Chapter 5. The long-term fiscal cost of financial incentives to promote savings for retirement
This chapter provides an assessment of the cost of promoting savings for retirement via tax and non-tax financial incentives from the point of view of fiscal policy and the budget. It first defines the concept of tax expenditure and presents how countries report the cost related to financial incentives for retirement savings. It then introduces a measure that allows for cross-country comparisons and calculates that measure for a selection of countries.
Financial incentives to promote savings for retirement represent a cost for governments. Tax incentives translate into lower personal income taxes paid by individuals and thus create a fiscal cost for the government. By contrast, non-tax financial incentives come directly from the general budget. The total cost of those tax and non-tax incentives depends on the generosity of the incentives and the amount contributed into retirement savings plans in each year. In an era of budget stringency, this type of cost has come under scrutiny.
This chapter provides an assessment of the cost of promoting savings for retirement via tax and non-tax financial incentives from the point of view of fiscal policy and the budget. Its main objective is to examine the fiscal cost of tax and non-tax financial incentives for retirement savings plans and its evolution over the next 45 years in selected OECD countries. This chapter first looks at national tax expenditure reports, focusing on the tax expenditures related to private pensions. After identifying the limitations of the measures used in these tax expenditure reports, this chapter estimates the current and future profile of the net tax expenditure related to private pensions in Australia, Canada, Chile, Denmark, Iceland, Latvia, Mexico, New Zealand, the Slovak Republic and the United States. The net tax expenditure in a given year is measured as the net amount of personal income tax revenues forgone on contributions, revenues forgone on accrued income and revenues collected on withdrawals that arise when comparing with the tax treatment of traditional savings accounts (“Taxed-Taxed-Exempt” or “TTE” tax regime benchmark). It adds to direct spending associated with non-tax incentives, when relevant, in order to estimate a total fiscal cost.
The analysis shows that the total fiscal cost of financial incentives varies greatly across countries, but remains in the low single digits of GDP. The time profile of the fiscal cost depends on the level of maturity of the pension systems and the countries’ age profile. In some countries, direct spending in the form of matching contributions or fixed nominal subsidies represents an important component of the total fiscal cost related to the promotion of savings for retirement. Accounting for corporate income tax revenues and the potential effects of new savings would reduce the cost of financial incentives.
The chapter proceeds as follows. Section 5.1 defines the concept of tax expenditure and provides the tax expenditures related to private pensions and their distribution by income for selected countries, as reported nationally. Considering the limited comparability of the measures used in tax expenditure reports across countries, this chapter looks at an additional indicator, the net tax expenditure, which is comparable across countries and measures the current and future cost of tax incentives for private pension plans. Before calculating the indicator for selected countries, Section 5.2 explores the relationship between the net tax expenditure and different economic variables, the maturing of pension systems, and demographic trends. Section 5.3 provides the results of estimating the net tax expenditure between 2015 and 2060 for selected OECD countries. Direct spending associated with government non-tax incentives is added to estimate a total fiscal cost. Section 5.4 discusses the potential impact of incorporating corporate income tax and new savings effects into the calculation of the net tax expenditure. Section 5.5 concludes. Annex 5.A presents the approach used to project the future profile of the cost of tax incentives for retirement savings, as well as country-specific assumptions used in the calculations.
5.1. Tax expenditures: Definitions, country reporting and limitations to comparability
Definition and calculation methods
The budgetary cost of a particular tax policy measure is generally referred to in the literature as a “tax expenditure”. A tax expenditure can be seen as a public expenditure carried out through the tax system via a special tax concession that results in reduced tax liability for certain subsets of taxpayers. In practice, tax expenditures are defined as “deviations from a tax norm or a benchmark that result in a reduced tax liability for the beneficiaries, who are generally a particular group of taxpayers or an economic activity” (OECD, 2010[1]). According to OECD (1996[2]), tax expenditures may take a number of different forms: exemptions (income excluded from the tax base), allowances (amounts deducted from gross income to arrive at taxable income), credits (amounts deducted from tax liability), rate relief (a reduced rate of tax applied to a class of taxpayers or activities) and tax deferrals (a relief which takes the form of a delay in paying tax).
The main challenge in any analysis of tax expenditures is to identify the benchmark tax system against which to establish the nature and extent of any tax concession (OECD, 2010[1]). In general, the benchmark tax system is the regular tax arrangements that apply to similar classes of taxpayers or types of activity within countries. This may involve an element of judgment on what the regular tax arrangements are. Benchmarks may therefore vary across countries and also within countries over time.
There are three principal methods for estimating tax expenditures. First, the revenue forgone method measures the amount by which tax revenues are reduced by a particular tax concession, under the assumption of unchanged behaviour from the taxpayer.1 Second, the outlay equivalent method measures the cost of providing the same monetary benefit through direct spending, assuming also that behaviour is unchanged as a result of the tax concession. Third, in contrast with the previous two methods, the revenue gain method takes potential behavioural responses into account and provides an ex ante measure of the expected increase in revenues if the concessions were repealed.
For each method, the cash-flow or present-value approaches can be used in the calculation. The cash-flow approach estimates the effect on government cash flow in a given year. The present-value approach estimates the effect on the tax liability accruing to government in a particular period. The cash-flow approach allows assessing the evolution over time of government revenue cash flows.
Country reporting of the cost related to financial incentives to promote savings for retirement
Tax expenditure reporting began in Germany and the United States in the late 1960s, with other countries introducing it in the late 1970s or later. In some countries, the authorities are legally obliged to produce a tax expenditure report (e.g. Austria, Belgium, France, Germany and the United States). The reports usually cover tax expenditures in the personal income tax as well as the corporate income tax. However, the actual definition and practice for estimating tax expenditures vary across countries, in particular for the choice of a tax benchmark. In the case of tax-favoured private pension plans, traditional savings accounts represent natural benchmarks. These usually follow a “TTE” tax regime. However, some countries have several types of savings account with different tax treatments, so the identification of the benchmark is not always straightforward.
Table 5.1 provides tax expenditures related to the tax treatment of private pension plans in selected OECD countries for 2017 or the latest year with available information. For most countries, the information comes from official tax expenditure reports published on the relevant authorities’ websites. Table 5.1 shows that the number of items included in the reports related to the tax treatment of private pension plans varies greatly across countries. The significance of the tax expenditure items as a percentage of GDP varies from negligible (e.g. Mexico and the Slovak Republic) to around 1% of GDP (e.g. Australia, Canada, the United Kingdom).
Some countries also report the distribution of tax expenditure across income groups. These tend to show that the tax expenditures are concentrated within high-income households. As shown in Figure 5.1, individuals in the top 20-25% of the income distribution receive around 80% of the tax expenditure related to retirement savings in Canada, 71% in Ireland and 66% in the United States.2 In voluntary pension systems, high-income earners are more likely to participate in private pension plans than other income groups, reflecting their higher propensity and capacity to save. Moreover, their higher income results in a greater value of the tax relief per household, even when the tax rate is same rate for everyone. Finally, in progressive tax systems where tax rates increase with taxable income, high-income earners typically face the highest marginal tax rates, thereby benefiting more on every unit of the flows that attract tax relief.
Note: “No TI” means no taxable income in the case of New Zealand. The distributions are on an individual basis, except for the United States (household basis).
Source: Macdonald (2016[4]) for Canada, Collins and Hughes (2017[5]) for Ireland, Inland Revenue data for New Zealand, Pensions Policy Institute (2016[6]) for the United Kingdom, and Congressional Budget Office (2013[7]) for the United States.
In the United Kingdom, 70% of tax expenditures related to the tax treatment of private pensions goes to individuals earnings between GBP 20 000 and GBP 75 000 and the part received by very high-income earners has declined over the period 2009-10 to 2014-15. The share of the tax expenditures paid to individuals earning above GBP 150 000 has declined from 20% to 7% over the period. This is due to the recent reforms that limited tax relief on pensions and affected more high-income individuals. Indeed, the maximum amount of pension contributions that an individual can get tax relief on in each tax year (the annual allowance) has been reduced from GBP 255 000 in 2010-11, to GBP 50 000 between 2011-12 and 2013-14, and is now GBP 40 000 since 2014-15. The lifetime allowance also limits the total amount that individuals can accumulate in a private pension plan. It was introduced in 2006-07 at a level of GBP 1.5 million. It then increased each year to 2010-11, when it reached a level of GBP 1.8 million. Since 2012-13, the lifetime allowance has been reduced and its level since April 2018 is GBP 1.03 million.
The KiwiSaver system in New Zealand illustrates the fact that matching contributions can reach a more equal treatment of government spending across income groups. The government makes an annual contribution towards KiwiSaver accounts (called member tax credit) of 50 cents for every dollar of member contribution annually, up to a maximum payment of NZD 521.43. Figure 5.1 shows that member tax credit payments are evenly distributed for individuals in the top 60% of the income distribution (i.e. fourth decile and up). This is due to the cap on the state contribution. However, individuals at the bottom of the income distribution get proportionately less.
