Chapter 3. Methodological approach to tax and skills statistics1

This chapter outlines an approach to estimating the financial incentives for individuals and governments to invest in individuals’ skills, and the effect of the tax system on these incentives. Specifically, it outlines the key indicators developed to examine the impact of the tax system on skills. These indicators include Marginal Effective Tax Rates (METRs) and Average Effective Tax Rates (AETRs) on Human Capital Investment, as well as Marginal Returns to Costs Ratios (MRCRs) and Average Returns to Costs Ratios (ARCRs) of Government Investment in Human Capital. The chapter cites the data sources used to develop results for these indicators, and explains how the results presented in Chapters 4 and 5 of the study should be interpreted.

3.1. Introduction

This chapter outlines the methodology used in this study to estimate the key indicators of the financial incentives surrounding skills investments. Specifically, it outlines the approach used to estimate the earnings necessary to breakeven on a skills investment (the Breakeven Earnings Increment or BEI). It also outlines the effects of the tax system on the incentives individuals face to invest in skills, both for a marginal student (the marginal effective tax rate or METR on skills), and for an average student (the average effective tax rate or AETR on skills).

The chapter also discusses the indicators developed that analyse the government’s net financial returns from investing in skills of students. Two indicators are developed from the government’s perspective, analogous to those for individuals. Specifically, an indicator is developed for a marginal student (the marginal returns to costs ratio or MRCR) and another for an average student (the average returns to costs ratio ARCR).

The study draws on the approaches used in the literature on the taxation of physical capital, in particular the succession of models by King and Fullerton (1984), Devereux and Griffith (1998), Devereux (2003) and Klemm (2008), which outline effective tax rates on physical capital investments in a net present value (NPV) framework. Marginal Tax Rates similar to those derived in Devereux and Griffith (1998) are provided, and equivalent AETRs as derived in Klemm (2008), but for human instead of physical capital.

All of these indicators have been built within the OECD’s Taxing Wages country models (OECD, 2016). This chapter, and the accompanying Annex A to this study, outline the methodology behind these additions to the Taxing Wages models. The chapter also discusses the data sources used, some of the assumptions made in the analysis, as well as some important caveats that need to be considered when interpreting the results.

The chapter proceeds as follows. Section 3.2 outlines the various cost components of education, and the data sources used to derive them. Section 3.3 discusses the costs that may be incurred when students must borrow to finance their skills investment. Section 3.4 outlines the various components of the necessary breakeven earnings level, and how it differs from the BEI. Section 3.5 outlines how the BEI differs in the presence and absence of taxes, and how it is used to calculate the METR on skills within the country models. Section 3.6 outlines how assumed post-education earnings levels are used to calculate the AETR on skills. Section 3.7 outlines some differences between the METR and AETR, and how they are both a function of two key tax rates: the tax rate on foregone earnings and the tax rate on the earnings increment. Finally, Section 3.8 outlines the average and marginal returns to costs ratio of government investment in education.

3.2. The costs of education

The main financial costs of a skills investment considered in this study for a student are:

Direct costs: tuition fees, books, computers, materials and other similar costs for students.
Foregone earnings: the reduction in earnings that takes place while a student is studying.

These costs are offset by:

Scholarship and grant income provided to the student by the government. The study abstracts from scholarship and grant income provided to students by other non-governmental entities.
Reduced taxes: as the students’ earnings decrease, they will pay less in tax. This offsets the cost of their skills investment, compared to a world without taxes where their foregone earnings would be larger.
Skills Tax Expenditures (STEs): special tax provisions that may offset the direct costs of skills, such as tax allowances and credits for tuition.

These separate factors are combined to calculate the overall cost of an education investment. In addition, it may be that a student is unable to finance their education with savings and so may have to incur debt to finance their education. These extra costs related to debt financing of education are discussed in Section 3.3. For the remainder of this section, it is assumed that a skills investment is financed with savings for ease of exposition.

Foregone earnings

Education is time-intensive. With many courses of study, full-time work and even part-time work cannot be continued. These lost earnings constitute a major barrier to education, especially for adults. In this study, a variety of assumptions about lost earnings during education are examined. Generally it is assumed that a student can earn 25% of their previous wage while they are in full-time education. In this way, the model captures the situations of older workers who may be able to continue to work part-time during education. In the case of 17-year-old university students, it is assumed that they can earn 25% of the average wage in their country during schooling, which is taken as an estimate of the earnings available from part-time low-skilled work.

Foregone earnings are affected by the tax system. The way in which this occurs depends on the tax system concerned. In a proportional tax system, foregone earnings will be reduced at the student’s statutory tax rate. In a progressive tax system a taxpayer may drop to lower tax brackets as they earn less during education. This will reduce the taxpayers’ foregone earnings even more. This means that while progressive tax systems discourage skills investments by taxing away the returns they also encourage investments in skills by reducing foregone earnings.

