Home | OECD iLibrary

5.1. Introduction

The support provided by the tax system to improve individuals’ financial incentives to invest in skills varies widely. Chapter 4 of this study has provided estimates of the overall impact of the tax system on different kinds of skills investment decisions. The incentives to invest in skills depend on the overall personal income tax (PIT) system, but also on specific skills tax expenditures (STEs) countries have in their tax systems.2 Some countries allow students to deduct the private costs of education from their taxable income or credit this expenditure against their tax liability; some countries exempt scholarship income from taxation; some countries exempt student income from social security contributions. Since all of these measures are designed to increase investment in skills, assessing their impact and comparing their impact to those of other skills policies is crucial in assessing the overall impact of the tax system on skills.

This chapter approaches these issues in several ways. It analyses the value of these financial incentives to different kinds of students at various income levels and for different kinds of skills investment scenarios. It also assesses the distributional impact of STEs, discusses students’ responses to these incentives and presents estimates of the effects of these expenditures, by drawing upon analysis from the economic literature. The analysis in the chapter makes several key points.

The value of STEs in the countries modelled is modest. Though dependent on the costs of skills investment and other assumptions, the results suggest that for university students, scholarship income and direct government support are much larger channels through which governments financially support skills investments.
STEs are progressive. This means that the estimated value of these STEs is larger for those on lower incomes in most countries. STEs are more progressive for university students than they are for workers job-related training. This is particularly the case where the STEs are reduced social security contributions (SSCs) on student income.
Direct support for university students in the form of direct grants and scholarships is more progressive than STEs, in part because direct support does not depend on an individual’s taxable income, which is often the case for STEs.
While the model notes that STEs can benefit those students on low incomes more than students on higher incomes, empirical research has suggested that the impact of these STEs on student outcomes such as enrolment and time-to-completion is modest or even zero.
In order to be effective, STEs must be very carefully designed. Existing STE policies are often designed in ways that make it difficult for students to access them, in ways that do not encourage completion of courses of study in a timely manner, and in ways that do not encourage the completion of the kinds of education that might be most effective in raising wages and employment prospects.
The impact of STEs on financial incentives to invest in skills is higher for those skills investments that yield low returns than for those skills investments that yield high returns. For high-return skills investments, the way the returns to skills are taxed away (i.e. tax progressivity) has a much larger impact than the way the costs of skills are offset (including STEs). For low-return skills investments, the support the financial system provides by reducing the costs of skills investments has a larger impact than the way these returns are taxed away.
As policy levers to encourage investments in skills, STEs may be less effective in comparison to direct spending on educational institutions or direct support to students in the form of scholarships and grants.
Existing skills tax expenditures are often only available where training pertains to a workers’ current employment, and may not facilitate workers who need or want to change careers. These tax provisions may therefore reduce labour market flexibility and exacerbate skills mismatch.

These key points lead to important conclusions that clarify choices for policy makers when it comes to tax and skills. A uniquely “best” set of tax and skills policies does not exist. The ways in which tax and skills policies and indeed financial support for skills in general should be designed depends on the goals of the policy maker with respect to skills outcomes in a given country. As discussed in Chapter 4, the tax system apportions the costs and returns of skills investments between the student, firm and government. Where governments want a very progressive income tax system, a large amount of government support for skills investment will be required to maintain and expand the level of human capital in the economy. Where the tax burden on human capital is low, government support for skills investments can be more modest without excessively impacting students’ and individuals’ incentives to invest in skills – provided adequate support is provided to credit-constrained and otherwise disadvantaged students.

Other considerations must also be borne in mind when considering the impact of the tax system on skills. Tax expenditures are often more complex for governments to administer and for taxpayers to understand than direct spending on skills investment. Many tax expenditures are only of value to those students who earn taxable income. This means that they may be a costly way of investing in human capital for governments, and may have low take-up rates, especially among those with low incomes and low skills. Tax expenditures may not be as effective an approach to supporting skills investments as direct spending. Tax expenditures may also involve significant deadweight losses, providing subsidies to skills investments that would occur even in the absence of government support (though this may be true of direct spending on skills as well). To raise the overall level of skills in the economy, a combination of reduced labour income taxation and increased government support for those on low incomes or those who are credit constrained may be appropriate.

The chapter proceeds as follows. Section 5.2 reviews the size and impact of STEs. Section 5.3 reviews other non-tax approaches to improving students’ financial incentives to invest in tertiary education, such as loans, scholarships and grants, and reducing tuition fees by providing more government support to educational institutions. Section 5.4 considers tax and non-tax issues targeted at lifelong learning and worker training. Section 5.5 concludes with an overall consideration of the role of the tax system in education finance and the system of incentives surrounding skills investment.

5.2. Tax incentives for tertiary education

This section analyses STEs for tertiary education. As has been outlined in Chapter 1, the returns to tertiary education are significant (OECD, 2015a). In most OECD countries, the government provides significant direct support for tertiary education (see Chapter 3, Section 3.1). By contrast, the support provided to tertiary education through the tax system is more modest. STEs supporting tertiary education come in several different forms, including:

Exemption of scholarship income from personal income taxes or employee and employer SSCs,
Exemption of student earnings from personal income taxes or employee and employer SSCs,
Deductibility of tuition costs and other educational expenses from the personal income tax base,
Tax credits by which skills expenditure can be credited against tax liability,
Deductibility of interest on student loans from the personal income tax base,
Income contingent loans.

This section will outline the approximate size of the value of these STEs for the stylised tertiary student outlined in Section 4.2, and will then discuss the impact of these expenditures on educational outcomes and income distribution. The section will conclude with a discussion of design considerations with regard to tax expenditures for tertiary education.

Size of tax incentives for tertiary education

This section provides estimates of the value of STEs in the sample of countries for tertiary education on a per-person basis. In assessing the value of these expenditures, the ‘skills spending’ that happens through the tax system can be assessed for stylised cases. This allows spending on skills that happens through the tax system to be compared with spending on skills that is provided directly, for example through direct grants to universities or students.

Assessing the overall value of STEs is challenging. These STEs may vary in value depending on the income of the taxpayer, the kind of education undertaken, the amount of tuition fees, the amount of scholarship income and other factors. Moreover, different STEs may exist in the same system, and may interact with each other. For example, student income may be exempt from social contributions, but increased student income may reduce eligibility for a tax-free scholarship. The model outlined in Chapter 3 of this study allows all STEs to be analysed together, to arrive at a general estimate of the value of these STEs for a stylised student.

In terms of the model as outlined in Chapter 3 of this study, the key parameter that will be analysed is the difference between the tax rates applied to earnings during education with and without STEs. This can also be expressed as:

,

where T_Ed is the rate of tax paid on in-education earnings E_d without STEs, and T^*_Ed is the rate paid on these earnings with STEs. Estimations of the value of STEs depend on a variety of factors and assumptions of the model, which are also discussed in Chapter 3 and in more technical detail in Annex A. For example, where scholarship income is tax exempt, those receiving larger amounts of scholarship income will receive a larger tax expenditure. In the model, an estimate of an average amount of scholarship income for each student in each country is taken, and the estimate of the size of the tax benefit stemming from the exemption of scholarship income depends on this assumption.

These estimates also depend on the income and family circumstances of the taxpayer. The results presented here are for single taxpayers without children. In addition, the model does not account for parental spending on education and for tax support for that spending.3 This is due to the challenges of accounting for the mix of spending between parents and children, the nature of that spending (whether parental spending on children’s education is an investment or consumption form of spending) and the apportionment of returns. These considerations all make it difficult to fully account for parental spending on education in the model.4

Figures 5.1 and 5.2 show an estimation of the size of STEs in the personal income tax system based on the stylised university education scenario outlined in Chapter 4. The university education case examines a 17-year old person undertaking a four-year course of study. University students receive an average level of government scholarship income and pay an average amount of direct costs.5 Their earnings during education are set at 25% of the Average Wage in any given country – that is, students are assumed to earn some part-time employment income, but at a comparatively low level. Figure 5.1 shows the value of STEs in PPP 2011 US dollars. Figure 5.2 shows the value of STEs as a share of the average wage in a given country.

The value of STEs as outlined in Figure 5.1 is comparatively small. Significant subsidies are provided through the tax system in Belgium, in Norway, and in the United Kingdom. These results are mainly driven by tax deductibility of scholarship income. In these three countries scholarship income is both tax deductible and large compared to the estimated cost of university education. Nonetheless, in most countries the value of STEs is modest: the typical student in the model receives USD 128 per year in indirect subsidies for tertiary education through the tax system.

It is possible that these data underestimate the true value the tax system provides to some students in the form of STEs. This may occur in several ways. First, an estimate of the average amount of scholarship income in each country is used in the model. As much of this income is tax-exempt, these deductions provide a significant form of tax expenditure for skills. However if certain students receive more scholarship income than other students, their tax expenditure will rise. The case of skills tax allowances is similar. If spending on tertiary education by some students is high – for example through tuition fees – they may benefit further from tax deductions for these skills investments than they do in the model.

It is notable that for several countries the estimated value of STEs at tertiary level is zero. For Iceland no such STEs are currently provided to students. For certain other countries such as Chile, Italy and Israel, STEs are available, but student income is assumed too low in the model for the STEs to be of benefit – students are already assumed to have little or no tax liability. This will be discussed further below.

It is notable that for Poland the estimates reported are negative. This can be explained by noting the results in the figure pertain to STEs in the personal income system only. This graph omits tax expenditures that may benefit students through SSCs. This conforms to the approach taken in Chapter 4, where only personal taxes are considered. A key method by which Poland and other countries support skills investments at the tertiary level is through reduced SSCs for students. However, as personal income tax in these countries is levied on income net of SSCs, reducing these contributions means that taxable income net of these contributions rises, and so PIT liability also rises. This results in the negative effect, when only PIT is accounted for in the graph.

Figure 5.3 reports the value of STEs incorporating SSCs. The graph shows that STEs that come in the form of SSC reductions (as is the case in Poland) are significant in size compared to STEs that come in other forms. For some countries it alters the value of STEs. Incorporating these provisions, the typical student in the model receives USD 860 per year in STEs for tertiary education through the tax system, a substantial increase from the USD 128 discussed above. Countries like Poland, with negative support for skills in the PIT system and countries such as Slovenia and the Slovak Republic who do not provide support for skills through the PIT and SSC systems are shown to have significant support when tax exemptions for SSCs on student income are taken into account. These results are dependent on the amount of student income being earned. Many of the STEs being provided through the SSC system are reductions on student wage income, meaning that those with higher wages will see the value of their STEs rise.

