal tax rate on pension distributions for an average-rate taxpayer in the United States falls from 25% to 15%, resulting in a drop in the METR from zero to -22.2%.
Figure 3.4. Marginal effective tax rates for private pensions (deductible contributions), 2016
Personal tax rate: 100% of average wage case. Inflation: country actual
While countries with fixed withholding rates are unaffected, rates for some countries that were already providing concessionary treatment of distributions also fall. For example, the already negative METRs in Japan, France, Luxembourg, the Netherlands, Slovenia and South Africa fall further (for at least the average-rate taxpayer type) as the concessionary taxation of distributions in these countries is linked to the marginal tax rate (e.g. Luxembourg taxes distributions at a rate equal to 50% of the marginal PIT rate).
Tax-favoured savings accounts
Tax-favoured savings accounts are present in several countries and are subject to varying degrees of restriction. For example, in the United Kingdom, an individual can save up to GBP 20 000 (EUR 22 500) in an Individual Savings Account (ISA) tax free. Similarly Korea provides an “Individual Savings Account” which allows up to KOR 20 million (EUR 15 800) to be saved tax-free. Similarly, South Africa’s “tax free savings accounts” provide an income exemption, with contributions limited to ZAR 30 000 (EUR 1 900) per year. In Belgium the first EUR 1 880 of interest income in qualified savings accounts is exempt, and any additional income is taxed at a concessionary (15%) withholding rate. In the United States, the tax-free “529 plan” and “Coverdell education savings account” are available, but are restricted to saving for particular education expenses.
In Spain, income from specified sources earned in “long-term savings plans” is exempt as long as the taxpayer does not withdraw any of the capital for at least five years. Contributions cannot exceed EUR 5 000 per year. In Hungary, interest, dividend and capital gain income earned in a “long-term investment account” is taxed at concessionary rates if the underlying assets are held for at least three years, and is exempt if held for at least five years. In Luxembourg, interest attributed to a building society account is exempt from taxation if the assets of the account are used to finance the construction, the purchasing, or the transformation of an apartment or a dwelling for the taxpayer’s own needs. The contributions are deductible up to an amount of EUR 672 per year (the modelling assumes the taxpayer is contributing more than this amount per year).
In Colombia, income in “AFC” accounts is exempt if used to either cover mortgage payments or the acquisition of real estate. Contributions made are also deductible as long as they remain in the account for at least five years (10 years as of 2017). For the medium (5 year) investment presented in the main results, this concessionary tax treatment leads to zero METRs for the low and average rate taxpayer types (because they pay zero income tax, so get no benefit from the deductibility of contributions). However, a high rate taxpayer faces a METR of -259% because they benefit from the immediate deduction, but never pay any tax on the income during accumulation or distribution.19
Several countries also have schemes that are not covered by the modelling results. For example, Austria provides a tax credit (maximum EUR 1 200) for contributions to housing savings accounts, but these most be held for at least 6 years (and the results presented assume a five-year expected holding period). In Denmark, children’s savings accounts are tax favoured, but they must be established before the child turns 14 years of age.
Equity-financed residential property
Using equity to save in owner-occupied residential property is highly tax-favoured in a majority of the 40 countries covered in this report. This is predominantly due to the non-taxation of imputed rental income and exemption of owner-occupied properties from capital gains taxation in most countries.
The significant impact of the non-taxation of imputed rental income can be seen when comparing owner-occupied with rented residential property. When a property is rented out, this rental income is taxed, typically at progressive marginal PIT rates. Additionally, most countries tax the capital gains from rented residential property, albeit some at reduced rates. These differences result in significantly higher METRs for rental properties than for primary residences, and are often higher than those that apply to other assets.
A key assumption necessary in modelling residential property is the split of the real return between rental income and capital gains. The base assumption made in the analysis is that 50% of the real return to residential property comes in the form of (imputed or actual) rental income and the remaining 50% as a capital gain. Where capital gains make up a significantly greater proportion of the overall return, METRs can be significantly lower than those presented in this section – due to the often concessionary taxation of capital gains. This is investigated further in the sensitivity analysis section.