This is a key issue for the design of tax systems. Highly progressive tax systems are generally thought to discourage skills investment, but this progressivity can also encourage skills development if tax progressivity moderates the amount of foregone earnings during education. If taxes are progressive, income net of taxes will fall at a lower rate than income gross of taxes, reducing the cost of skills investments.

The fact that many countries provide benefit support for those on low incomes may mean that those investing in skills may not see their incomes fall when they invest in skills – their incomes may be subsidised by social benefits of different kinds. In such countries income during a skills investment does not fall as much as it would in the absence of the tax-benefit system. If after-tax income does not fall much over the range over which pre-tax income falls during the skills investment, then the tax system raises the incentives to invest in skills.

The importance of reducing lost earnings as a form of support for skills is even more important when social benefits are considered. In some countries, students on low incomes can avail of a variety of social benefits such as housing benefits, medical benefits, or other social transfers. These benefits can further reduce forgone earnings and thus provide a significant extra incentive to invest in skills by reducing lost earnings. These social benefits are not, however, captured in the results presented in this study which focuses mainly on the tax system.

Direct private costs

Direct costs include costs directly paid by the student to acquire skills such as tuition fees, books and materials, computers, registration fees, transportation costs of attending school, and so on. In this model, estimates of the costs of education are taken from the OECD’s Education at a Glance data (OECD, 2014).2 Further details are provided in Box 3.1. Figure 3.1 shows the estimates of direct costs for OECD countries of tertiary education. Due to data limitations, the calculations in this study use tertiary education costs as a proxy for costs of lifelong learning, workplace training, and so on.

Box 3.1. Calculating BEI and ETR inputs from Education at a Glance

To calculate the Effective Tax Rates (ETR) on Skills and other skills indicators discussed in this study, three key data points are required regarding education spending in each country examined. These are:

The direct spending by each student on their education (DC_W).

The direct spending by the government on the education of each student (DC_G).

The scholarship and grant income received by each student (SG).

Detailed data at an individual level are not currently available. The approach taken in this study is to estimate average levels of spending on DC_W, DC_G, and SG, as proxies for spending by individuals.

It is also important to note that estimates of DC_W, DC_G, and SG are based on OECD Education at a Glance (EAG) figures for third-level education. These spending estimates are used as proxies for educational costs for graduate education, for continuing mid-career education, and for spending on life-long learning. An exception to this is in-work training, where the assumption is that no scholarship and grant income is received from the state, and so set SG = 0. Future editions of EAG should contain more detailed data on costs of these different types of education, in which case these figures could be revised. In this study, the latest data available is from EAG 2014 which contains data for 2011. This data is combined with the Taxing Wages 2011 models, to ensure conformity between estimates of education spending and the tax system.

Direct spending by each student on their education (DC_W)

The formula used is as follows

Total spending is total spending on third level education, expressed in equivalent PPP USD. It is taken from EAG Table B1.1a, “Annual expenditure per student by educational institutions for all services”. The column used is Column 9, “all tertiary education”.
Private Fraction is the fraction of total spending that is undertaken by households. It is taken from EAG Table B3.1, “Relative proportions of public and private expenditure on educational institutions, by level of education”. The column used is Column 12, “household expenditure”. This omits non-household non-government private education spending, such as spending by firms or NGOs.
Cur is a currency adjustment factor, converting the amount spent by each student in PPP USD to 2011 national currency units.

These data sources are used for Australia, Austria, Belgium, Canada, Chile, the Czech Republic, Estonia, France, Iceland, Ireland, Israel, Italy, Japan, Korea, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, the Slovak Republic, Slovenia, Spain, Sweden, and the United Kingdom. For some countries in the analysis, data omissions from EAG require other data sources to be used.

For Denmark, separate data on household spending as distinct from private spending are not available. Data on all private spending is used. This means that total costs to the individual of educational spending are possibly inflated compared to other OECD countries.
For France, separate data on household spending as distinct from private spending are not available. Data on all private spending is used. This means that total costs to the individual of educational spending are possibly inflated compared to other OECD countries.
For Germany, separate data on household spending as distinct from private spending are not available. Data on all private spending is used. This means that total costs to the individual of educational spending are possibly inflated compared to other OECD countries.
For Greece, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2008.
For Hungary, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2000.
For Luxembourg, data on both the share of private spending in total spending, as well as on the total educational spending per student are unavailable for the entire EAG time series. Data for Belgium are used to proxy for educational costs in Luxembourg.
For Switzerland, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2000.
For Turkey, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2006.