The model shows overall that the value of STEs remains modest for tertiary education, especially compared to direct support for tertiary education provided by governments. This section has also noted that there are reasons to believe that this support is heterogeneous across the population in the group of countries analysed. The value of STEs is often conditional on tuition fees, scholarship income, and wage income for students. In addition, significant support is provided in the form of reduced SSCs on student income. As students on low incomes or from low-income households are more likely to receive scholarship income, and work during their years of tertiary education, it is likely that those students from low-income households receive more benefit from the tax system than is estimated in the model.

Impact of tax incentives for tertiary education

This section assesses the impact of STEs on financial incentives to invest in skills. Specifically, this section considers the extent to which the presence or absence of these STEs impact the Breakeven Earnings Increment (BEI), the overall indicator of incentives to invest in skills in the model. It also assesses the extent to which the presence or absence of STEs affects the overall indicators of the impact of the tax system on financial incentives to invest in skills, the Marginal Effective Tax Rate on Skills (METR) and the Average Effective Tax Rate on Skills (AETR).

Figure 5.4 shows the BEI for the typical stylised tertiary education student. With STEs, the data are the same as those presented in Chapter 4. Also shown are values of the BEI in the absence of STEs. In most OECD countries the BEI is higher in a world without STEs than in a world with STEs. This suggests that in the OECD countries considered, STEs reduce the amount of earnings needed to break even on a skills investment over time for a hypothetical student. This is only for a specific educational context (the 17-year old university student as discussed above) and for a specific set of assumptions about educational costs and scholarship income levels. As the value of STEs is closely associated with the values of these educational costs and scholarship income levels, these results reflect these assumptions.

Figures 5.5 and 5.6 analyse the changes in the METR and AETR when STEs are not taken into account. While Figure 5.4 showed the impact of STEs on overall incentives to invest in education or not, Figures 5.5 and 5.6 show whether METRs and AETRs are impacted by STEs. The extent to which METRs and AETRs are impacts by STEs measures whether the impact of the tax system on financial incentives to invest in skills occurs mainly through the impact of STEs or it occurs through other tax factors such as the taxation of returns to education.

Figure 5.5 shows that the STEs do reduce the METR on skills for most of the countries considered, although modestly in most countries. For example, removing Norway’s tax exemption for scholarship income raises the METR on skills from -5.6% to -.85%. Similarly removing Belgium’s tax deduction for student income and tax exemption for scholarship income raises the METR on skills from 3.2% to 8.3%. By contrast, the absence of STEs (and income taxes at 70% of the average wage) in Chile means that the METR on skills is 0% in both cases. On average in the sample of countries considered, METRs for tertiary education students who would earn 70% of the average wage in the absence of training are 2.9% with STEs (as was discussed in Chapter 4) and 5.1% without STEs.

Figure 5.6 shows that the impact of STEs on AETRs is even more modest. The average AETR in the sample of countries considered for tertiary education students who would earn 70% of the average wage in the absence of training is 19.1% with STEs (as was discussed in Chapter 4) and 20.5% without STEs. As discussed in Chapters 3 and 4, the tax rate on an average skill investment is larger than on a marginal skill investment, as the larger-than-marginal return on an average skill investment is usually taxed away by progressive taxation.

A regression analysis of the STEs on these three key dependent variables suggests that the impact of STEs on incentives to invest in skills is indeed modest (see Box 5.1). However, this does not mean that the tax system overall has no impact on incentives to invest in skills. Specifically, the extent to which the tax system offsets foregone earnings, or taxes away the returns to skills was shown to have a larger impact on the overall financial incentives to invest in skills (as measured by the BEI) than STEs.

Box 5.1. The Drivers of the Tax and Skills Statistics

This box examines the drivers of the key indicators of the tax and skills system outlined in Chapter 4, the METR on Skills, the AETR on Skills, and the BEI on Skills. As discussed in Chapters 3 and 4, incentives to invest in skills depend on a wide variety of factors; lost earnings, the potential returns to skills, the amount of scholarship income available, tuition fees and non-pecuniary costs. The tax system impacts financial incentives to invest in skills in a variety of ways; reduced earnings during study lead to reduced taxation; increased earnings after study may lead to increased taxation; the costs of a skills investment can be set against tax liability. This box analyses how these various factors impact the key indicators developed in this study.

Approach

The approach taken is to regress key dependent variables on the three outcomes of interest: the METR, the AETR and the BEI. While regression techniques are used to analyse the results presented in this study in more detail, the data used in the analysis is not the observational data that is typically analysed using these techniques. Because the data is not observational data, but rather manufactured data as part of the indicator-building process, the results serve to illustrate the key moving parts of each indicator, but do not make claims about the deeper relationships between financial incentives to invest in skills and individuals’ responses to these incentives.

The key variables used are the tax rate on foregone earnings (TFE) and the tax rate on the earnings increment that individuals earn after education (TEI). A variable for STEs is also included (STE). This variable is expressed as the difference between the take-home income of a student who works while studying with and without STEs (normalised to thousands of 2011 PPP USD for all countries). Student income before education is also included to measure the opportunity costs of education (INCb). The number of years of education is also included (Years). A sample regression equation is presented below:

Each i is a particular person or education scenario, and each j is a country. The variable T_FE_ij is the tax rate on foregone earnings for person i in country j. The variable T_EI_ij is the tax rate on an earnings increment for person i in country j. STE_ij is the value of STEs accruing to person i in country j. The β coefficients measure the impact of a given explanatory variable on the outcome variable Y_ij. Country fixed effects are included throughout.

Results

Column 1 shows the impact of various aspects of the tax system on the BEI. As discussed in Chapters 3 and 4, this shows the extent to which gross earnings must rise after a skills investment is made in order to break even on that skills investment. The coefficient on the TFE variable shows that higher taxes on foregone earnings decrease the costs of investing in skills and so reduce the earnings increment needed for a worker’s investment to break even. For example, Column 1 shows that when the taxes on foregone earnings increase by one percentage point (such as from 18% to 19%), the BEI decreases by 0.19 percentage points.

This is a key means by which the tax system impacts on financial incentives to invest in skills. When workers engage in skills investments, their incomes fall. This is less the case when taxes are present than when they are not; the tax system offsets this cost component of skills. These indirect costs of skills investments including the impact of the tax system on foregone earnings seldom feature in the literature on ways to incentivise skills investments. This analysis shows that the BEI of a skills investment is quite sensitive to personal income taxes which reduce foregone earnings during periods of investment.

Converse results are obtained for the effects of the TEI on the BEI. The regression in Column 1 shows that the TEI has a comparatively higher, but negative impact on the incentives to invest in skills: an increase in TEI increases the BEI. The impact of STEs is also noticeable – a USD 1 000 increase in the extent to which the tax system offsets the direct costs of skills is associated with a reduction in the BEI by .5%.

Table 5.1. Regression results
Dependent Variable	(1)	(2)	(3)
Dependent Variable	BEI	METR	AETR
Taxes on Foregone Earnings (FE)	-0.192***	-1.357***	0.017
Taxes on Foregone Earnings (FE)	-0.006	-0.02	-0.012
Taxes on the Earnings Increment (EI)	0.227***	1.500***	0.461***
Taxes on the Earnings Increment (EI)	-0.006	-0.018	-0.011
Skills Tax Expenditures	-0.519***	-6.572***	-1.310***
Skills Tax Expenditures	-0.092	-0.296	-0.176
Income Before Education	0.366***	1.401***	1.887***
Income Before Education	-0.064	-0.206	-0.123
Years of Education	3.807***	-0.418***	-2.133***
Years of Education	-0.024	-0.079	-0.047
Sample Size	11 880	11 880	11 880
Adjusted R-Squared	0.743	0.426	0.508
Note: Data incorporate tax deductions and tax credits for direct costs, tax exemptions for scholarship income, and reduced taxes on student wage income. The results do not incorporate STEs that subsidise parental spending on education or that subsidise firm spending on education. Stars indicate for the statistical significance of the coefficients. * means that the variable is significant at the 1% level, at the 5% level and * at the 10% level.

The logic behind the key levers in the tax system in Column 1 is clear. The tax system impacts financial incentives to invest in skills in two important ways: it offsets costs both through tax expenditures and through reducing foregone earnings and it reduces the returns through the taxation of higher post-education earnings. These findings are further illustrated in Column 2, which shows similar results for the three key tax variables on the METR. Higher TFEs mean lower BEIs and METRs, higher TEIs mean higher BEIs and METRs. More generous STEs raise incentives to invest in skills for marginal students.

These similar coefficients highlight that the BEI and the METR are closely linked – the BEI measures financial incentives to invest in skills for a marginal student, the METR measures the extent to which the tax system affects these incentives. Figure 5.7 confirms the intuition regarding the overall impact of the tax system. This figure shows a positive correlation between the METR and the BEI. Each data point in the graph is a country-observation; multiple observations per country are different age and income levels in each country. The METR presented in the graph includes not only both tax rates on foregone earnings and on the earnings increment but also specific STEs. There is a strong positive correlation between the two variables. Higher taxes, as evidenced by the METR, impact the individual’s incentives to invest in skills.

The relationship between the METR, the TFE and the TEI is further illustrated in Figure 5.8. The horizontal and vertical axes show the TFE and TEI respectively. The strong positive correlation between the TFE and the TEI can be seen in the figure: on average, when TFEs are high, TEIs will be high as well. Figure 5.8 shows that the PIT systems of the countries included in the study are progressive on average as TEIs are typically higher than the corresponding TFEs. This progressivity effect results in higher METRs. In fact METRs are increasing in the difference between the TEI and the TFE. METRs are lower in cases where TFEs are more similar to TEIs, such as in the case of a proportional tax system.

The relationship between these two constituent taxes and the METR can be seen by examining the colours in Figure 5.8. Darker colours represent higher METRs. METRs are darker in the top left of the figure where tax rates on the earnings increment are high, and tax rates on foregone earnings are low. These are two ways in which the tax system can incentivise investment in skills: by reducing the costs and increasing their returns.

Column 3 of Table 5.1 shows the same variables regressed on the AETR. In this case, the effects of the TEI are the same as in the marginal case however the effect of foregone earnings is not statistically significantly different from zero. This means that for the average student, the extent to which the tax system reduces the costs of skills through reduced taxation of foregone earnings is not a significant driver of the overall impact of the tax system on incentives to invest in skills. The way the tax system taxes away the returns to skills is an important driver, but the extent to which it reduces the costs of skills is not.