Finally, fixed nominal subsidies also help rebalancing the distribution of government spending with income when combined with tax incentives. In Germany, the government encourages people to contribute to a Riester plan through two types of incentive: tax exemptions and fixed nominal subsidies. There are three types of subsidy: a basic subsidy of EUR 175 per year and per person, a child subsidy of EUR 300 per year and per child and a young worker subsidy of EUR 200 granted once at the age of 25. In addition, all Riester savings up to EUR 2 100 (including subsidies) can be claimed as a special expense deduction in the annual income tax return. Corneo, Schröder and König (2015[8]) assess the distributional impact of the Riester pension scheme using survey data from 2012. They find that the average subsidy (in the form of fixed nominal subsidies or tax deductions) increases with income, from EUR 23.56 in the bottom decile, to EUR 56.83 at the 6th decile and to EUR 156 for the top decile. Around 38% of the total subsidy paid by the government (tax expenditure and direct spending) goes to the 20% richest individuals, while only 7.3% accrues to the 20% poorest ones. The fact that high-income earners participate more to the Riester pension scheme explains this result. Although the total subsidy is still skewed towards high-income households, the share going to low-income households is much higher than in countries having only tax incentives (the 20-25% poorest individuals receive less than 1% of the tax expenditure related to private pensions in Canada, 0.4% in Ireland, and 2% in the United States).
Limitations of national tax expenditure reports for cross-country comparisons
Although all the countries listed in Table 5.1 calculate tax expenditures using similar methodologies (revenue-forgone method with the cash-flow approach), tax expenditures related to the tax treatment of retirement schemes are difficult to compare across countries.
The items reported as tax expenditures vary across countries. For example, some countries include the tax collected on pension withdrawals, reporting it as a negative tax expenditure, (e.g. Australia, Canada, Switzerland and the United Kingdom), while other countries only communicate tax revenues forgone (e.g. Belgium, France and Ireland). In the United States, both tax expenditure reports account for the net result of summing up revenues forgone on contributions, revenues forgone on investment returns and revenues collected on withdrawals, but individual flows are not reported (there is only a breakdown by type of plan). In Austria, Germany and New Zealand, tax expenditure reports account for subsidies or matching contributions paid into the pension accounts of entitled individuals. These are direct spending, not tax expenditures. Finally, in Sweden, the tax exemption of contributions to occupational pension plans is not considered as a tax expenditure. The tax expenditure report only accounts for revenues forgone due to the reduced tax on returns.
Some countries clearly warn that different tax expenditure items cannot be summed up, while other countries include net aggregate numbers (revenues forgone on contributions plus revenues forgone on accrued income minus revenues collected on withdrawals) in their report. For example, the Australian Treasury mentions in its Tax Expenditure Statement that it is not appropriate to aggregate revenue-forgone estimates because they do not take into account potential changes in taxpayer behaviour following the hypothetical removal of a tax concession. In addition, the tax expenditure resulting from the assumption that all the tax concessions were cancelled at the same time (e.g. changing simultaneously the tax treatment of pension plans for contributions, investment returns and withdrawals) is different from the sum of the tax expenditures of each tax concession. For instance, if only the tax concession on contributions was removed (i.e. contributions would no longer be deductible from income), individuals would save less through these plans, implying a reduction in tax revenues forgone on investment returns.
Finally, the revenue-forgone calculation of the tax exemption of investment returns accruing in pension funds is likely to be overestimated in a number of cases. Indeed, the investment income achieved by pension funds is likely to be larger than the one that would be achieved in the tax benchmark. Usually, the tax benchmark is some version of the comprehensive income tax regime (“TTE”), and so after-tax contributions accumulating in the benchmark are lower than those accumulating in a pension fund for which contributions are tax-deductible. In the United States, the cash-flow calculations do not take this into account and assume that all of the money contributed to a pension plan is, instead, contributed to a benchmark taxable account. This therefore overestimates tax revenues forgone on investment returns. The present-value calculations, however, assume that only after-tax contributions accumulate in the tax benchmark.
This chapter estimates subsequently the net tax expenditure associated with the tax treatment of retirement savings, today and into the future, based on the revenue-forgone method and using the cash-flow approach. Tax expenditure reports are not suited to compare the burden related to the tax treatment of retirement savings on the government budget across countries. There is therefore a need for another measure. The net tax expenditure in a given year is calculated as the net amount of revenues forgone on contributions, revenues forgone on accrued income and revenues collected on withdrawals. Revenues forgone on accrued income are calculated taking into account the reduction coming from taxed contributions in the benchmark.3 This methodology is well suited to capture the influence of demographic changes as it looks at the profile of the net tax expenditure at different points in time. Section 5.2 demonstrates the convenience of this measure with the use of simple and straightforward examples and Section 5.3 presents the results for ten selected OECD countries.4 Annex 5.A provides the full methodology.
5.2. Simple and straightforward models of the net tax expenditure
This section explores the relationship between the net tax expenditure and economic variables, the maturity of pension systems, and demographic trends with the help of simple and straightforward models. These models can assist in better understanding the mechanisms behind the net tax expenditure and thus understand the different factors behind the country-specific results (Section 5.3).
For retirement savings plans where both contributions and returns on investments are exempted from taxation while benefits are treated as taxable income upon withdrawal (“Exempt-Exempt-Taxed” or “EET” tax regime), the cash-flow net tax expenditure for year t, NTEt, with respect to a “TTE” benchmark, may be expressed as
Equation 1
Ct and Bt are contributions to and benefits paid out of private pension plans in year t, At–1 is the level of assets in the plans at the end of the previous year, i is the nominal pre-tax rate of return on plan assets in the year and µC, µA and µB are the average marginal personal income tax rates applicable to contributions, investment income and benefits respectively.5 The parameters i, µC, µA and µB are assumed constant over time.
When only taking into account personal income tax revenues, the net tax expenditure incurred by an “EET” arrangement as compared to a “TTE” benchmark has three components. Private pension contributions are tax-deductible, unlike savings under the benchmark tax system (e.g. a traditional savings account). The first term of Equation 1, µCCt, therefore represents tax revenues forgone on contributions, i.e. the amount of tax that would have been collected by the Treasury had the contributions been invested in a benchmark savings vehicle.
Investment income accrues tax free in a private pension plan but is taxed in the benchmark savings vehicle. The second term of Equation 1, µAiA't-1, therefore represents tax revenues forgone on accrued income. The relevant stock of assets to calculate the revenue forgone on accrued income is not the total assets invested in the pension scheme (At-1) but those accumulated from diverted savings (A't-1). The reason for including only a subcomponent of total assets is that some of the contributions would not have generated investment income in the absence of the tax-favoured scheme. Contributions can be decomposed between the tax break (forgone tax revenues on contributions), µCCt, and personal savings, (1 – µC)Ct. Personal savings can in turn be split between diverted savings, (1 – α)(1 – µC)Ct, and new savings, α(1 – µC)Ct, where α is the share of personal savings stemming from a decline in consumption (new savings). Since neither the new savings nor the tax break would have generated investment income in the absence of the tax-favoured scheme, they have to be excluded from the calculation of forgone tax revenues on accrued income, which requires a second asset accumulation based on diverted savings alone. The two asset accumulation equations may be expressed as
Equation 2
Equation 3
Finally, the third term of Equation 1, µBBt, represents tax revenues collected on withdrawals from the private pension system (withdrawals are tax-free in the benchmark).
Changes in the level of the net tax expenditure over time depend primarily on differences in the growth patterns of contributions, assets and benefits. To produce a negative value of the net tax expenditure, having a higher level of benefits than contributions is not sufficient. The tax collected on benefits must exceed the forgone tax on both contributions and accrued income.
Making the simplifying assumptions of a single marginal tax rate, µC = µA = µB = µ, and that all contributions come from diverted savings (i.e. no new savings, α=0), the net tax expenditure may be written as
Equation 4
This shows that the net tax expenditure is positive as long as the year-over-year rate of increase in assets is larger than µi.
Steady state
Assessing the steady state can help to better understand the mechanisms behind the net tax expenditure. In steady state, the pension system is mature (i.e. the contribution and benefit levels as a percentage of earnings are constant), there are no demographic shifts (i.e. the population at each age group is equal and grows at a constant rate over time), the rate of return on assets is constant, and earnings (and GDP) grow at a constant rate. In this case, contributions, assets, benefits and the net tax expenditure will all be constant fractions of GDP. Let Yt be GDP, k the constant share of assets over GDP, so that At = kYt, and g the growth rate of GDP, so that Yt = (1+g)Yt–1.6
In steady state, the net tax expenditure will be positive as long as the growth of GDP (determined by population growth, real wage growth and inflation) is greater than the product of the marginal tax rate and the nominal pre-tax rate of return on plan assets. The latter is a tax wedge on returns on investment that measures the rate of loss in returns due to taxation. Rearranging Equation 4 and using the steady state assumptions above, the net tax expenditure can be expressed as a fraction of GDP as
Equation 5
Another interesting steady-state result concerns the relationship between contributions and benefits. By rearranging Equation 2 and using the previous assumptions, the excess of benefits over contributions can be expressed as a fraction of GDP, as
Equation 6
This shows that benefits, funded out of both contributions and investment income, can exceed contributions on a permanent basis as long as the nominal pre-tax rate of return is greater than the GDP growth rate. The size of the excess depends on the asset level and the difference between the rate of return on assets and GDP growth. For example, assuming µ = 0.30, k = 1.3, g = 0.0328 (1.25% real growth and 2% inflation) and i = 0.0506 (a 3% real rate of return and 2% inflation), the net tax expenditure equals 0.66% of GDP and the excess of benefits over contributions equals 2.2% of GDP. With a contribution level of 3.0% of GDP for example, benefits are permanently 75% higher than contributions (5.2% vs. 3.0% of GDP).