Direct spending by the government on the education of each student (DC_G)

The formula used is as follows

Total spending is total spending on third level education, expressed in equivalent PPP USD. It is taken from EAG Table B1.1a, “Annual expenditure per student by educational institutions for all services”. The column used is Column 9, “all tertiary education”.
Public Fraction is the fraction of total spending that is undertaken by governments. It is taken from EAG Table B3.1, “Relative proportions of public and private expenditure on educational institutions, by level of education”. The column used is Column 11, “public sources”.
Cur is a currency adjustment factor, converting the amount spent by each student in PPP USD to 2011 national currency units.

These data sources are used for Australia, Austria, Belgium, Canada, Chile, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Iceland, Ireland, Israel, Italy, Japan, Korea, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, the Slovak Republic, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom. As with DC_W, for some countries in the analysis, data omissions from EAG require other data sources to be used.

For Greece, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2008.
For Hungary, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2000.
For Luxembourg, data on both the share of public spending in total spending, as well as on the total educational spending per student are unavailable for the entire EAG time series. Data for Belgium are used to proxy for educational costs in Luxembourg.
For Switzerland, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2000.
For Turkey, 2011 data on the share of public and private investment in total educational spending are unavailable. The shares of public and private spending used are from 2006.

Scholarship and grant income received by each student (SG)

The formula used is as follows

Sch % Total is total spending on scholarships and grants, expressed as a % of total spending on third level education, expressed in equivalent PPP USD. It is taken from EAG Table B5.4, “Public support for households and other private entities for tertiary education”. The column used is Column 2, “Scholarships and other grants to households”.
Total Spending % GDP is total spending on tertiary education, expressed as a % of GDP. It is taken from EAG Table B4.1, “Total public expenditure on education”. The column used is Column 7, “Tertiary education”.
GDP is simply a measure of GDP in PPP USD, taken from the OECD National Accounts.
# of Students is a measure of the total number of full-time students in a given year. It is calculated by dividing total education spending at third level by total spending per student at third level. The measure of total education spending is taken from EAG Table B4.1, “Total public expenditure on education”. The column used is Column 7. The figure for total spending per student is taken from EAG Table B1.1a, “Annual expenditure per student by educational institutions for all services”. The column used is Column 9, “all tertiary education”. Dividing these two figures yields an estimate of the number of students in third-level education.
Cur is a currency adjustment factor, converting the amount spent by each student in PPP USD to 2011 national currency units.

For the Greece, 2011 data on scholarship levels are unavailable. The shares of public and private spending used are from 2008.
For the Luxembourg, data on scholarship levels are unavailable for the entire EAG time series. Data for Belgium are used to proxy for educational costs in Luxembourg.
For the Turkey, 2003 data on scholarship levels are unavailable. The shares of public and private spending used are from 2008.

Scholarship income

Scholarship and grant income reduces the cost of education, and so is a key component of the net costs of investment in skills. Data on scholarship income is also taken from Education at a Glance (OECD, 2014).3 Details of scholarship income calculations are discussed in Box 3.1. It is assumed that scholarship income is available to university students, graduate students and to those engaged in lifelong learning. It is assumed that it is not available to those mid-career training. The amounts of scholarship income used for each country are shown in Figure 3.2.

Special tax expenditures for education spending or scholarship income

STEs are also included as a component of the “costs” of education, for the same reason as scholarship income. That is, tax expenditures that reduce the costs of education are a form of financial benefit that only accrues to someone pursuing a course of study (they would not be given to someone not pursuing such a course). These STEs can come in the form of reductions in taxable income by the amount spent on education, or reductions in tax liability in some proportion to education spending. They can also come in the form of tax benefits that reduce or exempt scholarship income from taxation. Finally, in some countries student wage income is subject to lower levels of income taxation or social security contributions. These tax expenditures are discussed in greater detail in Chapter 2.

The total costs of education

The previous sections have outlined the key cost components of education. Key private costs of education are foregone earnings, direct costs such as tuition fees, and extra taxes that may need to be paid on scholarship income. These costs are offset by reduced taxes on earnings, scholarship and grant support, and STEs. Foregone earnings are often the most significant cost component of education. In sum, the total costs of education consist of:

In the model, the shorthand used is:

represents Direct Costs minus Scholarships and Grants. represents the net impact of STEs. This equation is discussed in more detail in Annex A of this study. The value and impact of STEs is discussed in more detail in Chapter 5.

3.3. The financing of the student’s education costs

The student’s education can be financed in different ways. This study considers just two options: that the student finances a skills investment either with retained savings or with a loan from the government.

Loan features

Several features of the student loans modelled in this study are important to point out.