These results can be understood by considering the components of the METR and AETR. A marginal student is just breaking even on a skills investment; therefore their costs are high compared to their returns. An average student, by contrast, receives much higher returns than costs. The comparative weight of costs and returns is mirrored in the impacts of the tax system – the returns are higher in the average case, the tax rate on these returns has a greater weight in the AETR. By contrast, the TFE matters more for the marginal student, where the ratio of returns to costs for the student is lower. For the average student with fairly high returns to skills, the impact of the tax system on incentives to invest in skills through taxing foregone earnings is overshadowed by the impact through taxing away returns. The different effects are summarised in Table 5.2.

These results highlight key differences between average and marginal students in terms of how the tax system affects their decision to invest in skills. For marginal students, the returns to skills are modest compared to the costs. Therefore the tax treatment of the costs matters more for the overall METR than the tax treatment of returns for these students. By contrast, for a student earning a greater-than-marginal return, the tax treatment of costs may not matter as much as the tax treatment of returns in determining the tax systems overall impact on that student’s decision to invest in skills.

This in turn has important implications for how the tax system can be designed to encourage skills investment. For the average student, a tertiary skills investment yields a significant return (Heckman and Jacobs, 2010). To encourage students to make these investments, or to attract and retain new students into a country to make these investments, reducing the tax rate on the return to skills is an important policy lever – which may mean reducing the level or progressivity of the income tax system.

However, for low-return skills investments, lowering the tax rate on the return to skills may have little impact. Low-return skills investments may include skills investments in declining sectors of the economy, or skills investments by students who have low aptitude in a given area or low innate skills. To encourage these kinds of skills investments, reducing the progressivity of the labour income tax system may have little impact on skills investments. Action on the costs side of skills investment is what is required.

Table 5.2. Effects of income on the AETRs and METRs on average in the tax systems of 30 OECD countries
	Average Student	Marginal Student
	Impact of Tax System
Ratio of Earnings to Costs	Higher	Lower
Weight of Taxes on Costs	Lower	Higher
Weight of Taxes on Returns	Higher	Lower
Impact of TEI	Strong	Strong
Impact of TFE	Weaker	Strong

Regression analysis in Box 5.1 also suggests that STEs also had very little impact on the AETR. Average skills investments are usually profitable: the returns are much larger than the costs. This means that the fact that the tax system taxes away the returns to skills matters much more overall than the impact of the tax system on the costs of skills investments.

STEs are more impactful for marginal and low-return skills investments. The impact of the tax system on the costs of STEs is a much larger part of the overall skills investment calculus for skills investments where the costs are higher than the returns. In the modelled cases of a marginal skills investment where the student is just breaking even, the costs are the same as the returns. In these cases, the impact of STEs on the overall skills investment calculus is more significant.

Empirical analyses of impact of tax incentives for tertiary education

The above analysis has sought to quantify the overall size of STEs in the sample of countries considered, as well as to assess their impact on the key indicators developed in this study. The overall impact of STEs is modest for a typical student, though it may be larger for students in receipt of larger amounts of scholarship income, with larger amounts of tuition expenses, or with larger amounts of student earned income. This section provides some evidence of the impact of STEs on the tertiary enrolment and completion outcomes for students.

The country that uses the tax system to encourage skills investment most is the United States. Details of the STEs in the United States are discussed in more detail in Annex D of this study. Turner (2011) finds the introduction of tax-based student aid increases full-time enrolment in the first two years of tertiary education. However, recent empirical studies suggest that expansion of the key tax credits for education expenses used in the United States – the expansion of the American Opportunity Tax Credit over its predecessor, the Hope Tax Credit has no statistically significant effect on the tertiary enrolment decisions of students. Bulman and Hoxby (2015) find, using two separate identification strategies including regression kink and simulated instruments, that the credit does not increase enrolment. This is not to say that the credit provides no financial benefit to certain students who receive them. Rather, the results suggest that near the phase-out portion of the schedule that the reduction in credit amount does not impact enrolment, and also that the expansion of the credits does not induce students to enroll. Similar effects are found in other research for the tax deduction for post-secondary education (Hoxby and Bulman, 2015). Though only based on one country with comparatively high tuition fees, these results suggest that two key kinds of STEs used in the OECD – tax credits for education expenses and tax deductions for education expenses – may be limited in their impacts on student outcomes.

In a further US study based on the enactment of the tax credit programs in 1998, and from the increase in the value of the Lifetime Learning Tax Credit and introduction of the Tuition Deduction between the 1999-2000 and 2003-04 school years, Turner (2012) finds that for a subset of four-year institutions, that the schools reduced aid to students in response to the tax subsidies. By examining tertiary tuition prices, he finds that much of the increases in financial resources provided to students through the tax cut were captured by tertiary institutions by decreases in school grant aid. It may be that these extra resources increased the quality of education available to those students who did attend university, but this is difficult to demonstrate. These results suggest crowding out of student aid provided by tertiary institutions in the United States: credits targeted at reducing student costs may in fact be a form of subsidy to educational institutions through the tax system. Though this research raises questions as to the efficacy of tax credits as a means to raise skills investments, it is possible that tuition fee responsiveness will be lower in non-US OECD countries, which have more heavily regulated tertiary sectors. This means that other countries may have more success in using these kinds of credits to reduce the costs for students.

Some modest effects for tax credits are found by other authors. Guzman (2013) finds that in the US tax credits for education caused more students to attend private for-profit colleges, which may suggest that the credits allow students to attend more expensive for-profit schools, though effects at the intensive margin may not exist. In a subsample of 30 and 40 year-olds, LaLumia (2012), finds that while the Lifetime Learning Tax Credit had no statistically significant effect on average, eligibility for an education tax subsidy was associated with an increase in the probability of tertiary attendance among “men whose 1998 educational attainment falls short of early-life educational expectations”. This conclusion supports the OECD data analysis presented above. While STEs targeted at average students may not have significant impacts as a means of increasing skills investments, STEs may have an impact among students with low potential returns to skills investments. Positive impacts of STEs on enrolment are also found by Bednar and Gicheva (2013), but with respect to graduate education instead of tertiary education.

The evidence on reducing taxes on student labour income also suggests that this policy lever may not be ideal for governments. Reduced social contributions on student wage income are a significant source of government support for students. However, there is also need for caution – increasing the amount of student part-time work may reduce credit constraints for students from low-income families, but may negatively impact educational outcomes for these students. Using data from a Swedish reform, Avdic and Gartell (2015) find that when financial aid policies changed to give students an incentive to work longer hours, some students worked more. This in turn negatively impacted the amount of time required to complete a degree for some students. While reducing the tax burden on student work may provide students with increased resources with which to invest in education, it may do so while depleting their ability to study and complete their degrees.

There are different reasons why STEs may be of limited use in encouraging skills investments at tertiary level. Cameron and Heckman (1999) argue that a wider variety of factors are behind the decision not to attend university in addition to financial factors. For low-income families, the non-pecuniary costs of attending a tertiary institution (lack of motivation, lack of success in secondary education, lack of family role models, lack of information) may constitute an additional burden to tertiary attendance. This means that attempting to increase tertiary enrolment by using STEs without addressing other obstacles may prove ineffective. Similar conclusions are discussed in Cunha et al. (2006) and in Heckman and Jacobs (2010). They find that the overall education decision comes with both pecuniary costs (tuition fees, lost earnings) and non-pecuniary costs (study effort, lost leisure time). In such situations where pecuniary costs are heavily subsidised, they may crowd out non-pecuniary costs such as study effort. In such situations, they argue “high subsidies on education may then go hand in hand with long study durations, high drop-out rates and low student performance” (Heckman and Jacobs, 2010).

The evidence suggests that the existing STEs are of limited effectiveness in increasing the amount of skills investments at tertiary level. In addition, these policies may have unintended consequences such as increased tuition fees by universities in response to tax credits, or reduced student effort or longer student working hours which can increase time-to-completion of degrees. Dynarski et. al. (2015) point to many design flaws with the system of STEs in the United States, which limit their effectiveness.

Distributional impact of tax incentives for tertiary education

A key component in the discussion of tax and skills policy is the distributional impacts of STEs. As discussed in Chapter 1 of this study, skills policies are a key inclusive growth policy measure. Increasing the human capital for those whose existing levels are low has many beneficial effects; it raises wages, productivity and labour market attachment. By decreasing wage inequality, it reduces the need to implement distortionary taxes and pay benefits to reduce disposable income disparities. Assessing tax and skills policies from the perspective of those with low skills and low incomes is crucial for assessing these policies’ effectiveness.

The distributional aspects of tax and skills policies have two related but distinct dimensions. Tax and skills policies can be addressed towards those with low incomes but for whom skills investments will be very profitable; those with low incomes but strong innate skills. Skills investments yielding high returns are very beneficial from the perspective of addressing income inequality. But another distributional margin is also important: that between low-return skills investments and high-return skills investments. Box 5.1 has described how STEs have larger effects on a marginal skills investment than on an average skills investment. Policies that reduce skills costs may benefit low-return skills investments more than policies that reduce the taxation of those returns.

These policies – that may benefit a marginal skills investment more than an average skills investment – may also yield positive distributional outcomes. In providing benefits to skills investments with low returns, policies targeted at reducing skills costs are beneficial in that they may provide larger benefits to those with low innate skills, who may receive lower benefits from further skills investments than those with higher skills. In providing benefits to those making these investments, these policies may simply offer a form of insurance against the unlucky: those who have made a skills investment yielding positive expected net returns to investment but negative actual returns. This means that these policies are a potential targeted measure towards those whose lifetime income is expected to be low.

Figure 5.9 assesses STEs in the sample of countries by income before a skills investment. It shows the impact of removing all STEs including SSCs on the BEI – the same data as Figure 5.4 – but varies the results over the income distribution. It shows, in part, whether these STEs raise financial incentives to invest in skills more for those on lower incomes or those on higher incomes. In other words, analysing BEIs with and without STEs over the income level is the same as analysing whether these STEs are progressive with respect to income before education.