Even where benefits well exceed contributions, the net tax expenditure remains positive in the steady state if nominal GDP growth is above “µi”. Equation 2 and Equation 4 can be rearranged to show that the net tax expenditure will be positive as long as the amount of investment income generated in the year by personal savings is greater than the excess of benefits over contributions. This point is relevant to non-steady-state cases as well. In most situations where benefits substantially exceed contributions, they will be accompanied by high levels of assets and investment income.
The steady-state net tax expenditure is larger when the average tax rate is lower during retirement than while working. Individuals may have lower income during retirement than while working, thereby reducing their average tax rate on withdrawals and thus increasing the net tax expenditure. For example, assuming µB=0.20 instead of µB=0.30 (while keeping µC=µA=0.30), increases the net tax expenditure from 0.66% of GDP to 1.19% of GDP. This stems from the fact that tax revenues collected on pension withdrawals are lower, while tax revenues forgone on contributions and accrued income are still the same.
Steady-state results provide a reference point to be able to assess future tax expenditure levels. Changes in economic variables, such as investment returns and the rate of wage growth, the level of maturity of the pension system and demographics (different sizes of age cohorts, e.g. baby-boom cohorts, immigration) will provide results that deviate from the steady state. Notwithstanding, the steady state provides an indication of the magnitude of the deviations.
The maturing of the pension system
A mature pension system is one in which current retirees draw their pension based on a full career and constant contribution rules. Individuals who were already in the labour market when a new pension system was introduced will draw a pension which accumulated for part of their career only. Similarly, if the contribution rate to the system changes, it will take some time before retirees start drawing a pension which accumulated at the new contribution rate for their entire career. While the system is still maturing, aggregated asset and benefit levels increase over time until they reach a stable level (steady state).
A simple, straightforward model of the retirement saving process can help in understanding the effect of introducing a new private pension system on the net tax expenditure over time.7 In this model, individuals save from age 20 to 64 and draw pension benefits from age 65 to 84. The pension of each individual is drawn according to a certain annuity with fixed nominal payments that, by age 84 (when the individual passes away), exhausts the assets accumulated at age 64. The number of people in each single-year age cohort is assumed to be equal.
The net tax expenditure rises sharply when the contributions to the new pension system begin and converges to the steady state over time. Figure 5.2 presents projections of net tax expenditure levels for two cases. In the steady state, contribution levels are constant across (pre-retirement) age groups and constant over time. As above, the return on assets is 5.06% and the rate of wage growth is 3.28%. With an aggregate contribution level of 3% of GDP and a constant marginal tax rate of 30% on contributions, investment income and benefits, the resulting net tax expenditure is constant at 0.66% of GDP. In the second case of a pension system that is maturing, the model assumes that the pension system was introduced in year t and that contributions started in that same year. The pension of someone reaching age 65 in year t+10, for example, is based on only 10 years of contributions. The net tax expenditure first rises well above its steady-state level before declining back to it. The year when the pension system is introduced, NTEt= µCt (30% of 3% of GDP = 0.90% of GDP).8 The system attains full maturity and reaches its steady-state level of 0.66% of GDP only in year t+64, the first year in which all of the retirees (up to the 84-year-olds, who were aged 20 in year t) have contributed for their full careers.9
The lag in the growth of benefits behind that of assets and investment income is what creates the temporary increase in the net tax expenditure above its steady-state level. Figure 5.3 helps to explain the drivers of the previous result by displaying how the three components of the net tax expenditure (i.e. Ct, iAt–1 and Bt) evolve over time. As the density of contributions progresses, asset and investment income levels grow and so, with a lag, do benefit levels.
A maturing pension system can create a fairly long period of declining tax expenditures (from t+42 to t+64 in this straightforward model) but only from a level that is temporarily above its steady-state value.
Assessing a pension system’s development as it matures provides a cautionary tale to policy makers. Introducing a pension system with “EET” tax advantages for retirement savings creates a larger upfront net tax expenditure. It is only several decades later that the new pension system and its tax expenditure reach a stationary situation. Therefore, unless this is fully disclosed and accounted for when introducing reforms, the potential upfront fiscal cost may create a political backlash, especially in times of economic crises. This problem may be compounded if, at the time when the new system with “EET” tax advantages is being introduced, the cohorts entering the labour market are larger than those about to retire.
Impact of the ageing of population cohorts that are larger than the previous ones
Larger cohorts entering the labour market than those retiring translates into higher contributions and higher assets until they retire, bringing the net tax expenditure above the steady state for a while and below it afterwards, until the larger cohorts pass away and the steady state is reached again. To investigate the effects of a population bulge on pension flows and the net tax expenditure, the steady-state model is adjusted by increasing the sizes of the cohorts entering the labour market between t+5 and t+24 by 20% (compared to previous and following cohorts). These individuals reach age 65 between t+50 and t+69, and by year t+89 they have disappeared from the model. The results are shown in Figure 5.4 and Figure 5.5.
The net tax expenditure for the larger-cohorts case rises above that of the steady state to a peak in year t+49 and then declines sharply to reach a minimum value in year t+69, the only year in which all of the retired population is composed of individuals in the larger cohorts. After year t+69, the net tax expenditure rises, reaching its steady-state level again in year t+89.
Asset and benefit levels are slightly lower as percentages of GDP up to year t+54 than with no population bulge. After that point they rise to higher levels. This is because the first effect of the larger cohorts is to raise the aggregate levels of earnings, GDP and contributions. Asset and benefit levels increase only with a lag. It is also noteworthy that the kinks in the net tax expenditure and benefits lines in Figure 5.4 and Figure 5.5 result from the simplifying assumption that benefits commence abruptly at age 65 for everyone in the population.
In summary, these simple and straightforward models of the net tax expenditure and the retirement saving process show that both a maturing pension system and the ageing of larger cohorts can create substantial declines in the net tax expenditure level. These declines may even bring the net tax expenditure below the steady-state level.10 However, as long as the economy continues to grow in nominal terms above a rate at least equal to the product between the marginal tax rate and the rate of return (that measures the rate of loss on returns due to taxation), it may not be sufficient to reduce the net tax expenditure to zero or below zero.
5.3. Fiscal cost of tax and non-tax financial incentives for private pensions in selected OECD countries
This section presents the results of projecting the net tax expenditure between 2015 and 2060 arising from the tax treatment of private pension schemes in 10 OECD countries. The net tax expenditure in a given year is measured as the net amount of revenues forgone on contributions, revenues forgone on accrued income and revenues collected on withdrawals, relative to a “TTE” benchmark.11 It adds to direct spending associated with government financial incentives, when relevant, in order to estimate a total fiscal cost.
The amount of information necessary to make these calculations is quite substantial. Generating estimates of the future profile of the net tax expenditure requires projecting forward a number of key variables including the number of contributors, total contributions, total assets, accrued income from assets, number of beneficiaries and withdrawals, all of which will be strongly influenced by future demographic developments. First, in order to calculate the net tax expenditure in the first year (i.e. 2015) it is necessary to have information on the amounts of contributions, accrued income and withdrawals in that year and to have these amounts broken down by age groups. Indeed, individuals of different ages have different income sources and face different tax rates. Applying the same tax rate independently of age as done in the models of Section 5.2 would not reflect the actual dynamics of tax expenditures and receipts. Second, in order to calculate the future value of contributions, accrued income and withdrawals, it is important to have information on the number of people contributing, holding assets and withdrawing benefits in the first year, broken down by age groups. This information is used to build ratios of contributors, asset holders and beneficiaries with respect to the total population or to the total employed population. Keeping these ratios constant in the future and using population projections (e.g. from the United Nations) enables projections of the future number of contributors and beneficiaries, and in turn the future value of contributions, accrued income and withdrawals. In general, therefore, the estimation of the future profile of the net tax expenditure requires a breakdown by age in 2015 of the number of contributors, total contributions, total assets, the number of beneficiaries and total withdrawals.
The country coverage has been primarily conditioned by the amount of available information. The analysis herein focuses therefore on 10 OECD countries for which the OECD Secretariat was able to get appropriate data and that provide a range of different characteristics. The countries examined are Australia, Canada, Chile, Denmark, Iceland, Latvia, Mexico, New Zealand, the Slovak Republic and the United States. As shown in Table 5.2, the private pension plans in these countries have different tax treatments and in the case of Australia, Chile, Mexico and New Zealand, government non-tax financial incentives (matching contributions and/or subsidies) also apply. In addition, the systems covered have a different nature (i.e. mandatory vs voluntary, occupational vs personal) and various levels of maturity.
The maturity of a private pension system is not easy to assess, as systems continue to develop over time. In terms of existence, the private pension systems have been in place for more than 50 years in Canada, Denmark, Iceland and the United States, 35 years in Chile, 25 years in Australia, 20 years in the Slovak Republic and Mexico, 15 years in Latvia and 10 years for the KiwiSaver system in New Zealand.
In addition, changes to the systems may affect their maturity:
-
Australia: The mandatory contribution rate to the superannuation system, which was created in 1992, will continue to gradually increase up to 12% in 2025 (from 9% in 2014).
-
Denmark: Occupational pension plans have been in place since the 1950s for certain groups of workers (e.g. civil servants, teachers, nurses). In the 1990s, coverage of occupational plans was greatly increased through collective agreements. However, the contribution rates were initially set at very low levels and gradually increased to reach the full contribution rate of around 15% only in the mid-2000s. Therefore, in 2035 all new retirees will be able to get private pension benefits based on a full career, but only in 2050 will these benefits be based on a full career and a full contribution rate.