Given the typically progressive taxation of rental income, METRs are generally higher for higher rate taxpayers than lower rate taxpayers for rental property, whereas METRs tend to be constant for owner-occupied properties. That said, rates are constant for rental property, for example, in the Czech Republic and Hungary due to their flat rate PIT systems. In some cases, progressivity in property taxes – whether based on income or value of the property – can lead to higher METRs on owner-occupied property for higher rate or higher wealth taxpayers, such as in Korea.20 Meanwhile, in the United States, METRs on owner-occupied and rented property are actually lower for higher tax rate groups. This is due to the deductibility of sub-central property taxes against federal taxes which provides a greater benefit to taxpayers on higher marginal tax rates.
Transaction taxes (e.g. stamp duties) and recurrent property taxes typically apply to both owner-occupied and rented residential properties, and in some cases can be very substantial. For example, Belgium imposes a transaction tax of 12.5% on the purchase of all residential property, which even when spread over the entire expected holding period, leads – together with a significant recurrent property tax – to a substantial METR on owner-occupied residential property.21 Canada (Ontario), the United Kingdom and United States also impose large recurrent property taxes relative to most other countries, resulting in comparatively high METRs on owner-occupied residential property. In contrast, for example, Italy exempts primary residences from recurrent property taxes and applies a reduced transaction tax rate to owner-occupied properties as compared to rental properties.
The inclusion in the METR calculations of recurrent property taxes effectively assumes that these taxes are a tax on the asset stock. However, it is arguable that in many cases these taxes may act like a “benefits tax” (i.e. as a payment for services) – and thus should not be included in the METRs. It is also arguable that recurrent property taxes may in some cases be capitalised into prices and hence are not borne by the prospective investor. We make no definitive conclusion on these issues in the report. Instead we present, for comparison, results both with and without recurrent property taxes in Table 3.11 Table 3.13. Results for the average-rate taxpayer case are also summarised in Figure 3.5 Figure 3.6 for (equity-financed) owner-occupied and rented residential property, respectively.
Table 3.11. Marginal effective tax rates on residential property, 2016 – Personal tax rate: 67% of average wage case; Inflation: country actual
Table 3.12. Marginal effective tax rates on residential property, 2016 – Personal tax rate: 100% of average wage case; Inflation: country actual
Table 3.13. Marginal effective tax rates on residential property, 2016 – Personal tax rate: 500% of average wage case; Inflation: country actual
Figure 3.5. Marginal effective tax rates for equity-financed owner-occupied residential property, 2016
Personal tax rate: 100% of average wage case. Inflation: country actual
Figure 3.6. Marginal effective tax rates for equity-financed rented residential property, 2016
Personal tax rate: 100% of average wage case. Inflation: country actual
Results, in general, show a significant reduction in METRs in the absence of recurrent property taxes for both owner-occupied and rented residential property. Reductions are particularly large in Canada, Chile, South Africa, the United Kingdom and the United States. The reduction in Belgium is significant, but not as large as in these other countries as Belgium’s comparatively large transaction tax remains.
Results without recurrent property taxes are unchanged in Hungary, Sweden and Switzerland. In Hungary, a representative tax rate of zero is modelled as only 548 out of 3 178 municipalities levy a recurrent property tax (in 2016). Sweden applies a relatively low maximum recurrent property tax amount and hence all three taxpayer types are assumed to pay no additional property tax at the margin. In Switzerland, no recurrent property tax is applied in Zurich, the representative canton applied in the modelling (although a number of other cantons do apply a recurrent property tax). In Italy, no recurrent property tax is applied to owner-occupied residential property (except for luxury homes). Meanwhile, a recurrent property tax is only modelled in Estonia for high-rate taxpayers due to an exemption for small properties (of less than 1 500 square metres).
Debt-financed residential property
It is common to at least partially supplement equity-financing of residential property with debt, and this may have significant tax implications. As such, METRs are also calculated for a marginal debt-financed investment in either owner-occupied or rented residential property. The modelling approach taken is to consider a saver choosing whether to finance a marginal investment with either debt or equity. If they choose to finance with debt, they must pay the market interest rate on the debt (the final cost of which may be reduced due to tax deductibility of the interest payments). They can then also invest the equity that they had available in an alternative investment and derive a return.22 This alternative investment is assumed to be bank account interest earning the same pre-tax return as must be paid in mortgage interest (and as is assumed to be earned on all savings vehicles), and subject to the METRs as calculated in this report.