There is a fixed interest rate in place for the duration of the loan, which can be higher or lower than the risk-free interest rate that an alternative investment might earn.
The duration of the loan may vary, it may last the duration of the students remaining career in the workforce, or it may last for a shorter period. Repayments on the loan only begin once the student has left education (so interest or principal repayments may not be payable while the student is still upskilling).
The loan is structured as a bond. This means that if the term of the loan is years, then the student pays interest on the loan for the first periods. In the final period an amount of interest is paid, and the principal is repaid.
Both the amount of the interest payable in each period as well as the principal to be repaid at the end of the loan term are fixed in nominal terms.
Debt write-offs may exist; that is, some fraction of the principal to be repaid can be written off by the government in the final period. The effects of increases in income from loan write-offs on tax liability are not modelled.
The interest paid on the loan in each period after education may be set against tax liability (depending on the provisions in a given country’s tax system). This may take the form of deductibility of interest from taxable income. Other similar provisions can also be incorporated.
Loan repayments can be made dependent on income after education. This means that if a taxpayers’ income does not pass a certain income level, their repayments may be smaller, or they may not have to make any repayments at all.

Modelling loans

To model the impact of these loans on the incentives to invest in skills, the costs of education are weighted upward or downward depending on the specifics of the loan concerned. For example, if the interest rate on a loan is higher than the risk free rate which a student could earn on savings, then debt financing increases the costs of education compared to financing education with savings. By contrast, if loans are subsidised such that the interest rate is below the risk-free rate, then debt financing reduces the cost of education. The total costs of education incorporating these financing costs are referred to as Total Financing Costs, or TFC. If a fraction TFC of the total costs of education TC are borrowed, then TFC is expressed as follows:

Where the term F is a composite term of the different ways that

The value of the loan write-offs spread over the duration of the loan.
The value of the differential between the real interest rate at which the student can borrow r*, and the risk free interest rate which he can earn interest on savings r. In the absence of capital taxes, this value is simply , where r* is the rate at which the student borrows.4
The value of any tax deductibility of interest payments.

The overall value of F is

F is then multiplied by the total costs of education, the fraction of these costs that are borrowed, and appropriate discount factors, to arrive at the correct weighting of the true costs of education. This is discussed further in Annex A to this study. All in all the TFC measures the total cost of education accounting for how the education is financed. As outlined, government policies can reduce these costs to the individual directly, by reducing tuition fees or increasing scholarship income, or by reducing financing costs, by subsidising loans or expanding loan write-off provisions.

3.4. The returns to education and the breakeven earnings increment

In order to recoup the costs of education, the student expects to earn higher wages after education. The returns to education for the individual come in the form of higher after-tax wages. Increases in the probability of employment after education, increases in the pace of wage gains, and other benefits to education are not modelled. In this study, it is assumed that wages rise once post-education, and then continue to rise linearly with inflation thereafter. The term “breakeven earnings level” is used to describe the level of pre-tax earnings that must be earned after education in order to make the skills investment worthwhile. The previously-discussed BEI is the difference between the breakeven earnings level and income before the skills investment. It is helpful to think about the breakeven earnings level as being made up of several components. This is another way of asking what must a student earn to make a skills investment worthwhile?

First, earnings after education must be at least as much as earnings before education. The methodology assumes that wages rise naturally with inflation, but do not rise otherwise unless a skills investment is made. In the absence of education, a worker’s real wage will stay constant until retirement. For a skills investment to break even, workers must earn at least this level of income after education.

The second component that must be earned for a skills investment to break even is a return that allows the worker to recover the costs of the skills investment. Physical capital investments can be sold at the end of the period of use to recoup initial costs of the investment.5 Human capital cannot be readily sold; retirees cannot sell off their talent or knowledge when they retire. In other words, a skills investment depreciates entirely upon retirement. This means that, in contrast to a physical capital investment, a human capital (skills) investment must earn enough to repay these initial costs over the course of the investment. This raises the amount of earnings needed for a skills investment to break even relative to a physical capital investment.

The third component of the required return for a skills investment to breakeven pays for the opportunity cost of spending on skills. If a student or worker had not made a skills investment, they could have invested the cost of the skills investment in an alternative capital asset which would have yielded a return. The opportunity cost of such an alternative investment must also be recouped in order for a skills investment to break even.

Finally, a skills investment must earn enough to recoup whatever extra taxes are owed as a result of any extra earnings. This tax rate is referred to in this study as the tax rate on the earnings increment, T_EI. There is an element of endogeneity to this process; as a taxpayer must earn more after education, their tax rate will rise. This means that the component of the required return to pay for the taxes will rise also –students must earn even more to break even, increasing their tax liability even further and so on. This is illustrated in Figure 3.3.

In the taxation and skills model, the required level of earnings is calculated such that the pre-tax earnings are enough to break even. The total picture of costs and returns is illustrated in a simple way in Figure 3.4; post-tax income, flat before education, falls during a period of education; due in part to foregone earnings, and in part to direct costs. Wages then increase after education to break even on the investment. The necessary size of this increase has several components: the necessary amount to recover the initial costs of education, the necessary amount to recover the returns to an alternative investment, and the extra earnings needed to pay for the extra taxes incurred. The latter component is referred to as the skills tax wedge.