Figure 5.9 shows that in the OECD these STEs are usually progressive; though not always. For some countries, STEs are modest in size (for example, in Greece, Turkey, and Switzerland). In these countries, their impact is not particularly large for university students, either at the higher or lower end of the income distribution. However, in other countries, STEs exist that provide proportionally larger reductions in the necessary breakeven returns to skills to those on low incomes than to those on higher incomes. These BEI gaps are likely a result of provisions that exempt some student wage income from taxes or SSCs (e.g. Belgium). There are also some countries for which the proportional reduction in the BEI remains broadly constant across the income levels considered (e.g. for Norway and Sweden).

A key factor in the degree of progressivity of STEs is that many of them come in the form of allowances or credits, so that resources spent on skills investments can be claimed against taxable income or tax payable. However this means that the credits provide value when a taxpayer is making a skills investment and has some taxable income. Students with no taxable income will in general receive less benefit from tax allowances and credits than other students, unless the credit is fully refundable. This may make STEs in the form of deductions and credits less progressive than other forms of STEs. This is discussed further in Box 5.2.

Box 5.2. Carry-forwards of Tax Credits in the OECD

Six countries in the sample considered in this study provide tax credits for tertiary education. A drawback of tax credits for education is that they may not benefit students from low-income families, who may not have sufficient tax liabilities to benefit fully from the credit. This is also the case with respect to tax allowances. Details of the credits as they existed in 2011 are provided in Table 5.3. Different countries deal with this issue in different ways. Some countries (such as the United States) make the credit partially refundable, to ensure that low-income households benefit from the credit. Other countries such as Canada allow the value of the credit to be transferred to other members of the household, though this may not be effective if other household members also have low incomes. Finally, some countries including Canada and Israel allow a student to carry forward the value of the credit to future years. A full carry forward of STEs is further analysed by Stantcheva (2015).

Table 5.3. Tax credits for university education in the OECD
	Amount (for single households)	Minimum Value	Maximum Value	Refundable	Carry Forward
Canada	15% of eligible expenses	None	None	No	Yes
Israel	ILS 2 508	NA (fixed amount)	NA (fixed amount)	No	Yes
Ireland	20% of eligible expenses	EUR 2 0001	EUR 7 000	No	No
Italy	19% of eligible expenses	None	None	No	No
Portugal	30% of eligible expenses	None	EUR 475	No	No
United States	100% of first USD 2 000 of expenses less related scholarships and 25% of next USD 2 000 of expenses less related scholarships.	None	USD 2 500	Yes	No
← 1. In Ireland, this value was EUR 3 000 in 2016.

Canada’s Tuition Tax Credit provided a 15% non-refundable tax credit of the costs of tuition in 2011 (the year estimated in the model underlying this study). There was no limit on the amount of tuition that could be claimed, but claims must have been higher than CAD 100. In addition, the credit could be carried forward against future tax liability for students who do not exhaust the value of the credit in the first year, or transferred to a supporting individual. In addition to the AOTC (American Opportunity Tax Credit) presented in Table 5.3, the United States has the Lifetime Learning Tax Credit that is not refundable. The AOTC also phases out at USD 80 000 (see Annex D).

Figure 5.10 shows the value of the refundability of the tax credit for students compared to a no-refundability scenario at various income levels, for various costs of education. Students at low income levels gain more by being able to carry forward the value of the credit. Students with higher fees benefit more still. This shows that making tax credits refundable specifically benefits those with low resources during education and higher costs of education.

The richest academic studies assessing the distributional impact of STEs are based on the US experience (Bulman and Hoxby, 2015; Hoxby, 1998; Turner, 2012). The US tax credits are partially refundable, and also phase-out at higher income levels (see Annex D for further details). Bulman & Hoxby argue convincingly that the tax credits for education provide greatest benefits for middle classes, but do so inefficiently. Poorer households tend not to have sufficient tax liability to benefit from the credits, or are in receipt of direct grants to support skills investments. Higher income households tend to pass the phase-out period of the credit, and so receive little benefit. In the United States these tax benefits are focused on those at the middle of the income distribution.

One approach to increase the value of tax allowances or credits as a form of STEs has been to allow the value of these credits or allowances to be carried forward to future years. Stantcheva (2015) argues that a system of tax deductions with full carry-forward can ensure that those on low incomes or whose skill investments yield low returns benefit from STEs as much as higher income individuals or those whose skills investments yield high returns, providing similar benefits to a system of income-contingent loans. A partial carry-forward system has been implemented in Canada; the Canadian credit can be transferred (up to a dollar limit) to a supporting individual such as a parent, and/or carried forward for use in subsequent years (see Box 5.2). While this approach may not provide the full benefits of a credit to low-income households in net present value terms, it improves the equity of such credits relative to a non-refundable credit without a carry-forward provision.

Design considerations of tax incentives for tertiary education

The above discussion has highlighted potential issues with STEs as a means of increasing the level of skills investments. These instruments may be biased against individuals with low skills and households with low incomes. There is mixed evidence about the extent to which STEs achieve their goals when it comes to tertiary education. Nonetheless, these tools remain an important part of the policy mix in many OECD countries. With this in mind, this section discusses several key recommendations that should be considered when tax policy makers are designing STEs for tertiary education so that their impact can be maximised.

The literature in the area of tertiary-based STEs has highlighted that the complexity of STEs can reduce their effectiveness. OECD countries provide scholarships, grants, student loans, tax deductions and tax credits to students and parents, as well as tax or SSC reductions for student work. Many STEs interact with each other; in some countries it is not possible to benefit from some provisions with some others. Eligibility criteria for loans, deductions, credits and student grants can be designed in an uncoordinated way. Interactions between different aspects of the tax and student support system must be designed in a holistic way, especially where these provisions are administered and designed by different ministries or agencies. Where eligibility for one program removes eligibility for another, the effective marginal rates of support can be very far from policy makers’ original intentions.

A second consideration stemming from overlapping programs is that these programs are complex and difficult to understand for taxpayers. This can lead taxpayers to make poor choices as to which programs are best for them. This complexity has been argued to be a key component as to why some of these policy measures are not effective (Davis, 2002; Dynarski and Scott-Clayton, 2015). In one US-based study, one in four taxpayers were found to mis-claim credits or deductions or not to choose the method of support that would be most beneficial for them (Turner, 2011). This study also found that low-income taxpayers are also more likely to mis-claim tax support for skills investments. Simplicity of tax and other forms of support is a key means of both ensuring the effectiveness and the equity of this form of support.

A third issue that can be important in designing the tax benefits for skills is that many low-income families may be credit constrained with respect to human capital investments (Carneiro and Heckman, 2002). Grants, scholarships, and tax benefits may provide an important means of reducing these constraints. However, for many households, the costs of human capital investments are incurred up-front, while the tax benefits can only be realised later in the tax year. For some families, this may nullify the usefulness of STEs (Dynarski and Scott-Clayton, 2015). This is another reason why direct support in the form of grants made available to individuals and families when skills investments are made may be a superior policy tool when compared to STEs which are available later in the tax year.

5.3. Non-tax financial incentives for tertiary education

Section 5.2 of this chapter discussed the ways in which the STEs can impact incentives to invest in skills for tertiary education. This section focuses on non-tax provisions that can provide financial incentives for students to invest in skills, such as reduced tuition fees, scholarships, grants and subsidised loans. The model developed in this study also accounts for the wide variety of non-tax financial measures provided to support educational investments by governments. This section considers the impact of these provisions from an efficiency and equity perspective. Data on this support has been taken from the OECD’s Education at a Glance publication (OECD, 2014). Details of the assumptions made around these forms of support are provided in Chapter 3 of this study.

Size of non-tax financial incentives for tertiary education

Figure 5.11 shows the overall components of government spending on skills investments, incorporating three key aspects of government spending: direct spending on tertiary institutions, scholarship spending, and STEs. Data are expressed in 2011 PPP USD, and the graph uses the measure of STEs based on both personal income taxes and SSCs. Direct spending is larger than the two other categories combined (an estimated USD 10 227 per student per year), while spending via scholarship income is USD 1 492 per student per year. Spending through the tax system is estimated to be USD 860 per student per year when SSC-based STEs are incorporated. In every country, direct spending is larger than estimated scholarship income (though this may vary per student) in some countries the amount of support offered through the tax system is greater than the amount of scholarship income (Estonia, Iceland, Poland, Slovak Republic, Slovenia and Spain). These results depend on the amount of scholarship income received, and the amount of student income modelled. Nonetheless, non-tax support is far larger than tax support for tertiary education.

Figure 5.12 shows the impact of the withdrawal of scholarship income on BEIs in the countries included in the study. As in Section 5.2, the data used incorporates the impact of SSCs in the analysis. The increase in BEIs is substantial in almost every country. BEIs are estimated to be 15% on average with scholarship and grant income considered, and 19% with the level of all scholarship and grant income set to zero. Removing scholarship and grant support raises the estimated BEI by 22.6%.

The literature on the effect of non-tax incentives for tertiary education has broadly more positive findings of the impact of these measures on educational outcomes, especially when compared with tax policy levers. Direct scholarship and grant support has several theoretical advantages over tax-based aid; it is more likely to be immediately available at the time tuition fees are due, administrative barriers are fewer, and receipt is not limited by taxable income. Financial aid has been found to have positive impacts on tertiary enrolment in the United States (see Kane (2006) for a review), and in Germany (Baumgartner and Steiner, 2004). Denning (2016) finds that financial aid improves time-to-completion of degrees.

A large number of studies also suggest that higher tuition fees can negatively impact college enrolment in a variety of countries (Dynarski, 2005; Kroth, 2015). Studies have also found positive relationships between tuition fees and time-to-completion of degrees (Bruckmeier et al., 2013). It is not clear how the impact of tuition fees and scholarship income differs, or how the extent of the impact of tuition fees is dependent on the amount of scholarship income or vice versa. Nonetheless the literature does suggest that non-tax financial incentives may be more impactful than tax-based incentives.

Distributional impact of non-tax financial incentives for tertiary education

Figure 5.13 shows the impact of scholarship income on the BEI at various points in the income distribution, by showing BEIs with and without scholarship income in the model. The figure shows that in almost every country, removing scholarship income would increase the BEI, often substantially. It is instructive to compare Figure 5.13 to Figure 5.9, which shows similar results, but removing STEs instead of scholarship income. The impact of removing scholarships and grants is worse for BEIs on average than removing all STEs. This suggests that scholarships and grants are at present a larger form of government support for skills than STEs. Setting all grants and scholarships to zero would increase BEIs for all students in most countries in the OECD, but BEIs would increase more for low-income students based on this model. This suggests that, due to the features mentioned above, scholarship and grant income is progressive in OECD countries, and more so than STEs.