Under this scenario, if mortgage interest is deductible for tax purposes at the same rate as tax is paid on the taxpayer’s alternative investment, then the financing cost will be zero – as the after-tax income earned on the alternative investment will exactly offset the after tax interest cost of the loan. This is the case, for example, in the Czech Republic, Sweden and the United States, and hence METRs are identical for debt-financed and equity-financed residential property. It is also the case for rental property in the Netherlands for taxpayers saving above the EUR 24 437 exemption amount, as residential rental property and bank accounts are taxed identically under the deemed return system.
If, however, mortgage interest is deductible at a lower rate than the tax rate on the alternative investment, the financing cost will be positive, and METRs will be higher for debt-financed housing. This will happen if the tax rate on rental income and hence the rate of deductibility of mortgage interest is lower than that on bank accounts.
In many countries (e.g. Austria, Germany and the United Kingdom) mortgage interest deductions are only allowed for rental property where the corresponding income is taxed, not owner-occupied property where imputed rental is not taxed. The financing cost will consequently be positive and METRs will increase for owner-occupied property. This will reduce the difference in METRs between owner-occupied and rental properties for debt-financed as compared to equity-financed investment. Mortgage interest deductions may also be disallowed for high income taxpayers or on more expensive houses, as is the case in Italy and Korea, thereby increasing METRs on debt-financed as compared to equity-financed housing.
Meanwhile, if mortgage interest is deductible at a higher rate than the tax rate on the alternative investment, the financing cost will be negative, and METRs will be lower for debt-financed housing. Effectively, the after-tax cost of borrowing is lower than the after-tax return made on the alternative investment. This will happen if the alternative investment is tax-favoured such as is modelled for low- and average-rate taxpayers in the United Kingdom (for rental property). This is also often the case in countries that apply a withholding tax on bank interest but marginal rates on rental income, such as in Germany and Italy, where marginal PIT rates on rental income are always higher than the withholding rate on bank interest.
These results, of course, are based on the assumption that the pre-tax rate of return on the bank account and the interest rate paid on the mortgage are the same. In reality, a bank will charge a higher rate of interest on a mortgage than they will pay on a bank deposit. However, for simplicity, and to illustrate the impact of the tax system rather than broader factors, we assume that both rates are the same. Even with a lower pre-tax rate of return being paid on a bank account, the tax saving on an alternative investment may still be sufficiently large as to outweigh the effect of the lower return, and so still reduce the financing cost.
In some countries (Argentina, Estonia, Mexico and the Netherlands) negative METRs can occur for debt-financed owner-occupied property. This is because there is no taxation of imputed rental income, but a mortgage interest deduction is still provided. The deduction is then assumed to be utilised against labour income in that year.
The ability to debt-finance a property may also open up leverage-based tax planning opportunities that may increase demand for residential rental property, without necessarily encouraging increased savings in residential rental property. Box 3.2 discusses the type of tax planning opportunities that may in some circumstances arise.
Box 3.2. Debt financing and tax planning opportunities
To examine the impact of tax on the composition of savings, the basic structure of the analysis throughout this chapter has been to focus on a taxpayer with one marginal currency unit of savings. This remains the case when considering the debt-financed purchase of residential property – with the opportunity return on that one currency unit of savings then partially, fully, or more than fully compensating for the interest costs of the debt, depending on tax rules in place. However, the ability to debt-finance residential property may also create leverage-based tax planning opportunities.
However, tax systems may provide tax-induced incentives to invest in immovable property even in the absence of own funds (i.e. savings) that can be invested. Consider, for example, a scenario where a rental property is expected to generate a significant return to pay for all ownership and financing costs, including taxes. The return may come predominantly in the form of a tax-preferred capital gain and through (possibly) tax-privileged rental income. Mortgage interest deductions are allowable against rental income, and operating losses can be offset against other sources of income. The taxpayer fully debt-finances the purchase of a residential rental property (using their owner-occupied house as collateral). The principal investment is gradually paid back over time or is redeemed when the loan expires. Significant interest expenses are incurred in the first few years following purchase, with the consequent interest deductions more than fully offsetting the rental income earned. The operating losses are possibly offset against labour income that would otherwise be taxed at high marginal PIT rates.
In subsequent years interest payments fall and the operating loss approaches zero. However, at this point the taxpayer then fully debt-finances the purchase of another rental property. This process could continue as long as banks want to provide debt financing. On sale, the anticipated capital gain is made, and taxed at a concessionary rate, with the after-tax gain more than fully offsetting the operating losses incurred. A leverage-based profit is consequently made without any of the taxpayer’s actual savings being used.