3.5. The marginal effective tax rate

The METR is the effect of the personal income tax system on the incentives of a “marginal” student to undertake a skills investment. Specifically, it is the difference between the BEI in a world with taxation and the BEI in a world without taxation, divided by the BEI in a world with taxation. It answers the question of whether a student would need to earn more after education to break even in a world without taxes compared to in a world with taxes.

The METR on skills calculated in this study is marginal in the sense that it is the effect of taxes on a person who is just indifferent between making a skills investment and not making one. In the same way that a METR on labour is the tax rate on the return of the last unit of labour a person is supplying, so the marginal tax rate on skills is the tax rate on a person who is just considering making a skills investment. This is why the level of earnings at which the tax rate is calculated is the breakeven earnings level.6

The METR calculation procedure can be summarised as follows. First, the three elements of what a worker would need to earn to pay for a skills investment are calculated. The extra amount needed to pay for taxes on these higher earnings is then added. This means the model must find an equilibrium solution because the tax rate that is paid on these extra earnings changes as earnings rise, requiring yet more earnings to pay for extra taxes. The model finds a point where the worker is just indifferent between making the investment and not. Once the BEI has been found, it can be expressed as a share of previous earnings. The increase in the BEI as a result of the tax system can then be calculated by subtracting the BEI in the absence of taxes from the BEI in the presence of taxes. The share of the extra earnings that is necessary solely because of taxes, as a share of the earnings increment needed to break even on a skills investment in the presence of taxes, is called the METR.

3.6. The average effective tax rate

In addition to calculating METRs, this study also calculates an AETR, an Average Effective Tax Rate on Skills. This tax rate is similar to the METR in that it is the difference between the returns to skills in a world with taxes and a world without. The difference between the METR and the AETR is that the METR calculates the effect of the tax system on a breakeven skills investment, while the AETR calculates the effect of the tax system on a skills investment in a more general way.

Specifically, the model calculates the net present value of a skills investment with taxes, and the net present value of a skills investment without taxes. The AETR is the difference between these two values, expressed as a fraction of the net present value of the earnings increment resulting from education. Specifically, the model calculates:

To find the AETR, it is necessary to calculate the value of an average skills investment. This is calculated in a similar way to the marginal skills investment; the costs come in the form of foregone earnings and direct costs, offset by taxes on foregone earnings, scholarship income, and tax provisions for skills investment. On the returns side, however, an important difference is that the breakeven earnings level is not calculated. The AETR is instead based on a fixed return, which may be higher or lower than the breakeven earnings level.

For the estimation of the returns to undergraduate education, data on the actual tertiary education premium for 15-64-year-olds from Education at a Glance are used (OECD, 2011). Unfortunately the data on the earnings returns for graduate education, for lifelong learning and workplace training are not as extensive as for undergraduate education. Instead, an assumed 15% return on a year of education is used in these cases. Estimates of the tertiary education premium are given in Figure 3.5. In most cases, the average return to a tertiary education available in the labour market is well above the required BEI calculated in the tax and skills models. Chapter 4 discusses this point further.

Using these returns, the model calculates the net present value of a skills investment with and without taxes. The difference between these two values, expressed as a fraction of the net present value of the returns to education with taxes, is the AETR.

3.7. Understanding the results

The effect of the tax system on a skills investment, as discussed, is the difference between the returns to skills without and with taxes. In the METR case, this difference is measured at a breakeven earnings level; in the AETR case, this difference is measured at some other (usually higher) earnings level, or at the earnings level based on labour market data where available.

In practice, the two tax rates are a function of how the tax system reduces the cost of an investment in skills, and taxes away the returns to skills. As mentioned, the PIT system reduces the cost of upskilling by reducing foregone earnings; the amount of income foregone during education is offset by the fact that a student also foregoes the taxes which would have been paid on these earnings. In addition, tax liability can be reduced in some proportion to the direct costs of education. The tax system reduces the cost of skills in these two ways. As these two subsidies (the tax rate on foregone earnings and the subsidies for skills costs and scholarship and grant income) increase, the METR and AETR will fall.

On the returns side, the tax system reduces the returns to skills by taxing them away; as a worker earns more after education, tax progressivity often means that they pay taxes at a steadily higher rate. Higher taxes, and tax progressivity, tax away the earnings increment after a skills investment. Increasing this tax rate on the earnings increment will increase the METR and AETR.

The tax system affects the financial incentives to invest in skills both positively and negatively. It reduces the costs of skills but also reduces the returns; the former through the Tax Rate on Foregone Earnings and the Tax Expenditures for Direct Costs and Scholarship Income, and the latter through the Tax Rate on the Earnings Increment.