This conclusion is borne out by the literature on the impact of scholarships, grants and financial aid in OECD countries. This literature has long concluded that higher tuition levels reduce enrolment (Leslie and Brinkman, 1987). Kane (2006) reviews substantial evidence that demonstrates that there are greater responses to the tuition fee differences amongst lower-income households, with a USD 1 000 reduction in the private costs of education having an estimated impact that is twice as large for a household from the lowest income quartile compared to a household from the highest income quartile. The literature varies on whether these differences across incomes are due to the increased likelihood of credit constraints amongst lower households or other factors. In either case, the evidence suggests that spending through tuition reductions or increases in scholarship and grant spending currently has positive distributional and efficiency consequences; direct spending in the form of fee reductions or scholarship and grant provision is more effective at raising enrolment and completion rates compared to tax-based subsidies, it is also better targeted towards those on lower incomes, and finally it is more likely to raise the enrolment rates of those on lower incomes compared to STEs, which are often less beneficial to those on lower incomes due to a lack of taxable income or to the administrative burden of applying for STEs.

Debt-based support for tertiary education

While reducing tuition fees and increasing scholarship and grant support is important, students in OECD countries still bear significant private costs of education in the current policy environment. Figure 5.14 outlines the up-front cost components of a skills investment for the stylised 17-year old student.6 Data are presented in 2011 PPP USD, exclusive of SSCs. This data is similar to the BEI, with the exception that the BEI also accounts for extra taxes a student must earn after education as their earnings rise. The data in Figure 5.14 focuses solely on the net costs of education in terms of net fees and government sponsorship, as well as net lost earnings. The returns to skills are not examined.

The figure demonstrates that foregone earnings are the largest cost component for all OECD countries, usually substantially so. While direct private costs borne by students such as tuition fees can be large, in every case they are smaller than foregone earnings. Similarly, foregone taxes are, in the majority of OECD countries, the most significant means by which the government offsets the costs of skills investments. Foregone taxes are larger than both STEs and scholarships and grants for students.

The data suggests that even in the presence of the current set of government policies to support tertiary skills investments in OECD countries, the private costs are still significant. As has been shown in Section 4.2 of the study, however, the returns to skills are larger than the costs for a typical student in OECD countries; skills investments are profitable, both from the student’s and the government’s perspective. This raises the question as to why skills investments that yield positive returns are not undertaken.

A key explanation as to why skills markets may fail leaving worthwhile skills investments not undertaken is credit constraints on the part of students. Unlike physical capital, human capital or future labour supply cannot be offered as collateral on a loan to fund an investment (Cameron and Heckman, 1999). This means that loans may not be provided to profitable skills investments as recipients would rationally choose not to repay. This breakdown in private markets for skills investments is a key motivation for government intervention in the human capital market. In the presence of credit constraints there may be underinvestment in human capital, concentrated on those with low incomes. There is evidence that nearly 10% of the US population underinvests in human capital due to credit constraints (Carneiro and Heckman, 2002; Lochner and Monge-Naranjo, 2002). There is also evidence that these credit constraints persist throughout life (Popov, 2014).

To address these issues, a significant component of government education support also comes in the form of subsidised or guaranteed student loans. Chapter 4, Section 4.3 of this study has outlined how assuming that students finance their investments through debt instead of through savings can have significant effects on the affordability of education depending on the interest rate available to students. Increased or reduced interest on student loans can dramatically alter the attractiveness of human capital investments compared to other investments. While the model in this study has examined the impact of debt-financing and equity financing of skills investments, it has not considered the possibility that some students may not be able to access debt financing at all due to various forms of market failure.

The literature on student loans has long argued that government-sponsored income-contingent student loans can be a beneficial way to increase student enrolment in the third level education while sharing risk across students (Chapman, 1997; Findeisen and Sachs, 2014; Jacobs, 2002). Having the government guarantee loans as well as using the government’s power to tax as a means to ensure repayments can both reduce the burden on the lender, while ensuring that low-income students retain access to finance (see Lochner & Monge-naranjo (2014) for a review). The introduction of income contingent student loans in Australia in 1989 has been successful in expanding access to higher education (see Box 5.3, and the discussion in Chapman (1997)).

Income contingent loans can be an attractive policy option because the income-contingent nature of the loan means that redistribution occurs from those whose skills investments yield high returns (who repay their loans) to those whose skill investments do not yield high returns (who may repay only in part or not at all). The discussion in Chapter 4 shows that while on average, a skills investment will pay for itself, many students make skills investments that may just break even or may not breakeven at all. This can be the case even if the investment did have positive expected returns. Loans that feature income contingent repayment offer insurance for the student against these risks.

The discussion in Box 5.1 of this chapter has highlighted that the progressivity of tax and skills policies need to be considered along two dimensions. Tax and skills policies can redistribute between high and low income students making skills investments. These policies can also redistribute between those whose skills investments yield high and low returns. In other words the tax system can redistribute across the income distribution ex ante of skills investments, or ex post. Income contingent loans do both. By ensuring access to education finance amongst the credit constrained, they redistribute ex ante. This is because low-income students are most likely to be credit constrained ex ante. In addition, by making repayment of the loans conditional on earnings after a skills investment, they redistribute ex post – those skills investments yielding low returns are taxed at lower rates.

A system of loans instead of graduate taxes can reduce potential adverse selection issues that may result from graduate taxes. Graduate taxes are similar to student loans in that the returns to student income are spread between students and governments through means other than through taxes. However unlike income-contingent loans where the returns to education after costs have been repaid accrue to students, graduate taxes see the returns shared between government and students even after costs have been repaid. With graduate taxes, students making high-value education investments may choose to invest elsewhere or move in response to taxes, relative to a system of loans where high returns will accrue to students (after loans have been repaid (Dynarski, 2015).

Finally, loans can facilitate higher tuition fees, potentially ensuring that the government receives an acceptable rate of return on profitable skills investments without reducing access for those students who are credit-constrained, risk- or debt-averse, or making risky investments. By raising tuition fees, and at the same time expanding access to income contingent loans, the governments returns to education can be maintained, access for low-income students can be maintained, and risky-yet-worthwhile investment can be maintained (Blöndal et al., 2002).

Box 5.3. The Australian system of income-contingent loans

Income-contingent loans are available to Australian students enrolling in eligible university courses. Repayments are connected to a graduate’s ability to pay, not the amount of the loan, or its age. If a graduate loses their job or takes time out from work, no repayments are required where their income is below the repayment threshold. The repayment schedule for 2016-2017 is provided in Table 5.4.

Table 5.4. Repayment schedules for HELP loans, 2016-2017
Repayment income	% of Loan to be Repaid
Below AUD 47 196	0
AUD 47 196 - 52 572	4.00%
AUD 52 572.- 57 947	4.50%
AUD 57948 - 60 993	5.00%
AUD 60994 - 65 563	5.50%
AUD 65654 - 71 006	6.00%
AUD 71 007 - 74 743	6.50%
AUD 74 744 - 82 253	7.00%
AUD 82 254 - 87 649	7.50%
AUD 87 650 and above	8.00%

Figure 5.15 shows the impact of Australia’s Income Contingent Loan Scheme for students by mapping the BEI for various marginal students. These results are presented based on the case of the stylised university student borrowing at a 3% real interest rate (Sections 4.2 to 4.3 provide further details of the assumptions behind this case). The results clearly show that the income-contingent loan scheme substantially reduces the necessary BEI for low-income marginal students relative to the scenario where education is financed wholly with a students retained earnings.

5.4. Tax Incentives for mid-career training

Importance of lifelong learning has been long acknowledged by policy makers. Due to rapid technological change, globalisation, and increased longevity, increasing numbers of older workers face challenges in keeping their skills up-to-date in the face of changing work practices. This means that maintaining human capital levels throughout workers’ careers is an important part of skills policy.

Maintaining and increasing human capital can be challenging, particularly for older workers. Employers may be less likely to train workers who may soon leave the labour market; workers’ propensity to undertake new training declines with age; research suggests that workers’ ability to learn new skills also declines with age (OECD, 2013). For many, making significant skills investments in later life requires time out of the labour market. This often means very high levels of foregone earnings for older workers, dis-incentivising lifelong learning. These comparatively higher costs of training older workers have led some scholars to argue that investments in human capital are best focused on younger workers (Heckman and Jacobs, 2010). However addressing challenges such as increased market inequality, stagnant wages, and demographic challenges will require improved human capital of older workers as well as younger workers, and so providing effective training for older workers remains a significant policy objective (McCall et al., 2016).

This section provides an assessment of the existing policies, particularly STEs, designed to encourage worker training. Based on the model detailed in Chapters 3 and 4 of this study, the section first outlines the value of these provisions for the hypothetical example of job-related mid-career training discussed in Chapter 4. These provisions are then discussed from the perspective of their effectiveness in encouraging skills investment, as well as from a distributional perspective. The section will discuss non-tax approaches to encouraging lifelong learning, and will end by briefly discussing some issues policy makers may need to consider when designing STEs for worker training.

Size of tax incentives for mid-career training

Figure 5.16 shows the value of STEs for the in-work training example outlined in Chapter 4, expressed as a percentage of the costs of training. This example is for a worker at the average wage who undertakes a short period of job-related training. In this instance, incentives through the SSC system are not incorporated. Eleven of the countries modelled have no STEs for mid-career training, and so the value of STEs is zero in the model. The remainder of the countries modelled have STEs, mainly in the form of tax deductions in the personal income tax system. The model for in-work training assumes that this training is “job-related” in the sense that the training is necessary for a workers’ job and thus becomes a tax deductible expense in many OECD countries. As will be discussed below, many OECD countries offer a tax deduction for training costs only where this training is related to a worker’s current employment.

Across the modelled countries, the average value of the costs of training that are offset through STEs for a worker at the average wage is 15.3% of the costs of training. Amongst the countries that do have STEs for worker training, the average value is 24.7% of the costs of training. Because the STEs for worker training tend to come in the form of tax allowances, the value of these tax allowances depends on the marginal rate of income tax in the country concerned. So countries such as Belgium with high tax rates on the average worker also see comparatively more valuable STEs in the model.

Many of the countries that have large STEs to encourage college education do not have similarly large STEs aimed at worker training. This is in part due to that fact that 28 of the 30 countries examined in this study have tax exemptions for scholarship income. It is assumed that those workers undertaking a period of in-work training do not receive any scholarship income, and so do not receive any tax benefit through STEs that come in the form of tax exemptions for scholarships. This reduces the size of STEs measured in several countries, notably Finland and Norway where scholarship income is tax-exempt and comparatively generous.