Depending on the tax system in a particular country, variants to the above scenario could include the carry-forward of losses to be utilised against capital gains once realised, additional losses being incurred up front due to depreciation deductions (potentially then “clawed back” on realisation of the capital gain, but with consequent timing benefits), and timed realisations to access age-based concessionary tax rates.
Such tax planning opportunities effectively require the after-tax expected return to be significant enough to (largely or fully) outweigh the mortgage interest costs and any other costs incurred. The tax rules illustrated above – deductibility of mortgage interest and concessionary treatment of capital gains – make such a possibility more likely and imply that investing in immovable property, even in the absence of equity, can be attractive. Such a scenario might in particular be attractive when housing markets are booming and investors are expecting a significant capital gain to make a positive return on investment. As noted in Chapter 2, mortgage interest is indeed deductible in most countries on rental properties, while capital gains are also often taxed at concessionary rates.
These tax planning opportunities highlight how particular tax rules may lead to greater incentives to purchase residential rental property than the METRs presented in this chapter suggest is likely to occur. Analysing these types of investment within a METR framework is feasible, but would require major adjustments to the modelling underlying the calculations presented in this report, as the timing of repayment of the principal does not occur at the initial period (as currently is assumed) but gradually over time or at the moment when the loan is redeemed. This is left for future work.
Sensitivity analysis
Impact of inflation
The rate of inflation can have a significant impact on METRs. This is because, under nominal tax systems, inflationary gains will be taxed as well as real gains – so a higher rate of inflation will lead to a higher METR. Inflation rates in the five years to 1 July 2016 were low, averaging 1.59% across the OECD, although rates ranged from 8.14% in Turkey to -0.45 in Switzerland.
The impact of inflation on METRs is illustrated in Figure 3.7 which compares METRs faced by an average rate taxpayer on bank deposits calculated with the actual country inflation rate and with a zero inflation rate. The high inflation rate in Turkey can be seen to increase the METR on bank interest significantly from 15% (with no inflation) to 55.7%. However, even relatively low inflation rates can have a significant impact on METRs. For example, the METR on bank deposits for an average rate taxpayer in New Zealand is 41.0% when calculated using New Zealand’s actual inflation rate (1.11%), but would be only 30.0% with no inflation.
Figure 3.7. Marginal effective tax rates on bank interest: zero and actual inflation, 2016
Personal tax rate: 100% of average wage case. Real return: 3%. Expected holding period: 5 years.
In contrast, the slightly negative actual inflation rates in Greece and Switzerland, decrease the METRs on bank accounts as compared to the no inflation case (e.g., from 30.2% to 26.5% in Switzerland). This highlights that negative inflation acts as a tax subsidy as less than the full real return is taxed.
In Colombia and the Netherlands, METR results in Figure 3.7 are the same both with and without inflation. In Colombia this is because real rather than nominal interest is taxed, whereas it is a result of the deemed return approach in the Netherlands. Meanwhile, in seven countries bank interest earned by the average rate taxpayer is untaxed, so METRs are zero irrespective of the inflation rate.
Zero inflation results for other assets are presented in ANNEX B.
Impact of expected holding period
METRs are also sensitive to the assumed expected holding period. There are several reasons for this. For assets producing capital gains, a longer expected holding period will mean the asset may be taxed on realisation at a lower – or even a zero – tax rate. Deferral effects on realised capital gains will also increase the longer the holding period. Meanwhile, one-off transaction taxes will effectively be spread over a longer period of time under a longer expected holding period, thereby lowering their effect on the METR.
The impact that the holding period can have on METRs is illustrated in Figure 3.8 which compares METRs faced by an average rate taxpayer on an equity investment (with 50% dividend distribution) for a six-month, five-year, and ten-year holding period. The impact of transaction taxes can be seen in the three countries with the highest METRs for a six-month holding period (Greece, Ireland and Korea). METRs in these countries fall markedly once the holding period increases and the transaction tax is spread over either five or ten years.
Figure 3.8. Marginal effective tax rates on shares: varying holding periods, 2016
Personal tax rate: 100% of average wage case. Real return: 3%. Inflation: country actual.