-
Iceland: Occupational pension funds have existed since 1921 but became available to the general public in 1969, with mandatory membership from 1974 for wage earners and from 1980 for the self-employed. Full maturity will therefore be reached in 2045. In addition, contribution rates have been raised over time, from 10% of basic salary initially to 10% of total salary in 1989, 12% in 2007 and 15.5% from 2018 onwards.
-
Latvia: The mandatory funded pension system was created in 2001 and has seen several revisions of the contribution rate (6% from 2016 onwards).
-
Mexico: The funded private pension system was created in 1997 for private sector workers and in 2007 for public sector workers.
-
Slovak Republic: The mandatory contribution rate to the individual retirement accounts system, which was created in 2005, is being gradually increased between 2016 and 2024 from 4% to 6%.
-
United States: The share of workers participating in a private pension plan at any point in time has been broadly constant since 1960. However, new types of plan have been introduced over time (e.g. 401(k) plans in 1978, Roth IRAs in the late 1990s, and Roth 401(k) in 2006), leading to a shift from defined benefit to defined contribution plans that may affect contribution levels, and therefore the maturity of the system.12
The structure of the population is expected to change over the next 45 years, with individuals aged 65 and older representing an increasing share of the population aged 20 and older. Figure 5.6 shows the structure of the population aged 20 and older in the 10 countries for 2015, 2035 and 2060. Between 2015 and 2060, the share of individuals aged 65 and older in the population aged 20 and older will increase from around 20% to 30-33% in the developed countries (Australia, Canada, Denmark, Iceland, Latvia, New Zealand and the United States). The increase is steeper for developing countries: from 14% to 32% in Chile, from 11% to 26% in Mexico and from 18% to 42% in the Slovak Republic. Except for Mexico and the Slovak Republic, the largest increase will happen between 2015 and 2035 (by 5 percentage points in Australia, 7 pp in Denmark and Iceland, 8 pp in the United States, 9 pp in Canada, Latvia and New Zealand, and 10 pp in Chile).
Source: OECD Population and Labour Force Projections Database.
The total fiscal cost of financial incentives to promote savings for retirement varies greatly across countries, but remains in the low single digits of GDP. Figure 5.7 shows the total fiscal cost of financial incentives, including the net tax expenditure of tax incentives and its components, and the direct cost of non-tax incentives when they exist, for the 10 selected countries between 2015 and 2060. The fiscal cost varies from 2%-3% of GDP in Australia and Iceland to 0.1%-0.3% of GDP in Chile, Mexico, New Zealand and the Slovak Republic, and it even turns negative in Denmark, indicating an overall positive fiscal effect in the future.
The increase in the total fiscal cost of private pensions in Australia projected for the next 45 years is mainly driven by the rise in the mandatory employer contribution rate.13 The mandatory employer contribution rate is 9.5% of ordinary earnings for financial years up to 30 June 2021, after this, the rate will increase by 0.5 percentage points per year until it reaches 12% from 1 July 2025. The maturing of the Australian private pension system, as more individuals receive the full contribution rate for an increasing portion of their career, creates a rise in the amount of investment income, which is taxed favourably as compared to traditional savings accounts. Figure 5.7 shows that tax revenues forgone on contributions increase until the contribution rate stabilises at 12% in 2025, while tax revenues forgone on accrued income keep increasing between 2020 and 2060. There is no tax collected on withdrawals when assuming that individuals retire at age 60 or older. The net tax expenditure is therefore expected to increase from 2.1% of GDP in 2015 to 2.8% of GDP in 2060. The total fiscal cost of private pensions also includes direct public spending in the form of matching contributions (the super co-contribution and the low income superannuation tax offset).14 This cost is modest and constant over the projected period, at around 0.05% of GDP.
As the Canadian private pension system is already close to maturity, the remaining factor explaining the evolution of the net tax expenditure over time is future demographic developments. The main trend driving the U-shape of the net tax expenditure is the ageing of the baby-boom cohorts. As shown in Figure 5.6, individuals aged between 50 and 59 represent a large share of the population aged 20 and older today. These individuals will start retiring and withdrawing their pension benefits by 2025. Figure 5.7 indicates that tax revenues collected on withdrawals will reach a peak between 2030 and 2035, years during which all of the baby-boomers will be receiving pension benefits and paying personal income tax on them. As a result, the net tax expenditure is expected to decline from 1.5% of GDP in 2015 to 0.9% of GDP in 2035-2040 and to remain stable at around 1.1% of GDP thereafter.
The total fiscal cost of private pensions in Chile is modest. As shown in Figure 5.7, between 2015 and 2060, the net tax expenditure will be around 0.2% of GDP. Indeed, wages are low in Chile, so that most individuals either do not pay income tax or pay tax at the marginal rate of 4%.15 Tax revenues forgone on contributions and accrued income are therefore small. In addition, as most pensioners do not pay taxes because their pensions are below the exempt limit, tax revenues collected on withdrawals are negligible. Direct public spending in the form of matching contributions (for young workers) and fixed nominal subsidies (for mothers) is also modest, at around 0.05% of GDP over the entire period. The total fiscal cost will remain constant between 2015 and 2060 at about 0.24% of GDP.
In Denmark, the net tax expenditure is expected to decrease and turn negative in 2035. The main driver of this trend is the maturing of the system, as more and more people build a pension based on a full career and a full contribution rate. Consequently, tax revenues collected on withdrawals are expected to increase from 1.1% of GDP in 2015 to 3.4% of GDP in 2045. The decline observed in tax revenues collected on withdrawals after 2045 is due to demographic trends, as cohorts entering retirement after 2045 will be smaller than the previous ones. On the other hand, tax revenues forgone on contributions and on accrued income will remain broadly constant over the period 2015-2060, at around 1.9% of GDP and 0.6% of GDP respectively.16 The net tax expenditure will therefore decrease from 1.4% of GDP in 2015 to -0.8% of GDP in 2045, and increase afterwards up to -0.4% of GDP in 2060.
In Iceland, the increase in the contribution rate will create a transition period during which the system will be maturing, so that the net tax expenditure will first increase and then decrease. Between 2016 and 2018 the contribution rate to the mandatory occupational pension system for private sector workers is being raised gradually from 12% to 15.5%.17 The system will therefore be maturing until future retirees draw their pension based on a full career and the full contribution rate. As described in Section 5.2, in a maturing system the net tax expenditure first increases because there is a lag in the growth of benefits behind that of assets and investment income. It then declines when individuals start withdrawing benefits based on a full contribution rate. The net tax expenditure related to mandatory occupational pensions in Iceland is therefore expected to rise from 2.2% of GDP in 2015 to 2.4% of GDP in 2035, before declining to 1.7% of GDP in 2060.
The Latvian system is still maturing, leading to a temporary increase in the net tax expenditure, before declining. The net tax expenditure will peak around 2030, at 0.69% of GDP and will decline thereafter to become negligible in 2060. As the system matures, the assets in the system grow and so do tax revenues forgone on accrued income (from 0.14% of GDP in 2015 to 0.69% of GDP in 2050). In addition, as more individuals receive a pension income based on their full career over time, tax revenues collected on withdrawal increase from close to 0% of GDP in 2015 to just above 1% of GDP in 2055.
In Mexico, the total fiscal cost of mandatory individual retirement accounts is expected to remain stable at around 0.37% of GDP over 2015-2060, with a peak at 0.41% of GDP between 2035 and 2045. This is the result of opposite trends for the net tax expenditure and direct public spending. As the system was created in 1997 for private sector workers and in 2007 for public sector workers, the system is still maturing. As assets continue to grow, tax revenues forgone on accrued income increase from 0.11% of GDP in 2015, until they reach a plateau at around 0.17% of GDP in 2040. As the analysis assumes that pension benefits remain below the tax-exemption limit of 15 times the annual minimum wage, tax revenues collected on withdrawals remain null over the whole period. Overall, the net tax expenditure will therefore increase, from 0.24% of GDP in 2015 to 0.29% of GDP in 2060. In addition, the government makes two types of contribution into individual accounts, one expressed as a percentage of salary (0.225% for private sector workers) and one based on a fixed amount that depends on the number of days worked and the salary level (social quota). The social quota is the most expensive of the two. As the analysis assumes that the maximum level of the social quota increases in line with inflation, the amount spent by the government on that subsidy decreases relative to GDP.18
The declining trend in the total fiscal cost of KiwiSaver schemes in New Zealand is driven by the decrease in direct public spending. The government makes an annual contribution towards KiwiSaver accounts equivalent to 50 cents for every dollar of member contribution annually, up to a maximum payment of NZD 521.43.19 As the maximum amount for the government matching contribution has been divided by two in 2011 and has remained at the same level since then, the analysis assumes that it will stay at NZD 521.43 for the next 45 years. As a percentage of GDP, the expenses linked to the payment of the government matching contribution will therefore decline over time.20 Moreover, the net tax expenditure related to the tax treatment of KiwiSaver schemes is negligible. Indeed, contributions are taxed, but for employees in some income bands, the rate applied to employer contributions is lower than the personal income tax rate, creating tax revenues forgone equivalent to 0.06% of GDP. This is reduced by revenues collected on accrued income, as accrued income is taxed and higher than in the absence of the scheme (because the benchmark tax system would not receive the government matching contribution).