The relative size of the TFE and the TEI is a function of the PIT and SSC tax schedules. In a proportional tax system, the marginal tax rate is the same regardless of the income level. This means that necessarily the TFE and the TEI will be the same. If direct costs are fully tax deductible, then the tax system should be neutral with respect to skills: the METR on skills will be zero. This is because the costs of a skills investment are being subsidised by the tax system at the same rate at which the returns to skills are being taxed away. Moreover, this will be the case whether the tax rate is low or high. At a high tax rate, the TEI and TFE will both be high. At a low tax rate, the TEI and the TFE will both be low. In both cases, where the rate at which the tax system reduces costs is the same as the rate at which it taxes away returns, the tax system is neutral with regard to the skills investment. In such instances skills investments that would be profitable from the individual’s perspective in the absence of taxes will be profitable in the presence of taxes (Brys and Torres, 2013).

In progressive tax systems, the tax rate rises with income. As wages after a skills investment will typically be higher than wages during a skills investment, the TFE will be less than the TEI in progressive tax systems: the tax system will tax away the returns to skills at a higher rate than it subsidises the cost. In these cases the METR will be positive. In this way progressive taxation can act as a disincentive to invest in skills, holding other factors equal.

In instances where the TFE and TEI are not equal, the overall impact of the tax system depends on the size of the returns to skills. Where returns are very low; the tax rate on the earnings increment does not matter as much; low returns means that the impact of the tax system through the way these returns are taxed is comparatively small. The effect of the tax system on the costs of upskilling predominates. Where returns are high, the tax rate on these returns is usually also high, and so this effect dominates the overall METR and AETR; the tax rate on returns matters more. These details are summarised in Table 3.1.

Table 3.1. Components of the METR and AETR
Name	Effect on METR and AETR	Dominates the METR and AETR when:
Tax Subsidy for Direct Costs	Decreases	Direct Costs of Skills are High
Tax Subsidy for Scholarship and Grant Income	Decreases	Scholarship Income Is High
Tax Rate on Foregone Earnings	Decreases	Foregone Earnings are High
Tax Rate on Earnings Increment	Increases	Returns to Skills are High

Figures 3.6 and 3.7 illustrate these dual effects of the tax system on incentives to invest in skills using concrete data. The example used here is that of a 32-year-old worker investing in one year of education, earning 25% of their earnings before education while doing so. Income before education (and thus foregone earnings) varies along the x-axis. The y-axis shows three key tax rates in both figures. In Figure 3.6, the top line shows the marginal tax rate on the earnings increment; the rate at which the returns to skills are taxed away. It is also a measure of how the METR is being increased by the tax system. The bottom line is the negative of the tax rate on foregone earnings; it is a measure of how the tax system is subsidising skills investments. It is also a measure of how the METR is being reduced.

The line in the middle of Figure 3.6 is the METR on Skills. For each country, it is a weighted average of the tax rate on the returns to skills investments, and the tax subsidy of the costs. Where the tax rate on the earnings increment (the top line) rises sharply, the METR rises as well. Where the tax subsidy of the cost rises (where the bottom line falls) the METR falls as well. This clearly shows the overall impact of the tax system is a weighted average of its positive effects (by reducing the costs in terms of foregone earnings) and its negative effects (by reducing the returns in terms of the earnings increment).

Figure 3.7 shows AETRs, instead of METRs in Figure 3.6. As with Figure 3.6, The AETR is the line in the middle, with the tax rate on the earnings increment and the tax rate on foregone earnings being the lines above and below respectively. As mentioned above, the earnings level with which the AETR is calculated is usually higher than the breakeven earnings level. This means that skills investment usually makes enough to pay for itself and more; the breakeven earnings level is reached and then passed. This means that the returns to skills are higher in Figure 3.6 than they are in Figure 3.7, while the costs remain the same.

This can be seen by comparing the top and bottom lines in each figure. The costs remain the same in both cases; so the effect of the tax system on the costs of upskilling is the same. This is why the bottom line is the same for each country.

What changes is the estimation of the returns to skills, and the tax rate on these returns. In Figure 3.6, the model examines a METR, based on a breakeven return. In Figure 3.7 it examines an AETR, based on an assumed return that is higher than the breakeven earnings level. Post-education earnings are higher in the average case than they are in the marginal case; the tax rate on this earnings increment is higher as well. This is clear as the top line is higher in most country cases.

The difference between these two graphs also illustrates that both the AETR and METR are weighted averages of the positive and negative effects of the tax system on incentives to invest in skills. The weight is the earnings increment; where returns are high, the tax rate on the earnings increment is a larger determinant, where returns are low, the tax system on foregone earnings is a larger determinant. In Figure 3.7 returns are higher; the AETR (the line in the middle) is closer to the tax rate on the earnings increment than in the marginal case.