In most other countries, METRs fall as the holding period increases due to a reduced capital gains tax rate and/or deferral. For example, in the United States capital gains on equity are taxed at 25% if held for less than one year, but at 15% if held longer. This impact can be seen in the drop in the METR in Figure 3.8 from 29.7% for a six month holding period to 20.7% for a five year holding period. Deferral effects continue to lower the METR for longer holding periods, with a ten-year expected holding period producing a METR of 19.4%.
In seven countries, METRs are positive but stay constant across holding periods. In Bulgaria, Turkey and New Zealand this is because there is no taxation of capital gains even for shorter holding periods.23 , 24 In Lithuania, an average-rate taxpayer is assumed to receive less than an exempt amount of capital gains. In Argentina and Norway (given the modelling of the risk-free return allowance) only a wealth tax is applied to equity, which is not affected by the holding period. In the Netherlands, the deemed return approach similarly means METRs are not affected by the holding period. Meanwhile, METRs are zero in Colombia, Germany and the United Kingdom as neither dividends nor capital gains are taxed for an average-rate taxpayer.
In the Slovak Republic, METRs increase between the six month and five-year holding period. This is because capital gains are assumed to be below a minimum threshold for the six-month holding period scenario, but above the threshold for a five-year holding period. Consistent with other countries, the results then show the impact of deferral lowering the METR between five- and ten-year holding periods.
Results for alternate expected holding periods for different assets are presented in ANNEX B.
Impact of pre-tax real rate of return
The choice of pre-tax real rate of return will also affect METRs. This is because, in almost all cases, the base of a tax is not solely the pre-tax real rate of return – it is either the nominal rate, or some other base such as the asset value or increase in asset value. Additionally, a wealth tax or a one-off transaction tax will be a smaller proportion of a higher return. As a consequence, an increase in the pre-tax real rate of return will result in a less than proportional increase in tax liability.
The impact that the real return can have on METRs is illustrated in Figure 3.9 which again focuses on an equity investment (with 50% dividend distribution), but this time for a high-rate taxpayer. METRs are compared for real returns of 2%, 3% and 4%. As expected, METRs fall as the real return increases in almost every country.
Figure 3.9. Marginal effective tax rates on shares: varying real rates of return, 2016
Personal tax rate: 500% of average wage case. Holding period: 5 years. Inflation: country actual.
The greatest variation in METRs occurs in Argentina and the Netherlands. In the Netherlands, this is a logical consequence of the deemed return approach – which applies a 30% tax rate on a deemed 4% real return, irrespective of the actual return. The METR presented in Figure 3.9 is, by design, 30% when the real return is 4%. However, when the real return falls below 4%, the deemed rate, and hence the tax burden, does not change – and so METRs increase to 40% and 60%, for a 3% and 2% real return, respectively.
In Argentina, the wealth tax imposed on equity (and other asset holdings) has a similar effect on METRs as the Netherlands’ deemed return approach (indeed a wealth tax can be thought of as equivalent to applying a fixed tax rate on an assumed rate of return). As the rate of return falls, the wealth tax burden does not change and hence METRs increase. A wealth tax has a similar impact in Norway and France, and to a lesser extent in Switzerland (and in Spain, though this is not captured in the modelling due to its high minimum threshold).
Results for alternate pre-tax real rates of return for different assets are presented in ANNEX B.
Residential property
METRs on residential property can be particularly affected by the assumptions made in the analysis. For example, the preponderance of property transaction taxes and recurrent property taxes can be expected to lead to significant variation in METRs depending on holding period and real return. This can be seen in Figure 3.10 which replicates Figure 3.9 but for debt-financed residential rental property and the average-rate taxpayer. Figure 3.10 shows METRs falling in all countries as the real return increases.
Figure 3.10. Marginal effective tax rates on debt-financed rented residential property: varying real rates of return, 2016
Personal tax rate: 100% of average wage case. Holding period: 20 years. Inflation: country actual.
In addition to the basic assumptions regarding inflation, holding period and real return, an assumption regarding the split of return between (imputed or actual) rental income and capital gain is necessary. The METRs presented in the main results assume that the real return is split equally between rental income and capital gains. However, in some countries, and depending on market conditions, capital gains may be a far more significant component of the overall return from residential property investment. Given that capital gains are often taxed at lower rates as compared to rental income, this can significantly reduce the tax actually faced on a residential rental investment.