The individual retirement accounts system in the Slovak Republic (pillar 2) is still maturing, leading to a temporary increase in the net tax expenditure. This trend will be reversed due to a drop in the population combined with a change in the age structure of plan participants. As the system was introduced in 2005, the assets have not reached their long-term level yet (cf. Section 5.2). In addition, the mandatory contribution rate is being gradually increased between 2016 and 2024 from 4% to 6%. Tax revenues forgone on contributions are therefore expected to rise from 0.12% of GDP in 2015 to 0.20% of GDP in 2025. Tax revenues forgone on accrued income are also expected to rise from 0.08% of GDP in 2015 to 0.31% of GDP in 2045. However, the Slovakian population aged 20 and older is expected to decline after 2035, from 4.3 million to 3.8 million in 2060. In addition, rules for the participation in the system have changed several times.21 According to the Ministry of Labour, Social Affairs and Family, the distribution of participants today is biased towards younger generations. This is expected to change since current participants will be ageing and mandatory participation has been removed. The net tax expenditure is therefore expected to peak in 2045 at 0.45% of GDP and decline thereafter. There is no tax collected on withdrawals (“EEE” tax system).
The net tax expenditure related to supplementary pension plans in the Slovak Republic (pillar 3) may turn negative in 2020 and go further down thereafter.22 For these plans, contributions are tax deductible, but only up to EUR 180, investment income is tax free during the accumulation period and, upon withdrawal, the part of the assets originated from investment income is taxed.23 As these plans have been introduced in 1996, the system is still maturing. This explains the increase in tax revenues collected on accrued income upon withdrawal, from 0.01% of GDP in 2015 to 0.31% of GDP in 2045. From 2020, tax revenues collected on accrued income will offset tax revenues forgone on contributions (stable at around 0.1% of GDP) and on accrued income during the accumulation phase (increasing from 0.02% of GDP in 2015 to 0.08% of GDP in 2050). Combining both systems, the net tax expenditure related to private pensions in the Slovak Republic will remain positive over the period 2015-2060, increasing from 0.23% of GDP in 2015 to 0.32% of GDP in 2030 and then declining to 0.21% in 2060.
Finally, in the United States, the net tax expenditure related to private pensions is expected to decline from 1.0% of GDP in 2015 to 0.7% of GDP in 2040 and then to remain broadly constant. There are two potential driving factors. The first one is the ageing of the baby-boom cohorts born between 1946 and 1964. Tax revenues collected on withdrawals are expected to increase as these cohorts enter retirement, from 0.77% of GDP in 2015 to 0.92% of GDP in 2025.24 The decline in tax revenues collected on withdrawals thereafter (to 0.74% of GDP in 2060) stems from the fact that the baby-boom cohorts gradually pass away and are replaced by smaller cohorts of retirees. The ageing of the baby-boom cohorts may not explain however the continuous declining trend in tax revenues forgone on accrued income over the entire period (from 0.8% of GDP in 2015 to 0.6% of GDP in 2060).25 Another explanatory factor could be the shift from defined benefit to defined contribution pension plans (OECD, 2016[9]). This shift may have reduced overall contribution levels, in turn reducing assets accumulated by younger generations.
5.4. Potential impact of incorporating corporate income tax and new savings effects into the calculation of the net tax expenditure
The calculations presented earlier only take into account the impact of tax incentives for retirement savings on personal income tax revenues. However, the preferential tax treatment of private pension plans may have an impact on other government revenues, either directly or indirectly.
This section complements the previous analysis by incorporating corporate income tax revenues and new savings effects into the calculation of the net tax expenditure. It first presents the potential effects of financial incentives on general government revenues other than personal income tax. It then adjusts the earlier model of the net tax expenditure to account for corporate income tax and new savings effects. This section, however, does not update the net tax expenditure calculations of Section 5.3 for the selected OECD countries, as this would require too many assumptions.
Effects of financial incentives to promote savings for retirement on other general government revenues
The tax treatment of retirement savings may affect the general government budget through its impact on different sources of revenues. As discussed in the previous sections, the tax treatment of retirement savings has a direct impact on personal income tax revenues. Other government revenues may be affected as well.
First, the tax treatment of retirement savings has a direct impact on corporate income tax revenues, provided that some of the savings are invested in domestic equities or corporate bonds. Feldstein (1995[10]) argued that traditional estimates of the net tax expenditure related to tax-favoured private pension plans systematically over-estimate the true cost because they fail to account for corporate income tax revenues. Looking at the effects of tax incentives for Individual Retirement Arrangements (IRAs) on government revenue and national savings in the United States, he claims that IRAs increase personal savings and that these additional savings are divided between the corporate sector capital stock, the non-corporate sector capital stock and net foreign investment. Some of the increase in national savings therefore raises the corporate capital stock. This in turn increases the productive capacity of companies and thereby the overall level of profits, which are subject to corporate income tax. He finds that tax incentives may lead to a revenue increase for the government in the medium term. Ruggeri and Fougère (1997[11]) agreed but questioned some of the assumptions used in Feldstein (1995[10]). They concluded that tax-favoured savings plans reduce tax revenues but that the magnitude of the effect may not be large once corporate income tax is taken into account.
Second, the tax treatment of retirement savings may also affect consumption tax revenues. This potential impact applies only if some of the savings originate from money that the individual would not have saved in the absence of the tax-favoured private pension plan (i.e. new savings). New savings are indeed financed by a reduction in consumption, therefore lowering consumption tax revenues during contribution years. In addition, as new savings would not have accumulated in the absence of tax-favoured schemes, the pension benefits withdrawn from the tax-favoured plan will be higher than those withdrawn from the tax benchmark, therefore increasing consumption tax revenues during the retirement period.
Third, there are other indirect effects on tax revenues that may also be triggered by the tax treatment of retirement savings. New savings provide a source of funds available for domestic investment, which in turn is a key driver of labour productivity, higher future living standards, economic growth and wage growth.26 Therefore, new savings may lead to higher future personal income tax revenues collected on wages and higher overall taxation from higher economic growth. In addition, tax incentives may encourage people to save for retirement, therefore relieving the government of the payment of selected social benefits. For example, in countries with means-tested public pensions, if private pensions turn out to be high, a lower proportion of the population may fall under the means-tested threshold and qualify for the public pension, therefore decreasing the cost for the government.
Finally, how private pension plans are treated when calculating social contributions also has a direct impact on the social security budget. As shown in Chapter 2, contributions to private pension plans are sometimes exempt from social contributions or enjoy a reduced rate for social contributions, for example employer contributions to occupational pension plans in Austria and Belgium.27 By contrast, contributions to the “TTE” tax benchmark would typically be subject to full social contributions. The social treatment of private pensions may therefore lead to revenues forgone in social security funds.
The following analysis focuses on the direct effects of the tax treatment of retirement savings on tax revenues. It therefore accounts for the direct impact on personal income tax, corporate income tax and consumption tax revenues. It abstracts from indirect effects at the macro-economic level that are more difficult to measure as it may involve a large degree of guesswork to come up with appropriate assumptions on productivity, wages and income. The analysis also abstracts from the treatment of private pensions when calculating social contributions.28
Incorporating corporate income tax and new savings into the net tax expenditure calculations
The total net tax expenditure (TNTE) calculated in this section can be split into three components, depending on the tax source: personal income tax (PIT), corporate income tax (CIT) and consumption tax (VAT)
Equation 7
The net tax expenditure related to private pension plans where both contributions and returns on investments are exempted from taxation while benefits are treated as taxable income upon withdrawal (“EET” tax regime), with respect to a “TTE” benchmark and when only taking into account personal income tax, PITt, corresponds to Equation 1 in Section 5.2.
The net tax expenditure related to corporate income tax is negative as it is a revenue increase. Corporate income tax is levied on company revenues when assets are invested in domestic equity.29 The rate of return earned on pension contributions is therefore net of corporate income tax, i.e. i = p × (1 – µCIT) × r + (1 – p) × δ where µCIT is the average marginal corporate income tax rate, p is the proportion of assets invested in domestic equity, r is the pre-tax rate of return on domestic equities and δ is the rate of return on other asset classes.30 In the absence of tax-favoured pension plans (i.e. for the “TTE” tax benchmark), corporate income tax would be collected on the returns on investment from diverted savings. The additional corporate income tax revenues collected on the returns on investment from the total assets invested in the pension plan comes from non-diverted savings (i.e. the tax break and the new savings) and equals: µCIT × p × r × (At-1 – A't-1). The net tax expenditure related to corporate income tax is the inverse of that expression.
Equation 8
Finally, new savings lower consumption tax revenues during the accumulation period and increase consumption tax revenues during the retirement period. Considering µVAT as the tax rate on consumption, the net tax expenditure related to consumption tax may be expressed as
Equation 9
The total net tax expenditure is therefore expressed as w
Equation 10
Total net tax expenditure
This section adjusts the simple and straightforward models of the net tax expenditure to account for the effect of corporate income tax and new savings. It assumes a pension system introduced in year t and maturing over time, using the same assumptions as in Section 5.2. In addition, the model assumes that the tax rates on personal income, corporate income and consumption are all equal to 30%, new savings represent 25% of all personal savings and the assets are fully invested in domestic equity.
Figure 5.8 displays how the three components of the total net tax expenditure (i.e. personal income tax revenues, corporate income tax revenues and consumption tax revenues) evolve over time when a pension system with an “EET” tax regime matures. As seen previously, the net tax expenditure related to personal income tax rises when contributions to the new pension system begin and then declines and converges to a steady state (cf. “PIT” line in Figure 5.8). The lag in the growth of benefits behind that of assets and investment income is what creates the temporary increase in the net tax expenditure.