In addition to the support provided by the tax system to training through the personal income tax system as modelled in this study, significant support is also provided through the corporate income tax system. Details of the support provided to firms through the corporate income tax are outlined in Torres (2012). These provisions are not modelled in this study, but a study by (Bassanini et. al., 2007) suggested that three-quarters of all worker training was provided by employers in the EU. The costs of this firm-provided training are usually deductible from corporate taxable income. STEs in the personal and corporate income tax system should be designed in a coherent way. In the absence of STEs for job-related training in the personal income tax system, the tax system will induce businesses to pay for the job-related training for workers. Workers will request that their training be paid for by employers, as this will be deductible from the corporate income tax base. Moreover, these payments will not be included as a fringe benefit from the taxable personal income of the worker. Such a system will however add a distortion between the levels of tax support provided to workers in larger firms versus workers in smaller firms, the self-employed, and those not in employment.

A key issue highlighted by this literature is that, just as the government shares in the returns to human capital investment by workers, so too do firms share in the returns to human capital investment, by having more productive employees.7 However, as firms may not fully internalise the returns to training to the government and to broader society they may under-invest in training. This positive externality affects firms in the same manner as individuals who do not fully internalise the broader positive effects of skills investments. This under-investment can be exacerbated by fear of highly skilled employees being poached by other firms. These externalities provide a rationale for government intervention into the market for worker training.

Impact of tax incentives for mid-career training

Though STEs for mid-career training exist in many OECD countries, the evidence on their effectiveness in terms of encouraging worker training is limited.8 There are several reasons for this comparative lack of evidence. A large part of mid-career training is provided by employers, who may select high-performance employees into training programmes: this means that the causal effect of training (as opposed to simply being a high-performance employee) cannot be identified. Similar problems exist with studies examining training financed by individuals: talented individuals may be more likely to self-select into training, so comparing wages across individuals who do and do not invest in training will pick up effects that are partially driven by training, and partially by talent. In addition, on-the-job learning is difficult to measure, and many employers do not keep records of the amount of training their staff receive. These effects make it difficult to examine the returns to work-related training; and the impact and value-for-money of tax subsidies in this area.

Despite these difficulties, some studies evaluating the impact of STEs for mid-career training on training participation do exist. An early study by Holzer et. al. (1993) based on a grant program for firms in the United States, found that the receipt of training grants was associated with a significant, though “one-off”, increase in training hours. Another recent study used a regression discontinuity approach based on a Dutch provision that granted a training deduction for workers over 40 years of age. This study found an increase in training rates for workers over 40, however, this was driven by workers postponing training, not by an increase in training overall (Leuven and Oosterbeek, 2004). A further study by the same authors using kinks in the income tax schedule to identify the impacts of tax incentives on training participation found positive effects: a 10% increase in the tax deductibility increases the training rate by between 10% and 25%. (Leuven and Oosterbeek, 2006). A study based on Italian data also found positive effects of tax incentives on training participation (Brunello et al., 2012). So the evidence suggests that STEs for training do have an impact – in contrast to those at third level. However, more research is needed in this area to properly assess the size of the impact of different provisions, and to assess how the impact varies across countries, demographics, and kinds of training.

The existing evidence on the financial returns to worker training is also mixed. Observational studies tend to find positive impacts of worker training on productivity and on wages. For example, Brunello (2004), using survey data for Italian large enterprises finds that a 10 percent increase in the average number of hours of training per head increase productivity in his sample by 1.32 percent. Brunello et. al. (2012) find that one additional week of training increases monthly after-tax earnings by 1.4 percent.

By contrast, more recent studies involving random assignment or instrumental variables find smaller effects on wages or no effects at all. For example, Görlitz and Tamm (2016) find no effect of a randomly assigned voucher for training on wages and employment, though those who are trained are assigned more cognitive tasks. Overall, the literature suggests that while training may have an impact, it is comparatively small and dependent both on the kind of training, whether it is related to current work, and whether is supplied by the employer. The returns to training also vary across the population and between countries. There is some evidence that training low-skilled workers may yield higher returns than higher-skilled workers (Fouarge et. al. 2013; Schwerdt et. al. 2012). For older workers, one study found that training did not improve wages, but did improve employability (Brunello, 2007).

Distributional impact of tax incentives for mid-career training

A key concern when evaluating worker training is assessing the distributional impact of policies that affect training. This is in part due to the fact that worker training is often cited as a key policy to improve distributional outcomes in OECD countries (OECD, 2015b). Figure 5.18 shows the value of STEs for mid-career training as a percentage of the direct costs of training across the OECD as derived from the model used in this study. The absence of tax deductions for training as seen in Figure 5.16 is also clearly visible for the countries that do not have STEs for worker training. Also apparent is the increasing benefits of existing STEs for workers on higher incomes. Most STEs for worker training come in the form of tax deductions for the private costs of training. As with tax deductions for university training, these provisions may not benefit low-income taxpayers to the same extent as higher-income taxpayers, due to low-income taxpayers’ lack of tax liability. Unemployed or non-participating workers may benefit from active labour market programmes but these programmes may not be useful for those who are working but at low incomes (McCall et al., 2016). In addition, as discussed in Section 5.2 tax deductions for worker training also do not address liquidity constraints for low-income taxpayers stemming from the fact that the costs of training may accrue during the tax year, but the benefits of the tax deduction do not accrue until the end of the tax year.

On the corporate side, most mid-career training is financed by firms and not by workers themselves. However, this firm training may not be optimal from a social perspective. The literature suggests that those with low skills are less likely to be trained by their employers (Hansson, 2008). There is also evidence that women and older workers are also less likely to receive employer training: women are more likely to self-finance their worker training (Bassanini, et. al. 2007). So firm-sponsored training may have unintended distributional consequences.

Figure 5.18 compares the value of training for the stylised worker training example outlined in Chapter 4 of this study. The results for “job-related” training are the same as those in Figure 5.16 - incentives through the social contribution system are not incorporated. However, in this figure the results for job-related training are compared to those for non-job-related training. Many countries that allow training to be tax-deductible where it is job-related do not allow deductibility where the training is not job-related (including Australia, Austria, Belgium, Finland, Iceland, Israel, Mexico, and the United Kingdom). The average value of STEs for non-job-related training is 6.9%, compared to 15.3% for job-related training.

Restrictions of training eligible for tax deductions to “job-related” training can also potentially be problematic from a distributional perspective. Many OECD countries that have tax deductions for training require that the training be related to a worker’s current employment. This is designed to prevent inefficient subsidisation of skills spending that is consumption. However, these measures may also mean that workers in secure employment or in a fixed career may receive more benefits than workers who need to train to move careers, or who are in casual employment.

Finally, a key concern about financing worker training through tax deductibility in the personal and corporate income tax system is that these provisions may not encourage training participation in the SME sector (Müller and Behringer, 2012; Stone et al., 2008). The evidence shows that SMEs are less likely to invest in training than larger firms (Bassanini et al., 2007). This may be because SMEs are often less profitable than larger firms; they may also have cash-flow issues that means that even training that is eligible for tax deductions is deemed unfeasible. This means that employees of SMEs may receive an inefficiently low level of training compared to their counterparts in larger firms. As SMEs are a key source of job growth and innovation, low human capital levels can be very problematic for longer-term productivity.

Design considerations of tax incentives for mid-career training

The existing STEs in the personal income tax system do seem to play a role in encouraging workers to invest in skills. There are concerns about the distributional consequences of these measures. Both the economic literature and the model developed in this study suggest that STEs may provide larger benefits to men, to those on higher incomes, to those in secure jobs, to those working for larger firms, and to those with higher skills. By contrast, women, older workers, those with lower incomes, those in insecure employment, those working for SMEs, and those with lower skills may not receive the same benefits from STEs designed to encourage worker training.

In implementing STEs, design considerations are important in ensuring that STEs are effective in increasing skills investments. Many of the same considerations discussed with respect to STEs for tertiary education also apply to worker training throughout life. Complexity of STEs, as well as their interaction with other similar provisions, raises the burden on taxpayers who might want to receive certain STEs. Low-skill, low-income taxpayers are more likely to be negatively affected, either by not claiming or mis-claiming STEs to which they are entitled. Similar effects pertain with respect to the corporate tax system and SMEs. Smaller firms are less likely to take part in complex training programs, even if these programs are tax-deductible and are beneficial to a firm (Müller and Behringer, 2012).

Design of STEs should also consider deadweight losses. As with all tax expenditures, a key concern is the additionality of the effect of the tax expenditure: how much extra training STEs actually generate (OECD, 2010a). Poorly-designed tax expenditures may result in deadweight losses of up to 60% (European Centre for the Development of Vocational Training, 2009). Thresholds, for example, can lead to unintended effects. One study of a Dutch scheme gave an extra tax deduction to older workers (workers above 40), which had the effect of causing workers in their late 30s to postpone training in order to avail themselves of the tax incentive later. The overall effect of the measure on training was found to be minimal (Leuven and Oosterbeek, 2004).

Non-tax incentives for mid-career training

In addition to providing tax incentives for training, governments across the OECD use a wide variety of non-tax policy approaches to encourage worker training and lifelong learning. A detailed discussion of these policies is beyond the scope of this study, but these policies include active labour market programmes targeted towards the unemployed, vouchers to workers to choose their own training, and employment funds financed by firms that provide training. Detailed discussion of these policies is available elsewhere for Europe (European Centre for the Development of Vocational Training, 2009; European Commission, 2015) and more broadly (Bassanini et al., 2007; McCall et al., 2016).

Vouchers have been argued to be less administratively costly than STEs at the personal level (Messer and Wolter, 2009; Müller and Behringer, 2012). However, many of the distributional biases that occur with respect to STEs also obtain with respect to vouchers: those low-skilled workers most likely to benefit from them are also least likely to use them (Schwerdt et al., 2012). Deadweight losses also arise with respect to vouchers: increased government provision of vouchers may crowd out firm spending on skills, in part because firms do not fully internalise the benefits of investing in their workers (their workers may not remain with the firm) (European Centre for the Development of Vocational Training, 2009).