To illustrate the potentially large effect on METRs of a market generating substantial returns in the form of capital gains, Figure 3.11 compares the standard METR results (from Table 3.3) for debt-financed residential property with a high capital gain scenario. This scenario assumes a real return of 7% is generated entirely from capital gains.25 Figure 3.11 shows returns significantly fall in such a scenario compared to the standard results. For example, in Australia the METR falls from 88% to just 25%, with similar reductions occurring in other countries under this scenario.
Figure 3.11. Marginal effective tax rates on debt-financed rented residential property: high capital gain scenario, 2016
Personal tax rate: 100% of average wage case. Holding period: 20 years. Inflation: country actual.
Another factor that can significantly affect METRs on debt-financed residential property is the presence of tax relief for mortgage interest payments. In 26 of the 40 countries covered in this study tax relief (either a deduction or tax credit) for mortgage interest is provided for debt-financed rental property. Figure 3.12 compares the standard METR results for these countries (from Table 3.3) with the METRs that would apply in the absence of mortgage interest relief.
Figure 3.12. Marginal effective tax rates on debt-financed rented residential property: with and without mortgage interest tax relief (selected countries), 2016
Personal tax rate: 100% of average wage case. Holding period: 20 years. Inflation: country actual.
Figure 3.12 shows that, in all but three countries, the presence of mortgage interest tax relief significantly reduces METRs. In Colombia and Korea the deductibility of mortgage interest against rental income does not impact the METRs as an average rate taxpayer faces no tax on rental income. Meanwhile in Estonia an average rate taxpayer is assumed to already be deducting the maximum EUR 1 200 per year, so receives no further benefit on the marginal euro of saved.
Individual country graphs
Figure 3.13. Marginal effective tax rates: Argentina, 2016
Actual inflation rate
Figure 3.14. Marginal effective tax rates: Australia, 2016
Actual inflation rate
Figure 3.15. Marginal effective tax rates: Austria, 2016
Actual inflation rate
Figure 3.16. Marginal effective tax rates: Belgium, 2016
Actual inflation rate
Figure 3.17. Marginal effective tax rates: Bulgaria, 2016
Actual inflation rate
Figure 3.18. Marginal effective tax rates: Canada, 2016
Actual inflation rate
Figure 3.19. Marginal effective tax rates: Chile, 2016
Actual inflation rate
Figure 3.20. Marginal effective tax rates: Colombia, 2016
Actual inflation rate
Figure 3.21. Marginal effective tax rates: Czech Republic, 2016
Actual inflation rate
Figure 3.22. Marginal effective tax rates: Denmark, 2016
Actual inflation rate
Figure 3.23. Marginal effective tax rates: Estonia, 2016
Actual inflation rate
Figure 3.24. Marginal effective tax rates: Finland, 2016
Actual inflation rate
Figure 3.25. Marginal effective tax rates: France, 2016
Actual inflation rate
Figure 3.26. Marginal effective tax rates: Germany, 2016
Actual inflation rate
Figure 3.27. Marginal effective tax rates: Greece, 2016
Actual inflation rate
Figure 3.28. Marginal effective tax rates: Hungary, 2016
Actual inflation rate
Figure 3.29. Marginal effective tax rates: Iceland, 2016
Actual inflation rate
Figure 3.30. Marginal effective tax rates: Ireland, 2016
Actual inflation rate
Figure 3.31. Marginal effective tax rates: Israel, 2016
Actual inflation rate
Figure 3.32. Marginal effective tax rates: Italy, 2016
Actual inflation rate
Figure 3.33. Marginal effective tax rates: Japan, 2016
Actual inflation rate
Figure 3.34. Marginal effective tax rates: Korea, 2016
Actual inflation rate
Figure 3.35. Marginal effective tax rates: Latvia, 2016
Actual inflation rate
Figure 3.36. Marginal effective tax rates: Lithuania, 2016
Actual inflation rate
Figure 3.37. Marginal effective tax rates: Luxembourg, 2016
Actual inflation rate
Figure 3.38. Marginal effective tax rates: Mexico, 2016
Actual inflation rate
Figure 3.39. Marginal effective tax rates: Netherlands, 2016
Actual inflation rate
Figure 3.40. Marginal effective tax rates: New Zealand, 2016
Actual inflation rate
Figure 3.41. Marginal effective tax rates: Norway, 2016
Actual inflation rate
Figure 3.42. Marginal effective tax rates: Poland, 2016
Actual inflation rate
Figure 3.43. Marginal effective tax rates: Portugal, 2016
Actual inflation rate
Figure 3.44. Marginal effective tax rates: Slovak Republic, 2016
Actual inflation rate
Figure 3.45. Marginal effective tax rates: Slovenia, 2016
Actual inflation rate
Figure 3.46. Marginal effective tax rates: South Africa, 2016
Actual inflation rate
Figure 3.47. Marginal effective tax rates: Spain, 2016
Actual inflation rate
Figure 3.48. Marginal effective tax rates: Sweden, 2016
Actual inflation rate
Figure 3.49. Marginal effective tax rates: Switzerland, 2016
Actual inflation rate
Figure 3.50. Marginal effective tax rates: Turkey, 2016
Actual inflation rate
Figure 3.51. Marginal effective tax rates: United Kingdom, 2016
Actual inflation rate
Figure 3.52. Marginal effective tax rates: United States, 2016
Actual inflation rate
References
Harding, M. and M. Marten (2018), “Statutory tax rates on dividends, interest and capital gains: the debt equity bias at the personal level”, OECD Taxation Working Papers, No. 34, OECD Publishing, Paris.