Note: “PIT” means personal income tax, “CIT” means corporate income tax and “VAT” means consumption tax.
Assuming that part of the savings are new savings reduces the net tax expenditure, as compared with no new savings. For example, the steady-state net tax expenditure is equal to 0.66% of GDP in the absence of new savings (cf. Figure 5.2), and to 0.33% of GDP when assuming that new savings represent 25% of all personal savings (cf. “PIT” line in Figure 5.8). Indeed, the amount forgone on personal income tax revenues on accrued income is lower when part of the savings are new savings. This is because new savings would not have generated investment income in the absence of the tax-favoured pension plan. By contrast, new savings neither affect revenues forgone on contributions nor revenues collected on benefits.
Under the assumptions used here, the net tax expenditure related to corporate income tax is always negative. Non-diverted savings (i.e. the tax break and the new savings) generate additional accrued income (as compared to the tax benchmark) which is subject to corporate income tax. Corporate income tax revenues therefore increase in line with the development of the assets and investment income generated by these non-diverted savings, until reaching a steady state. In addition, corporate income tax revenues collected on accrued income are higher in the case of new savings (as opposed to no new savings) as non-diverted savings are then equal to the tax break plus the new savings. Higher non-diverted savings generate higher investment income, which is subject to corporate income tax.
Finally, the net tax expenditure related to consumption tax changes sign as the pension system matures. New savings in tax-favoured pension plans first generate consumption tax revenues forgone as new savings are financed by a decline in consumption. However, consumption tax revenues become higher than in the absence of new savings during retirement as benefits received (net of personal income tax) are larger than in the absence of the tax-favoured pension plan. This implies that new savings allow the Treasury to collect more consumption tax revenues (as compared to the tax benchmark) once the volume of benefits paid to retirees is high enough to offset the volume of new savings done by workers.
Combining the three components, the total net tax expenditure is higher at the onset of the system and declines to its steady state over time (Figure 5.9). The system attains full maturity and its steady-state level only in year t+64, the first year in which the 84-year-olds, who were aged 20 when the system was introduced, have contributed for their full career.
Under the assumptions used in this analysis, the total net tax expenditure reaches a negative steady state level of -1.08% of GDP (plain line in Figure 5.9). It becomes negative 33 years after the introduction of the pension system. However, whether the steady state is negative and the year in which the total net tax expenditure becomes negative is sensitive to the assumptions. For example, the steady-state level is negative at -0.16% of GDP in the case of no new savings (“long dash and dot” line in Figure 5.9).31 Alternatively, assuming that only 50% of the assets are invested in domestic equity (instead of 100%) gives a negative steady-state level of -0.32% of GDP.
5.5. Conclusion
This chapter has examined the fiscal cost of tax and non-tax financial incentives for private pension plans and its evolution over the next 45 years in selected OECD countries. The main conclusions are the following:
-
The tax treatment of private pension plans represents a significant fiscal cost in some countries but remains in the low single digits of GDP. This cost is mostly concentrated on high-income earners as these individuals participate more in tax-favoured retirement savings plans and pay the largest share of total taxes.
-
Introducing a pension system with an “EET” tax treatment for retirement savings creates a larger net tax expenditure that comes upfront. It is only several decades later that the new pension system and its net tax expenditure reach a stationary situation.
-
The maturing of the pension system and the ageing of larger cohorts can create substantial declines in the net tax expenditure level over time, but only from a level that is temporarily above a steady-state value that results from the introduction of the pension system or from the working years of larger cohorts.
-
Countries should always expect a positive net tax expenditure from a tax treatment of retirement savings following an “EET” regime, including in the long-term. Even the maturing of the pension system and the ageing of larger cohorts may not be sufficient to reduce the net tax expenditure to zero or below zero, except when the amount of investment income generated in the year by personal savings is lower than the excess of benefits over contributions.
-
Most of the private pension systems analysed in this document are still maturing, due to the recent introduction of the system (Latvia, Mexico and the Slovak Republic), changes in contribution rates (Australia, Iceland, Latvia and the Slovak Republic) or past changes in coverage and contribution rates (Denmark). Depending on how far back these changes have taken place, the net tax expenditure may be in an increasing or decreasing trend. The ageing of the baby-boom cohorts is the other main driving factor for changes in the net tax expenditure (Canada, Denmark and the United States).
-
Direct spending in the form of matching contributions or fixed nominal subsidies is modest compared to the net tax expenditure in Australia and Chile, while it represents an important component of the total fiscal cost related to private pensions in Mexico and New Zealand.
-
Accounting for corporate income tax revenues and the effects produced by new savings would reduce the net tax expenditure arising from the tax treatment of private pensions.
References
[13] Austrian Federal Ministry of Finance (2016), Subsidy Report 2016, https://www.bmf.gv.at/budget/das-budget/Foerderungsbericht_2016.pdf?69voz6.
[15] Belgian House of Representaives (2018), Inventory of federal tax expenditures, https://finance.belgium.be/en/figures_and_analysis/figures/federal_tax_expenditures_report.
[5] Collins, M. and G. Hughes (2017), “Supporting Pension Contributions Through the Tax System: Outcomes, Costs and Examining Reform”, The Economic and Social Review, Vol. 48/4, pp. 489-514.
[12] Congressional Budget Office (2016), “The distribution of household income and Federal taxes, 2013”, Congress of the United States, Congressional Budget Office, Vol. Publication No. 51361.
[7] Congressional Budget Office (2013), “The distribution of major tax expenditures in the individual income tax system”, Congress of the United States, Congressional Budget Office, Vol. Publication No. 4308.
[8] Corneo, G., C. Schröder and J. König (2015), “Distributional effects of subsidizing retirement savings accounts: Evidence from Germany”, Discussion Paper, School of Business & Economics: Economics, Vol. No. 2015/18.
[16] Department of Finance Canada (2018), Report on Federal Tax Expenditures: Concepts, Estimates and Evaluations, https://www.fin.gc.ca/taxexp-depfisc/2018/taxexp18-eng.asp.
[10] Feldstein, M. (1995), “The effects of tax-based saving incentives on government revenue and national saving”, The Quarterly Journal of Economics, Vol. 110/2, pp. 475-494.
[17] French Ministry of Economy and Finance (2017), Evaluations of ways and means: tax expenditures 2018, https://www.performance-publique.budget.gouv.fr/actualites/2017/publication-evaluations-voies-moyens-tomes-1-2-projet-loi-finances-2018#.W08RZrpuJCo.
[18] German Federal Ministry of Finance (2017), 26th Subsidy Report of the Federal Government, https://www.bundesfinanzministerium.de/Content/DE/Downloads/Broschueren_Bestellservice/2018-08-23-subventionsbericht-26.pdf?__blob=publicationFile&v=2.
[23] Government Offices of Sweden (2018), Recognition of Tax Expenditures in 2018, https://www.regeringen.se/rattsliga-dokument/skrivelse/2018/04/skr.-20171898/.
[25] HM Revenue & Customs (2018), Registered pension schemes: cost of tax relief, https://www.gov.uk/government/statistics/registered-pension-schemes-cost-of-tax-relief.
[19] Irish Tax and Customs (2018), Costs of tax expenditures (credits, allowances and reliefs), https://www.revenue.ie/en/corporate/information-about-revenue/statistics/tax-expenditures/costs-expenditures.aspx.
[20] Italian Ministry of Economy and Finance (2011), Working Group on tax erosion: Final report, http://www.mef.gov.it/primo-piano/documenti/20111229/Relazione_finaledel_gruppo_di_lavoro_sullxerosione_fiscale.pdf.
[3] Joint Committee on Taxation (2018), “Estimates of Federal Tax Expenditures for Fiscal Years 2017-2021” JCX-34-18, https://www.jct.gov/publications.html?func=startdown&id=5095.
[4] Macdonald, D. (2016), Out of the Shadows: Shining a light on Canada's unequal distribution of federal tax expenditures.
[21] Mexican Secretariat of Finance and Public Credit (2018), Tax Expenditure Budget 2018, https://www.gob.mx/shcp/documentos/presupuesto-de-gastos-fiscales-2018.
[9] OECD (2016), “The changing pensions landscape: The growing importance of pension arrangements in which assets back pension benefits”, in OECD Pensions Outlook 2016, OECD Publishing, Paris, https://doi.org/10.1787/pens_outlook-2016-4-en.
[1] OECD (2010), Choosing a Broad Base - Low Rate Approach to Taxation, OECD Tax Policy Studies, No. 19, OECD Publishing, Paris, https://doi.org/10.1787/9789264091320-en.
[2] OECD (1996), Tax Expenditures: Recent Experiences, OECD Publishing, Paris.
[6] Pension Policy Institute (2016), Pension Facts.
[11] Ruggeri, G. and M. Fougère (1997), “The effect of tax-based savings incentives on government revenue”, Fiscal Studies, Vol. 18/2, pp. 143-159.
[24] Swiss Federal Department of Finance (2011), What are the tax breaks granted by the Confederation?, http://www.artias.ch/wp-content/uploads/2012/02/all%C3%A9gements-fiscaux-rapport-2011.pdf.
[14] The Australian Government, the Treasury (2018), Tax Expenditure Statement 2017, https://treasury.gov.au/publication/2017-tax-expenditures-statement/.