Another non-tax policy approach to worker training is communal training funds. These are funds that can be organised at the local, sectoral, or national level. Employers pay mandatory fees into these funds – either as a share of sales, payrolls, or profits – and then have access to the training programmes provided by these funds. Stone & Braidford (2008) argue that these funds may be more effective in encouraging worker training than tax deductions for training as firms, once having paid into a fund, will be more likely to use it. Moreover, as payments into these funds are usually compulsory, SMEs are also more likely to use them, addressing the issue of reduced SME participation in worker training (Stone et al., 2008). A potential issue with sectoral level funds is that firms – and their employees – in small or low-growth sectors in an economy may lose out through having access to a smaller fund compared to larger or high-growth areas of the economy. Similar effects could arise between wealthier regions and poorer regions if funds are set up on a regional basis. Maintaining a multiplicity of funds can result in political economy challenges if different funds are in states of varying levels of financial security.

5.5. Tax, skills and education finance

How does the tax system impact skills investments?

The preceding sections have discussed the effects of tax expenditures designed to encourage skills investments. The model developed in this study illustrates several key issues concerning STEs:

In most OECD countries STEs are modest in size, though they can be larger when social contributions are taken into account.
As a result, their impact on the overall financial incentive to invest in skills is also modest, particularly at tertiary level.
In the hypothetical cases outlined, STEs benefit workers on higher incomes more than workers on lower incomes.
In spite of this, STEs are a larger factor in the skills investment decisions of those making low-return skills investments than those who may see higher returns from their skills investments. This is because the costs of skills investments are a larger component of the overall financial decision to invest in skills than the returns for marginal students, so the way the tax system impacts those costs has a large impact on the overall effects of the tax system.

These arguments are supported by the findings from the academic literature. Though this literature is limited in many ways, some key conclusions are as follows:

Tax expenditures have limited effects on students’ decisions to enrol in tertiary education.
Tax expenditures for mid-career training do encourage training, but can come with significant deadweight losses.
Those on low incomes are less likely to benefit from themselves of STEs, both for college education and for worker training. STEs for training are often poorly designed: credits accrue to students at the wrong time of the tax year and provisions are overly complex. Training provided by employers can disadvantage certain workers due to employer bias.
The literature is broadly supportive of non-tax approaches to providing support for skills investments in the form of tertiary education, including scholarships, reduced tuition, and income-contingent loans. These approaches may be more beneficial for low-income students.
Support to students through income-contingent loans has been found to be particularly effective, in terms of ensuring access to education for low-income students, sharing the financial burden between government and students, distributing the risk of human capital investments, and balancing equity and efficiency considerations.

The largest impact of the tax system on skills investments does not come through STEs but through the broader income tax system. The regression analysis discussed in Box 5.1 has highlighted the impact of the tax system on financial incentives to invest in skills, which goes far beyond the impacts of STEs. The strong relationship between the METR on skills and the BEI is shown in Figure 5.7. This impact occurs in two ways: through the way the tax system reduces foregone earnings, and through the way that the returns to skills investments are taxed away.

The tax rate on foregone earnings is a key driver of financial incentives for skills investment for marginal students who just break even on a skills investment. For these students the costs of skills are large relative to the returns. The ways in which the government reduces these costs is therefore crucial to financial incentives to invest in skills or not. Figure 5.14 shows that foregone earnings are a large component of total educational costs; the extent to which the tax system reduces these earnings is also large. For the typical college student in the model, foregone taxes are larger than direct government support to tertiary educational institutions in 24 of the 29 countries modelled. This highlights the impact of the tax schedule on education decisions for students.

The way the tax system taxes the returns to skills is just as important as the impact of the tax system on foregone earnings, but mainly for students earning higher returns. The higher the return on a skills investment, the greater the extent to which high and progressive taxes will act as a disincentive. These taxes may also reduce incentives to participate in the labour market, further reducing the incentives to invest in skills. The literature on the impact of tax progressivity on skills investment is quite limited, especially in comparison to the literature on STEs (Cameron and Heckman, 1999). Nonetheless, Figure 5.19 shows a suggestive relationship in this area: those countries with higher levels of tax progressivity also have higher levels of public expenditure on education as a percentage of GDP. This could suggest that those countries that reduce the incentives to invest in skills through the tax system may compensate for this by reducing the costs of skills for students through increased government support: policies are aligned, albeit imperfectly, with the recommendations in this study.

Policy mixes for skills: taxes, spending, and debt support

Many academic studies have highlighted the potential problems that high and progressive taxes can have on incentives to invest in skills. Tax progressivity can raise the tax rate on the earnings increment thus raising tax rates on skills. However, tax progressivity can also raise the tax rate on foregone earnings, thus reducing tax rates on skills. The impact of progressivity on skills investments may thus be smaller than has been argued previously. For each individual skills investment, this impact will depend on where in the tax schedule a potential student currently sits: local progressivity matters.

Holding other factors constant, the negative impact of tax progressivity on incentives to invest in skills will be strongest when the gap between the tax rate on the earnings increment on skills and the tax rate on foregone earnings is large. This typically occurs when the returns to skills are highest. This means that while tax progressivity can negatively impact skills investments, it does so at its highest rate for those skills investments that are most profitable: those skills investments that are most likely to be undertaken even in the presence of taxes. Therefore, incentives to invest in skills could be thus increased by reducing income taxes, and these effects would be strongest for high-return skills investments.

The discussion in this chapter has also highlighted the importance of direct spending as a means to improve incentives to invest in skills, particularly for those who are credit constrained. Reducing income taxes and increasing spending are imperfect substitutes from a skills perspective: reduced income taxes may provide larger incentives for high-return skills investments compared to low-return skills investments. Reduced income taxes may also not assist affect credit-constrained students to the same extent as additional education spending.

Additional education spending or reduced income taxation may be complemented by income-contingent loans for students. Income contingent loans address credit constraints that may be present for positive-return skills investments. In addition, they may also provide a way to increase skills investments without placing a severe burden against short-term fiscally constrained governments. A further benefit of income contingent loans is that they can potentially provide a level of insurance for risky investments: the risk of skills investments are shared not just between governments and students, but across the population of students. Those human capital investments that yield high returns subsidise those skills investments that yield low returns. Complementing the loans with government support for education can further allow the costs and returns to be shared between the government and students (Gottardi et al., 2014; Krueger and Ludwig, 2013).

Skills investments are crucial for productivity, for growth, for wages, employment and for well-being in OECD economies (see Heckman & Jacobs, 2010, and Chapter 1 of this study). While government spending on skills is substantial, continued high returns to skills investments, as well as the high returns to costs ratios developed in this study, suggest that further government support for skills investments can yield positive returns for many countries. These findings echo on the returns to education elsewhere in the literature (de la Fuente and Jimeno, 2008). Though this study has focused on skills investments from college to later in life, the literature suggests that by far the highest returns to further skills spending come early in life, particularly through increased spending on early childhood education (Bulman and Hoxby, 2015; Cameron and Heckman, 1999).9

A key insight of this study has been that the tax system apportions both the returns and costs of skills between the government and the individual. Where governments have both high income taxes and low levels of skills spending, the incentives for individuals to invest in skills will be low compared to those countries that have low income taxes and high levels of skills spending. In these latter cases, countries’ incentives to invest in skills will be high.

There is significant heterogeneity both within and between countries in terms of the distribution of returns and costs between governments and students. For some countries, the current policy mix of below-average education spending and an above-average labour income tax and social security contribution burden suggests that more needs to be done to incentivise skills investments. This is particularly the case where high labour market premiums for tertiary education suggest a lack of highly-educated workers in the workforce. In such countries, the policy mix is discouraging education: extra education spending could result in governments recouping the cost of their investment in higher income tax revenues, with further returns accruing from other forms of social and economic benefits. The Average Returns to Costs Ratio (ARCR) indicator suggests that altering the tax and spending mix to increase incentives to invest in education could be particularly beneficial and even self-financing, in Hungary, Slovenia and Portugal. These countries have both high ARCRs and high college premiums; suggesting that skills investments will yield positive returns for governments.

However, it is not the case that more government educational spending is always optimal. For other countries such as Norway and Denmark, the current policy mix features a high level of education spending, often combined with very low college premiums due to high levels of education in the population. In some countries the analysis suggests that while on average skills investments provide positive returns for governments, the costs of low-return skills investments may not be recouped in their entirety, at least through income tax.10 There are two potential reasons for this. The first is that extensive government funding of education could be resulting in poor educational choices by students: students who may be better off working may be choosing to invest in skills that will not yield sufficiently high returns. Students may also choose to make skills investments that do not have high financial returns but have high non-pecuniary benefits (e.g. students may prefer certain subjects because they enjoy them). High or poorly-designed educational subsidies may distort student choices in ways that lower the returns to education.

A second reason why ARCRs may be low is that the returns to education may be high, but may be captured by students and not by government. Where the costs of education are heavily subsidised by governments relative to the extent to which the returns are taxed away, then the educational decision may be very profitable for students, at a cost to the government, and thus, to the taxpayer. Whether governments want to allow these private returns to accrue to students or not is a normative policy decision, but in such instances it may be possible to reduce, or better target education expenditures without negatively affecting investment in human capital in an economy.

The optimal mix of tax and spending policies with regard to skills will depend on a variety of factors. This study has sought to highlight the ways in which loans, spending, and taxes can be complements and substitutes for governments seeking to both finance skills investments themselves, and provide incentives for individuals to finance their own skills investments. Different countries may successfully choose different policy mixes. Countries that prefer high or progressive income taxes may need to provide extra education financing to students. Countries that have high levels of private education or high tuition fees may want to provide direct support or income contingent loans to students. Countries wishing to raise the overall level of skills investment in their workforce may want to reduce the AETR on skills by reducing income taxes. Countries that want to increase the skills of low-skilled, marginalised or disadvantaged groups may want to focus on the METR on skills; they could lower the METR by providing extra income support through scholarships and grants to certain students. Countries where credit constraints or risk-aversion is reducing skills investment may want to implement an income-contingent loan program. Different policy objectives will require different policy levers.

The issues addressed in this tax policy study only begin to address the policy questions concerning the impact of the tax system on skills investment. While this study has calculated measures of the impact of the tax system on financial incentives to invest in skills, the overall incentive to invest in skills, and the returns to government from skills investments, more needs to be done. Detailed work on how the returns to skills investments varies by field of study, by type of education, and by demographic would inform policy analysis on government spending in these areas. While the literature on the responses of tertiary education decisions to financial incentives is growing, further research is needed to separate out the effects of financial incentives, credit constraints, and variation in the value of education on skills investment decisions, especially for those on lower incomes. For lifelong learning, there remains a lack of detailed evidence of both the impact of training on wages and other economic outcomes and of the impact of financial incentives on training participation. Finally, this study has not considered in detail the impact of migration on skills policies. In a globalised world, countries with high skills spending and high taxes may attract more foreign students, but lose workers after education. The extent to which education and work decisions across borders are impacted by financial incentives will be increasingly important for policy makers as economies become increasingly more integrated. These issues are left for future research.