IBFD Tax Research Platform, www.ibfd.org.
King, M. and D. Fullerton (1984), Taxation of income from capital: a comparative study of the United States, United Kingdom, Sweden and West Germany, Chicago University Press, Chicago.
OECD (1994), “Taxation and Household Saving”, OECD Publishing, Paris.
OECD Tax Database, www.oecd.org/tax/tax-policy/tax-database.htm.
Notes
← 1. The Taxation and Household Saving study (OECD, 1994) calculated METRs for a group of five different assets: bank deposits, government bonds, equities, pensions and owner-occupied housing. The case of government bonds was subdivided into those issued at a discount and those issued and redeemable at par (in order to illustrate the effect of differential treatment of interest and capital gains). Different effective tax rates for pensions were calculated as well, focusing on the cases where pension premiums were deductible or non-deductible (if applicable) and on the cases where pension income was taxed at the standard tax schedule or at lower rates. Investment in owner-occupied housing was either fully financed with equity (own funds) or partly with equity (own funds) and borrowed funds.
← 2. With the exception of Argentina, where, due to recent inflation volatility, the five-year average from Q1 2008 to Q4 2013 has been used.
← 3. For example, Denmark has a semi-dual tax system where returns to household savings are mostly taxed separately from (and at different rates than) labor income, and where the (marginal) tax rates that apply to a given category of savings return will often depend on the size of this specific type of return – meaning the appropriate tax rate depends on the size and composition of the household’s savings return – but not necessarily the size of the household’s labour income and thus the household’s overall level of income. METR results are consequently driven significantly by the assumptions made regarding capital income levels.
← 4. Assumptions for average- and high-rate taxpayers regarding wealth levels, and the mix between housing and non-housing wealth, are approximations based on unweighted averages of asset holdings for different income groups across eighteen European Union countries calculated from 2016 European Central Bank Household Finance and Consumption Survey (HFCS) microdata.
← 5. For the Netherlands, general scheme SSCs (“premies volksverzekeringen”) are included in the modelling. The SSCs paid by employers and employees have not been included.
← 6. An outlier is defined as a result more than 1.5 times the inter-quartile range below the first quartile or above the third quartile. Where outliers are present, the upper and lower points of the box-and-whisker plot are set equal to this limit.
← 7. In a small number of countries a transaction tax is also applied to deposits in bank accounts. In Italy a reduced 0.2% transaction tax is applied under the assumption that the account is subject to some form of time constraint, in return for better interest rates.
← 8. All interest income from bank accounts is exempt in Argentina. In Estonia, interest income is exempt as long as it is received from deposits with a credit institution which is a resident of a Contracting State of the European Economic Area (EEA) or on account of a permanent establishment of a credit institution located in a Contracting State of the EEA Agreement, and the interest does not partially or completely depend on the value, or change in value, of a security, deposit, currency, other instrument or the underlying assets thereof.
← 9. In such cases, the level of interest income earned for each taxpayer type is determined by applying a 3% real return to the wealth levels for each taxpayer type specified in 3.2.
← 10. Note that, as of 2017, the deemed return applied on net assets increases with the net value of the assets. The tax rate applied continues to be 30%.