The impact of taxes on skills investment is a function of the subsidies the tax system gives to skills investments that reduce their costs and the taxes on the returns to skills investments that reduces these returns. Moreover, for low-return skills investment, it is the subsidies on costs (both directly and as foregone earnings) that have the largest effect. For high return skills investments, it is the tax rate on returns (not subsides of costs) that matters most. This can be seen in the below formula which describes the relationship. This result is derived in Annex A.

This result shows that as the return to skills grows large, the tax rate on the earnings increment, T_EI, has a higher weight in the AETR. When the returns to skills fall to the breakeven level, the AETR falls to the METR. This is the same result that can be seen in Figures 3.6 and 3.7.

3.8. The returns to costs ratio of investment in education

In addition to examining the financial costs and returns to education from the perspective of the individual, this study also explores the financial costs and returns to investment in skills from the perspective of the government. This is done through the MRCR and ARCR, the Marginal and Average Returns to Cost Ratios. These are measures of the ratio of the returns to education to the costs of education for the government.

Costs of education for government

On the cost side, the government’s cost of education has six main components.

The first is the government’s direct spending on education; on teacher salaries, grants to private universities, public universities and so on. As with the data on private educational costs and scholarship income, estimates of government direct educational spending are taken from Education at a Glance 2011. These estimates are shown in Figure 3.8. This is discussed further in Box 3.1.
The second is the scholarship and grant income that is provided by the government to the student. Note that the model assumes no private scholarship income; it is assumed that all scholarship income received is received by the student from the government.
The third component is lost taxes while a student is working. This is exactly the converse of the after-tax foregone earnings component of costs to the student. While the student loses their after-tax earnings when they are in school instead of working, the government loses the tax revenue it would have earned had the student continued working instead of educating themselves.
The fourth component is the lost tax revenue that results from STEs. The government may lose tax revenue if STEs exist that defray the costs of education for the student of the kind outlined in Section 3.1.4.
The fifth component is lost taxes that may result from deductibility of interest payments from workers as they repay student debts.
The sixth component is fiscal costs that may accrue to the government if it lends to students to finance education at rates lower than it can borrow, or from the government writing off student loans, including in systems when repayment of these loans is contingent on a worker’s income. By contrast, if the government is able to earn real returns on student loan provision, these returns may offset government costs elsewhere.

In terms of returns, the model considers only returns to government in the form of increased tax revenue from the higher wages that workers earn after education. In doing so, it abstracts from the many other positive benefits that increased education will bring to the government, such as increased indirect taxation when the extra income earned is spent, reduced spending on employment and social benefits due to a more productive population, higher growth, lower probability of unemployment, and potentially greater social cohesion. The true returns to governments from education are likely to be considerably higher than those estimated in the model.

Average returns to cost ratio

A very simplified formula of the ARCR can be expressed as follows:

In the model, this formula is appropriately time-discounted. A more technical expression of the ARCR can be found in Annex A of this study.

This formula is, like the METR, AETR and BEI, incorporated into the Taxing Wages models to calculate future tax revenue based on earnings assumptions. As with the AETR, labour market estimates of the lifetime tertiary education premium are used to estimate the increased tax revenues from upskilling for the government.

From a technical perspective, the ARCR functions as a ‘Tobin’s q’ in economic investment theory. Values of the ARCR below one suggest that the returns to education (in tax revenue alone) are insufficient to recoup the costs of education for the government. Values of the ARCR above one suggest that the returns to education in the form of tax revenue more than cover the government’s costs.

This in turn has implications for government spending on education, and for the government’s taxation of the returns to skills. In investment theory, a Tobin’s q above one suggests that the investment should be continued. In similar terms, ARCRs that far surpass one suggest that governments could increase tax revenue by increasing skills investments. By contrast, very low ARCR values could suggest that the returns to government in income tax revenue do not cover its costs from investments in upskilling at current spending levels.

It is important to realise that the ARCR does not provide information about the overall ratio of costs and benefits of education for society. It merely does so for the government. For example, it is possible for the ARCR to be far below one for the government, but for the skills investment to be profitable from an overall social perspective. It could simply be that the tax rate is too low for the government to recoup its spending on education; meaning that returns to a skills investment for the student are very high. Similarly, a high ARCR could mean that the government is spending little on education; that a large fraction of education spending is being financed privately by students, but the returns are being taxed away by the government.

Essentially, then, the ARCR is a measure not of the total ratio of costs to returns of education, but of the way these costs and returns are being shared between the government and the student. Low ARCRs suggest that the government is receiving a lower share of the returns to education than it is bearing in costs. This may not mean that education is not profitable or worthwhile overall, but simply that large shares of the returns are being captured by the student. Higher ARCRs suggest that the government is receiving a higher share of the returns to skills investments than it is bearing in costs. In such cases, the financial incentives to invest in education on the part of the student may be lower than they should be. This is illustrated in a very simple way in Table 3.2.