[22] The New Zealand Government, the Treasury (2018), 2018 Tax Expenditure Statement, https://treasury.govt.nz/publications/tax-expenditure/2018-tax-expenditure-statement.
[26] U.S. Office of Management and Budget (2017), Analytical Perspectives, Budget of the U.S. Government, Fiscal Year 2018, https://www.whitehouse.gov/sites/whitehouse.gov/files/omb/budget/fy2018/spec.pdf.
This annex presents the approach used to project the future profile of the cost of tax incentives for private pensions, taking into account current and future contributions, asset accumulation and withdrawals, all of which will be strongly influenced by future demographic developments.
Methodological framework
Generating estimates of future costs and benefits arising from the tax treatment of retirement savings plans requires projecting forward a number of key variables including the number of contributors, total contributions, total assets, accrued income from assets, number of beneficiaries and withdrawals.32 In each case, the total figure is obtained from aggregation across 13 heterogeneous 5-year age groups from ages 20 to 80+. They are calculated for each of the ten 5-year periods from 2015 to 2060, given initial conditions in 2015.33 The projections also require estimates of relevant tax rates associated with each component of the net tax expenditure, as identified in Section 5.2.
Net tax expenditures at each period t are obtained as the net sum over all age groups g of the revenues forgone on contributions (RFC), revenues forgone on accrued income (RFAI) and revenues collected on withdrawals (RCW)
Equation 11
Revenues collected on withdrawals (RCW) are the product of the age-specific marginal income tax rates on withdrawals (µB,g) and the amount withdrawn by each age group (Bt,g)
Equation 12
In order to project forward withdrawals/benefits levels, it is necessary to project the total amount of assets, as benefits generally depend on total assets accumulated in private pension plans at the time of retirement.34 Assets for each age group (At,g) are determined according to the nominal pre-tax rate of return i on previous period assets, new contributions (Ct,g) and withdrawals (Bt,g)
Equation 13
Withdrawals are modelled based on the assumption that the total amount of assets accumulated until the age of 65 (or 60, depending on the county) is run down according to a constant annuity formula until full exhaustion at the age of 85.
A85 = 0, P84 = A84 and given that Pt = P, Equation 14
where P is the annuity payment.
As contributions can generally be deducted from taxable income, revenues forgone on contributions (RFC) made by each age group are the product of the age-specific marginal income tax rate on contributions (µc,g) and the total amount contributed in age-group g, (Ct,g)
Equation 15
In order to project forward the level of contributions, there is a need to project the number of contributors and the average contribution level per contributor. In turn, the number of contributors per age group in a given time period is determined by the number of individuals in employment in the age group multiplied by the participation rate in private pension plans. In addition, the average contribution level per participant in each age group can be expressed as a ratio of the economy-wide average wage. All in all, total contributions per age group in a given time period can therefore be expressed as
Equation 16
where E is the number of individuals in employment, L is the number of contributors, and W stands for nominal wages.
The ratio of average contribution level per participant in each age group to economy-wide average wage (first term in Equation 16) and the age-specific rates of participation in private pension plans (third term) are assumed to remain at their current level in the future. The average nominal wage in the total economy (second term) is assumed to grow at a constant rate. Finally, employment projections (fourth term) are based on labour force projections from the OECD Population and Labour Force Projections Database, combined with assumptions regarding the future evolution of unemployment rates.35
Forgone tax revenues on accrued income from investment (RFAI) measures taxes that would have been collected on investment income if private savings had been invested in a benchmark savings vehicle. It thus depends on the age-specific tax rates on accrued income from alternative savings (µA,g), the nominal rate of return on assets (i), and the amount of assets accumulated ( ).
In contrast to the calculation of revenues collected on withdrawals, the relevant stock of assets in this case is not the total assets invested in the scheme (At-1,g) but only those accumulated from diverted savings. The reason for including only a subcomponent of total assets in the calculation of revenue losses on investment income is because not all contributions to retirement saving plans would have generated investment income in the absence of the retirement scheme. Contributions indeed comprise the tax break (forgone tax revenues on contributions), µCCt , and personal saving, (1-µC)Ct. The latter can in turn be split into diverted savings, (1-α)(1-µC)Ct, and new savings, α(1-µC)Ct, where α is the share of personal savings financed by a decline in consumption (new savings). Since neither the new savings nor the tax break components would have generated investment income in absence of the retirement scheme, they need to be excluded from the calculation of forgone tax revenues. Hence, the model requires a second asset accumulation equation based on diverted savings alone to determine forgone tax revenues on accrued income from investments
Equation 17
Withdrawals B't,g (which also differ from withdrawals used in Equation 13) are calculated as a constant annuity on accumulated assets from diverted savings, A’, at the age of 65 and fully exhausted by the age of 85. Forgone revenues on accrued investment income are thus determined by the age-specific tax rates on investment income multiplied by the return on assets accumulated from diverted savings
36 Equation 18
Substituting Equation 12, Equation 15 and Equation 18 into Equation 11 leads to the following relation for net tax expenditures
Equation 19
For those countries that tax accrued income in retirement saving plans (e.g. Denmark and Sweden), albeit at a usually favourable rate, , Equation 20 includes one extra term which captures tax revenues collected on the return to total assets accumulated
Equation 20
Key parameters
The relevant tax rates used to estimate revenues forgone on contributions and accrued investment income, as well as revenues collected on withdrawals are assumed to remain constant during the whole simulation. They are calculated based on a number of assumptions.
First, age-specific marginal tax rates are determined according to the age-specific average income in 2015. Average income by age group comes from the OECD Employment database. The average income is compared to the country-specific income tax brackets in 2015 to choose the appropriate marginal tax rate. In all countries where contributions to private pension plans are deductible from taxable income, these effective marginal tax rates measure the tax revenue forgone on a unit of contribution.
Second, the current tax treatment of traditional savings accounts in each country is taken as the benchmark tax system. In all cases, this is some version of the comprehensive income tax regime (“TTE”). Detailed information on the tax treatment of these traditional savings accounts is used to derive tax rates on the return on investment.
Third, given the lack of sufficient information about the overall income of private pension beneficiaries, the general rule has been to set the tax rate applied on benefit withdrawals from private pensions equal across all retired age groups (µB,g = µB ) at 5 percentage points below the marginal tax rate of the average earner. A proper calculation would require adequate information about the level and the various sources of taxable income of pensioners who have participated in a private pension scheme. These can be quite different from the average level and sources of taxable income of all pensioners. On the one hand, if benefits from private pension schemes were the sole source of taxable income, then the appropriate rate applied to measure revenues collected on withdrawals would be the average tax rate corresponding to the value of the annual benefit withdrawn. In practice, however, most recipients of private pension benefits usually receive income from various other taxable sources, in which case applying the average tax rate corresponding to the level of private pension benefits would most certainly underestimate the amount of tax revenues recovered. On the other hand, applying the corresponding marginal tax rate would most likely lead to overestimate the tax revenues, if only because pensioners generally benefit from special tax rebates and other benefits. On balance, a 5-percentage point reduction probably represents in most countries a conservative spread, in particular in those with steep tax schedule.37
Finally, the pre-tax nominal rate of return on assets is set at 5.06% per annum, including 2% inflation. Annex Table 5.A.1 presents the broad set of key assumptions made to generate the base case projections.
Country-specific assumptions
The following assumptions apply to the countries covered by the analysis.
Australia
The analysis covers both concessional and non-concessional contributions to the superannuation system.
The maximum amount for the low-income superannuation tax offset and for the super co-contribution (AUD 500 each) is assumed to remain constant going forward.38
As pension benefits by age are not available for 2015, the analysis assumes that the minimum age of withdrawal is 60 and that the distribution of benefits by age is the same as the distribution of individuals by age in the total population.
The calculations do not include the impact of the 2016-17 Budget measures and subsequent amendments. According to the Australian Treasury, these measures are expected to have a net effect of reducing the tax expenditure estimate.
Canada
The analysis covers Registered Pension Plans (RPPs), Registered Retirement Savings Plans (RRSPs), Pooled Registered Pension Plans (PRPPs) and Registered Retirement Income Funds (RRIFs).
Chile
The analysis covers the mandatory pension system only. Voluntary contributions are not taken into account.
The minimum wage is assumed to be indexed to wages, so that the maximum amount for the matching contribution for young workers and the subsidy for mothers also grow in line with wages.
Denmark
The analysis covers all pension plans for which an “EtT” tax regime applies (i.e. ATP and quasi-mandatory occupational plans). It therefore excludes “age savings” plans, for which a “TtE” tax regime applies.
Iceland
The analysis covers mandatory occupational pension plans only (pillar 2). Voluntary personal pension plans (pillar 3) are not taken into account as the distribution of assets by age is not available.
Data on total contributions and total assets by age are not available. Total assets were approximated using accrued liabilities for a sample of pension funds. Total contributions for 2015 were approximated assuming a weighted average total contribution rate of 12.24% and using the distribution of employee contributions by age.
All benefits paid are assumed to be retirement income (i.e. no disability and survivor’s pensions).
Latvia
The analysis covers the mandatory funded pension scheme only. Voluntary private pension funds are not taken into account as the distribution of contributions by age is not available.
The distribution of contributors by age is not available, but the distribution of participants by age is. The analysis assumes that the proportion of deferred members (those not paying pension contributions) is constant across age groups (21.8% in 2015).