← 11. A taxpayer does not have to return income from a financial arrangement on an accrual basis if they are a cash basis person. To be a cash basis person, the difference in the person’s annual income on a cash basis and on an accrual basis must not exceed NZD 40 000; and either: the person’s income and expenditure under all financial arrangements for the income year does not exceed NZD 100 000; or the value of the person’s financial arrangements during the year does not exceed NZD 1 million.
← 12. The Estonian results provide an example of the limitation created by only modelling taxation at the personal level. In Estonia, tax is not paid at the corporate level until distribution of profits as dividends, with the dividends then exempt in the hands of the shareholder. Our modelling results consequently show a zero METR at the personal level for Estonia with 100% distribution. The results imply that there will be an incentive to distribute dividends to avoid paying the 20% capital gains tax, whereas there will in fact be an incentive to retain profits in the company to defer payment of the corporate income tax.
← 13. Effective 2006, Canada introduced an enhanced gross-up and dividend tax credit regime for dividends distributed by large corporations, which are subject to a higher statutory rate than small businesses. As a result, Canada operates a dual rate gross up and dividend tax credit system that provides full imputation at the federal level (a number of provinces responded to the federal initiative by adjusting their own dividend tax credit rates). Rates modelled are those applicable to large corporation dividends.
← 14. Note that thresholds can vary across asset types. For example, Spain provides a general threshold (which covers most assets including investment fund interests) and a specific threshold for property which is lower.
← 15. The statutory rate itself need not be reduced as long as there is some form of concessionary treatment on distribution. For example, distributions in the United Kingdom are taxed at the taxpayer’s marginal PIT rate, but 25% may be distributed tax free immediately as a lump sum lowering the effective tax rate on distribution.
← 16. The “Riester-Rente” pension fund is modelled for Germany. Private cash-value life insurance plans are also available. Premiums paid for such life insurance plans are not tax deductible.
← 17. In the Netherlands, tax deductible contributions to private pension funds are limited to 13.8% of annual income (with a ceiling of EUR 101 519 in 2016) minus a threshold for the general state pension and taking into account accrued pensions rights in an occupational pension plan. Consequently, the marginal contribution by a high-rate taxpayer will not be deductible. The marginal contribution will instead be subject to a TEE regime, resulting in a zero METR.
← 18. A similar result occurs for low-rate and average-rate taxpayers in Australia. However, a top-rate taxpayer is assumed to be above the “untaxed plan cap amount” (AUD 1 395 000 in 2015-16) and, hence, liable to the top marginal tax rate on distributions which results in a large positive METR.
← 19. Given the requirement for retention in the account, full retention has been assumed in the modelling for Colombia, Hungary and Spain, whereas for other countries contributions to tax-favoured savings accounts are assumed to be distributed annually and reinvested.
← 20. This is likely to be the case for other countries. However, current limitations of the modelling, specifically the use of “average” property tax rates, may prevent their appearance in the current METR results. This modelling aspect could potentially be adjusted for the final project report.
← 21. It should be borne in mind that the choice of expected holding period can significantly affect METRs on residential property, particularly in countries that impose substantial transaction taxes as these fixed taxes are effectively apportioned across a larger or smaller number of years depending on the expected holding period modelled. The longer the expected holding period, the lower the METR will be.
← 22. METR calculations for saving in residential property when the household does not own the necessary funds would require a different modelling approach than the one followed in this report. Households who do not have savings, but want to purchase a house, will typically borrow from a financial institution with the commitment to pay for the principal investment in the house at a later point in time. The interest, net of mortgage interest relief, that has to be paid on the loan will then increase the cost of the investment. Instead of saving in the initial period (as assumed in this report), the saving event would arguably then occur in the future period when the household pays the financial institution for the principal. Such analysis is left for future work.
← 23. In the case of Chile, capital gains are untaxed as long as the taxpayer is not habitual in the trade of shares and the capital gains are less than approximately CLP 5.5 million (in 2016). In New Zealand, capital gains on shares are also untaxed as long as the taxpayer is not considered to be in the business of trading shares.
← 24. The higher METR for a holding period of six months in Belgium is due to a short-term capital gains tax that was in force at the reference date (1 July 2016). The tax has now been repealed, so there is no longer any differentiation of METRs across holding periods in Belgium.
← 25. For simplicity of modelling, it is also assumed that there is no economic depreciation.