Table 3.2. Components of the MRCR and ARCR
	High Taxes on Earnings Increments	Low Taxes on Earnings Increments
Low Government Spending/Low Taxes on Lost Earnings	High ARCR – Government receives a higher share of the returns than its share of the costs	ARCR near 1 – low share of the returns, but a low share of costs as well
High Government Spending/High Taxes on Lost Earnings	ARCR near 1 – high share of the returns, but a high share of the costs as well	Low ARCR – Government receives a smaller share of the returns than its share of the costs

The results also suggest that raising the employment rate provides significant returns to the government by increasing the returns to its skills investment. For existing members of the labour force the formula suggests that unemployment can be very costly for governments who pay a large fraction of the costs of skills investments; a high unemployment rate means that sunk costs of skills investment are not being recouped by the government in the form of tax revenue.

It is important to note that many simplifying assumptions are made in the model. As mentioned, the model does not account for a wide variety of the financial and social benefits to education. These other benefits may occur in terms of tax revenue from taxes other than income tax, reducing social and other expenditure, and other social benefits. In addition, the model is static in its approach; the tax system is assumed to remain the same over years to come, as are the responses of wages to increased skills. Present taxes and present tertiary education premiums proxy for taxes and tertiary education premiums in years to come. As such, it is likely that the model underestimates the returns to education for both the student and government. The returns to skills are likely to rise steadily over the years to come, and comprise much more than just income tax revenue. The ARCR indicator is more useful as a guide of the relative position of countries with respect to how the costs and returns to education are distributed between governments and students, as opposed to an overall measure of the benefits of skills investment itself.

Marginal returns to cost ratio

Sections 3.4 and 3.5 discussed two tax rates on individual’s investment in skills, the METR and the AETR. As discussed, the AETR measures the effect of the tax system on the financial incentives to invest in skills for a student earning the average return to skills in the labour market. By contrast, the METR on skills measures the effect of the tax system on the financial incentives to invest in skills for a student earning a breakeven return on a skills investment.

In addition to the Average Returns to Cost Ratio, the model allows for the calculation of a Marginal Returns to Cost Ratio. As outlined, the ARCR measures the ratio of returns to costs from the government’s perspective where a student earns an average amount of returns on their skills investment. This means that from the government’s perspective, returns in the form of tax revenue will be average as well. By contrast, the MRCR measures the ratio of returns to costs for the government when the student earns just a breakeven return. This indicator measures the incentives of the government to educate a student who will just break even on a skills investment. As with the ARCR, values of the MRCR above one suggest that the government will more than recoup their costs from skills investments.

As will be discussed in Chapter 4, breakeven earnings levels in the OECD are usually well below the average tertiary education premium: students more than recoup the costs of their education. This means that a hypothetical “breakeven” student will earn much less in this analysis than the average student. This in turn means that the tax revenue the government receives from a breakeven student will usually be significantly lower than that of the average student. This is true regardless of the other components of the ARCR and MRCR.

This in turn means that MRCRs are usually much lower than ARCRs in OECD countries. This is illustrated for a sample case in Figure 3.9. It can be seen here that ARCRs are larger than one for most countries at most income levels. However for most countries and for most income levels – though not all – MRCRs are not above one. This illustrates that for the government, the returns from educating a student who is just breaking even on a skills investment are lower than the returns to educating an average student.

References

Brys, B. and C. Torres (2013), “Effective Personal Tax Rates on Marginal Skills Investments in OECD Countries: A New Methodology”, OECD Taxation Working Papers, No. 16, https://doi.org/10.1787/5k425747xbr6-en.

Klemm, A. (2008) “Effective Average Tax Rates for Permanent Investment”, IMF Working Paper, No WP/08/56

OECD (2014), Education at a Glance 2014: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/eag-2013-en.

OECD (2016), Taxing Wages 2016, OECD Publishing, Paris, https://doi.org/10.1787/tax_wages-2016-en.

Notes

← 1. The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.

← 2. Data for direct costs of education in Luxembourg are not available in Education at a Glance; data for Belgium are used as a proxy.

← 3. Data on scholarship and grant income in Luxembourg are not available in Education at a Glance; data for Belgium are used as a proxy. More details are provided in Box 3.1.

← 4. In the presence of capital taxes, the capital tax inserts a further wedge between what the student can earn on a risk-free investment and the rate at which he can borrow. This is because capital taxes must be paid on the nominal return on any investment opportunity. In this case, instead of , the difference would be .

← 5. This is true in a setting without uncertainty, as is the case in the framework presented here.

← 6. The mechanics of the METR and BEI are discussed in detail in Brys and Torres (2013).