Benefits paid by age are not available. The analysis assumes that the money transferred at retirement from the funded scheme to the pay-as-you-go scheme is instead run down according to a constant annuity formula until full exhaustion at the age of 85.
Mexico
The analysis covers mandatory individual retirement accounts and some voluntary pension savings (i.e. complementary contributions, short-term voluntary contributions, long-term voluntary contributions and special "saving for retirement" accounts). It excludes solidarity savings, as well as occupational and personal pension plans, for which detailed data broken down by age are not available for all variables. The results for voluntary pension savings are not shown but can be provided upon request.
The maximum amount for the social quota is assumed to be inflation indexed.
Returns on complementary contributions and short-term voluntary contributions are assumed to be taxed yearly instead of upon withdrawals.
The analysis assumes that withdrawals occur from the age of 60 years old.
The analysis assumes that the benefits received from mandatory individual retirement accounts are not taxed due to the exemption limit of 15 times the annual minimum wage. This exemption applies to the sum of all pension payments or benefits paid by the federal government, by pension funds, by occupational pension plans and by personal pension plans. However, only a minority of workers receive pension payments from different schemes and are likely to receive a total pension income that exceeds the tax-exemption limit. This assumption may therefore underestimate tax revenues collected on withdrawals and overestimate the net tax expenditure.
The assumed productivity growth rate of 1.25%, which is common for all 10 countries, is high in the context of Mexico (0.76% in recent years).
New Zealand
The analysis covers the KiwiSaver system only. Other superannuation plans are not taken into account.
Data on total assets by age are not available. The analysis approximates the distribution of total assets by age in 2015 by taking into account the fact that KiwiSaver started in 2007 and assuming a contribution rate of 7.5% (3% from the employee, 3% from the employer and 1.5% from the state – “member tax credit”).
As the maximum amount for the government matching contribution has been divided by two in 2011 and has remained at the same level since then, the analysis assumes that it will stay at NZD 521.43 for the next 45 years.
The analysis assumes that the assets are withdrawn as a lump sum at age 65.
Slovak Republic
The analysis covers the individual retirement accounts system (pillar 2) and supplementary pension plans (pillar 3).
Participation in individual retirement accounts by age is assumed to be dynamic rather than mirroring the distribution in 2015. The model assumes that the participation rate in 2060 will be around 40% of the labour force.
The tax deductibility of employee voluntary contributions in their individual retirement account was temporary, from 1 January 2013 until 31 December 2016. The analysis assumes that this measure will continue in the future. The impact on the results is however very negligible as the proportion of voluntary contributions in total pillar 2 contributions was 0.26% in 2015 and 0.29% in 2016.
The analysis assumes that the tax-deductibility limit for employee contributions into supplementary pension plans (EUR 180) is not indexed.
Contributions are not tax deductible for individuals who opened a supplementary pension plan before 2014. These individuals represented 85% of all participants in 2015 and their share is assumed to decline, such that by 2060 all participants have a plan opened from 2014.
Returns on investment are taxed upon withdrawal for supplementary pension plans. The analysis therefore calculates tax revenues forgone on accrued income during the accumulation phase and tax revenues collected on accrued income just before getting pension benefits (using an approximation of the share of accrued income in total assets at age 65).
United States
The analysis covers voluntary occupational and personal plans. It only takes into account the net tax expenditure for the federal government (income taxation at state levels is not included).
The breakdown of the different relevant variables by age is approximated using the 2013 Survey of Consumer Finances. The analysis applies the distribution by age to the aggregated numbers from the OECD Global Pension Statistics database.
Notes
← 1. This means that the counterfactual scenario assumes that individuals continue saving the same amount in a non-tax favoured benchmark vehicle.
← 2. The calculations for Canada have been undertaken by the Canadian Centre for Policy Alternatives and are not supported by the Canadian government. They only reflect tax expenditure for private pension plans and do not encompass the whole government tax-related support for retirement savings, in particular tax assistance on Canada Pension Plan (CPP) savings.
← 3. The net tax expenditure is calculated with respect to a “TTE” benchmark. This benchmark may differ from the one used in national tax expenditure reports.
← 4. The net tax expenditure adds to direct spending associated with non-tax financial incentives when relevant in order to estimate a total fiscal cost.
← 5. µC and µA represent the tax rates that apply respectively to contributions and returns in the benchmark tax regime, and that would apply to contributions and returns in the tax-favoured pension plan, would this plan stopped being tax-favoured. By contrast, µB represents the actual tax rate that applies to benefits paid by the tax-favoured pension plan.
← 6. In steady state, the value of k depends on contributions as a share of GDP (Ct/Yt), i, g, and the duration of the contribution and pay-out periods.
← 7. The mechanisms are the same when assuming that the contribution rate increases.
← 8. As the model assumes that contributions are made at year-end, there is no investment income and no pension benefits in year t.
← 9. It also assumes constant demographics.
← 10. The steady-state level corresponding to the “EET” tax regime entails a long-term fiscal cost.
← 11. Annex 5.A presents the methodology as well as country-specific assumptions used in the calculations.
← 12. In addition, existing and proposed policy efforts are aimed at increasing participation in private pension plans in the future. These include automatic enrolment in defined contribution plans and state initiatives to boost retirement savings.
← 13. The calculations do not include the impact of the 2016-17 Budget measures and subsequent amendments. According to the Australian Treasury, these measures are expected to have a net effect of reducing the tax expenditure estimate.
← 14. The low income superannuation tax offset replaced the low income superannuation contribution as of 1 July 2017 and kept the same characteristics.
← 15. Based on data from the Pension Superintendence of Chile on average wage by age group, individuals aged between 35 and 69 would face a marginal income tax rate of 4%, while younger and older individuals would be tax exempt.
← 16. Investment income is taxed favourably as compared to traditional savings accounts (fixed rate of 15%).
← 17. It is already 15.5% for public sector employees.
← 18. The net tax expenditure related to voluntary pension savings (i.e. complementary contributions, short-term voluntary contributions, long-term voluntary contributions and special “saving for retirement” accounts) is negligible, below 1 basis point of GDP. The results are not shown but can be provided upon request.
← 19. The government used to kick-start every KiwiSaver accounts with a tax-free contribution of NZD 1 000. New members who joined on or after 21 May 2015 are not eligible for this payment. This contribution is therefore not taken into account.
← 20. As explained in the annex, the model assumes that GDP grows in line with employed population growth and nominal productivity growth (1.25% real plus 2% inflation).
← 21. Participation was mandatory for new labour market entrants between 2005 and 2007. Then participation became voluntary until April 2012. From April 2012 until the end of 2012, new labour market entrants were automatically enrolled with an option to opt-out within 2 years. Participation became voluntary again since 2013, but once the individual has joined the system (before age 35), s/he cannot opt out.
← 22. The calculations do not include the fact that employer contributions are deductible from corporate income tax (up to 6% of the employee’s salary). The fiscal cost of exempting employer contributions is higher than the one related to the partial exemption of employee contributions (EUR 14 million vs. EU 1.3 million in 2015).
← 23. Contributions are not tax deductible for individuals who opened a supplementary pension plan before 2014. The tax deductibility amount for those who opened their plan since 2014 is not indexed.
← 24. The first individuals from these cohorts reached retirement age in 2011, while the last ones will enter retirement in around 2029.
← 25. As seen in Section 5.2 (Figure 5.5), accrued income increases as larger cohorts build pension assets.
← 26. Admittedly, lower consumption may also reduce economic growth and company profits.
← 27. These social contributions are usually levied on gross salaries and wages to finance among others, health care insurance, unemployment insurance, public pensions and disability pensions.
← 28. Although this effect is not explicitly taken into account in this analysis, it could be easily added by increasing the average marginal personal income tax rate applicable to contributions (µC) by the appropriate social contribution rate.
← 29. This applies also when assets are invested in corporate bonds. For simplicity of presentation, the analysis focuses on domestic equity.
← 30. For simplification, the analysis further assumes that δ = r, so that i = r × (1 – µCIT × p).
← 31. As a way of comparison, the model in Section 5.2 (which also assumes no new savings, but only accounts for personal income tax) reaches a steady-state net tax expenditure of 0.66% of GDP. Accounting for corporate income tax therefore reduces significantly the steady-state level.
← 32. The calculations use the number of contributors rather than the number of active members in private pension plans to project future contribution levels.
← 33. Unlike the models in Section 5.2, the simulation is done with a five-year pace, rather than one-year. This is because demographic projections are only available with a five-year pace.
← 34. In the case of defined benefit pension plans, the analysis assumes full-funding (i.e. total assets are equal to total liabilities). As an approximation, benefits can also be considered to depend on assets accumulated at retirement.
← 35. Unemployment rates are assumed to converge to their long-term equilibrium level (non-accelerating inflation rate of unemployment), as published in the OECD Economic Outlook.
← 36. Diverted accrued income is equal to the proportion of diverted savings multiplied by the total accrued income (i.e. ).
← 37. Two factors could contribute to a wider spread. One is the possibility in many countries to withdraw benefits in the form of a lump sum, which in some cases is treated more leniently from a tax standpoint. The other factor, which is more relevant for European Union countries, concerns the possibility for pensioners to migrate to a lower tax country. On the other hand, some pensioners could face very high marginal rates if they lose means-tested benefits as they withdraw pension benefits.
← 38. The low income superannuation tax offset replaces the low income superannuation contribution as of 1 July 2017 and keeps the same characteristics.