References

Avdic, D. and M. Gartell (2015), “Working while studying? Student aid design and socioeconomic achievement disparities in higher education”, Labour Economics, Vol. 33, pp. 26–40.

Bassanini, A., A. Booth, G. Brunello, M. De Paola, and E. Leuven (2007), “Workplace Training in Europe”, in Brunello, G., Garibaldi, P., Wasmer, E. (Eds.), Education and Training in Europe, Oxford University Press, Oxford.

Baumgartner, H.J. and V. Steiner (2004), “Student Aid, Repayment Obligations and Enrolment into Higher Education in Germany: Evidence from a Natural Experiment”, DIW Discussion Paper, No. 444.

Bednar, S. and D. Gicheva (2013), “Tax Benefits for Graduate Education: Incentives for whom?”, Economics of Education Review, Vol. 36, pp. 181–197.

Blöndal, S., S. Field, and N. Girouard (2002), “Investment in human capital through upper-secondary and tertiary education”, OECD Economic Studies, Vol. 2002/1, https://doi.org/10.1787/eco_studies-v2002-art3-en.

Bruckmeier, K., G.-B. Fischer, and B.U. Wigger (2013), “Does Distance Matter? Tuition Fees and Enrollment of First-Year Students at German Public Universities”, CESIFO Working Paper, No. 4258.

Brunello, G. (2007), “The effects of training on employment, wages and productivity: a European perspective”, Thematic Review Seminar of the European Employment Strategy, http://www.voced.edu.au/content/ngv%3A28646.

Brunello, G., S.L. Comi, and D. Sonedda (2012), “Training subsidies and the wage returns to continuing vocational training: Evidence from Italian regions”, Labour Economics, Vol. 19, pp. 361–372.

Bulman, G.B. and C.M. Hoxby (2015), “The Returns to the Federal Tax Credits for Higher Education”, NBER Working Paper, No. 20833.

Cameron, S. V. and J.J. Heckman (1999), “Can Tuition Policy Combat Rising Wage Inequality”, in Kosters, M.H. (Ed.), Financing College Tuition: Government Policies and Educational Priorities, The AEI Press, Washington, D.C.

Carneiro, P. and J.J. Heckman (2002), “The Evidence on Credit Constraints in Post-Secondary schooling”, The Economic Journal, Vol. 112, pp. 989–1018.

Chapman, B. (1997), “Conceptional Issues and the Australian Experience with income-contingent charges for higher education”, The Economic Journal, Vol. 107 No. 442, pp. 738–751.

Cunha, F., J.J. Heckman, L. Lochner, and D. V. Masterov (2006), “Interpreting the Evidence on Life Cycle Skill Formation”, in Hanushek, E.A. and Welch, F. (Eds.), Handbook of the Economics of Education, Elsevier, https://doi.org/10.1016/S1574-0692(06)01012-9.

Davis, A.J. (2002), “Choice Complexity in Tax Benefits for Higher Education”, National Tax Journal, Vol. 55 No. 3, pp. 509–538.

Denning, J.T. (2016), “Born Under a Lucky Star: Financial Aid, College Completion, Labor Supply, and Credit Constraints”, Unpublished Working Paper, https://aefpweb.org/sites/default/files/webform/41/BornUnderALuckyStar.pdf.

Dynarski, S. (2005), “Building the Stock of College-Educated Labor”, NBER Working Paper, No. 11604.

Dynarski, S. (2015), “An Economist’s Perspective on Student Loans in the United States”, CESifo Working Paper, No. 5579.

Dynarski, S. and J. Scott-Clayton (2015), “Tax Benefits for Education: History and Prospects for Reform”, https://doi.org/10.1017/CBO9781107415324.004.

European Centre for the Development of Vocational Training (2009), Using Tax Incentives to Promote Education and Training, https://www.cedefop.europa.eu/files/5180_en.pdf.

European Commission (2015), An in-Depth Analysis of Adult Learning Policies and Their Effectiveness in Europe, https://doi.org/10.2767/076649.

Findeisen, S. and D. Sachs (2014), “Education Policies and Taxation without Commitment”, University of Mannheim Department of Economics Working Paper Series, No. 14–16.

Fouarge, D., T. Schils, and A. de Grip, (2013), “Why do low-educated workers invest less in further training?”, Applied Economics, Vol. 45 No. 18, pp. 2587–2601.

Görlitz, K. and M. Tamm (2016), “The returns to voucher-financed training on wages, employment and job tasks”, Economics of Education Review, Vol. 52, pp. 51–62.

Gottardi, P., A. Kajii, and T. Nakajima (2014), “Optimal Taxation and Debt with Uninsurable Risks to Human Capital Accumulation”, Federal Reserve Bank of Atlanta Working Paper Series, No. 2014–24.

Guzman, T. (2013), “To Choose or Not to Choose: Federal Income Tax Credits and Deductions and College Choice Decisions”, Indiana University School of Public and Environmental Affairs Research Paper, No. 2013-01–01.

Hansson, B. (2008), “Job-Related Training and Benefits for Individuals: A Review of Evidence and Explanations”, OECD Education Working Papers, No. 19.

Heckman, J.J. and B. Jacobs (2010), “Policies to Create and Destroy Human Capital in Europe”, NBER Working Paper, No. 15742.

Holzer, H.J., R.N. Block, M. Cheatham, and J.H. Knott (1993), “Are Training Subsidies for Firms Effective? The Michigan Experience”, Industrial and Labor Relations Review, Vol. 46 No. 4, pp. 625–636.

Hoxby, C.M. (1998), “Tax Incentives for Higher Education”, in Poterba, J.M. (Ed.), Tax Policy and the Economy, Vol. 12, MIT Press, Cambridge, MA, pp. 49–82.

Hoxby, C.M. and G.B. Bulman (2015), “The Effects of the Tax Deduction for Postsecondary Tuition: Implications for Structuring Tax-Based Aid”, Economics of Education Review, Vol. 51, pp. 23–60.

Jacobs, B. (2002), “An investigation of education finance reform - graduate taxes and income contingent loans in the Netherlands”, CPB Discussion Paper, No. 9.

Kane, T.J. (2006), “Public Intervention in Post-Secondary Education”, in Hanushek, E.A. and Welch, F. (Eds.), Handbook of the Economics of Education, Vol. 2, Elsevier, https://doi.org/10.1016/S1574-0692(06)02023-X.

Kroth, A.J. (2015), “The Effects of the Introduction of Tuition on College Enrollment in Germany: Results from a Natural Experiment With Special Reference to Students from Low Parental Education Backgrounds”, University of Michigan, https://deepblue.lib.umich.edu/handle/2027.42/111597.

Krueger, D. and A. Ludwig (2013), “Optimal Progressive Labor Income Taxation and Education Subsidies When Education Decisions and Intergenerational Transfers are Endogenous”, American Economic Review: Papers & Proceedings, Vol. 103 No. 3, pp. 496–501.

de la Fuente, A. and J.F. Jimeno (2008), “The Private and Fiscal Returns to Schooling and the Effect of Public Policies on Private Incentives to Invest in Education: a General Framework and Some Results for the EU”, Barcelona Graduate School of Economics Working Paper, No. 342, http://www.barcelonagse.eu/research/working-papers/private-and-fiscal-returns-schooling-and-effect-public-policies-private.

LaLumia, S. (2012), “Tax Preferences for Higher Education and Adult College Enrollment”, National Tax Journal, Vol. 65 No. 1, pp. 59–90.

Leslie, L.L. and P.T. Brinkman (1987), “Student Price Response in Higher Education: The Student Demand”, The Journal of Higher Education Studies, Vol. 58 No. 2, pp. 181–204.

Leuven, E. and H. Oosterbeek (2004), “Evaluating the Effect of Tax Deductions on Training”, Journal of Labor Economics, Vol. 22 No. 2, pp. 461–488.

Leuven, E. and H. Oosterbeek (2006), “The responsiveness of training participation to tax deductability”, Unpublished Working Paper.

Lochner, L. and A. Monge-Naranjo (2014), “Student Loans and Repayment: Theory, Evidence and Policy”, in Hanushek, E.A., Machin, S.J. and Woessmann, L. (Eds.), Handbook of the Economics of Education, 1^st ed., Vol. 5, Elsevier B.V., pp. 397–498.

Lochner, L. and A. Monge-Naranjo (2002), “Human Capital Formation with Endogenous Credit Constraints”, NBER Working Paper, No. 8815.

McCall, B., J. Smith, and C. Wunsch (2016), “Government-Sponsored Vocational Education for Adults”, Handbook of the Economics of Education, Elsevier B.V., Vol. 5, pp. 479–652.

Messer, D. and S.C. Wolter (2009), “Money Matters – Evidence from a Large-Scale Randomized Field Experiment with Vouchers for Adult Training”, CESifo Working Paper, No. 2548.

Müller, N. and F. Behringer (2012), “Subsidies and Levies as Policy Instruments to Encourage Employer-Provided Training”, OECD Education Working Papers, No. 80, OECD Publishing, Paris, https://doi.org/10.1787/5k97b083v1vb-en.

OECD (2010a), “Tax Policy Reform and Economic Growth”, OECD Tax Policy Studies, No. 20, OECD Publishing, Paris, https://doi.org/10.1787/9789264091085-en.

OECD (2010b), “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.

OECD (2013), OECD Skills Outlook 2013 - First Results from the Survey of Adult Skills, OECD Publishing, Paris, https://doi.org/10.1787/9789264204256-en.

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

OECD (2015a), Education at a Glance 2015: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/eag-2015-en.

OECD (2015b), In It Together: Why Less Inequality Benefits All, OECD Publishing, OECD Publishing, Paris, https://doi.org/10.1787/9789264235120-en.

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

Popov, A. (2014), “Credit constraints and investment in human capital: Training evidence from transition economies”, Journal of Financial Intermediation, Vol. 23, pp. 76–100.

Schwerdt, G., D. Messer, L. Woessmann, and S.C. Wolter (2012), “The impact of an adult education voucher program: Evidence from a randomized field experiment”, Journal of Public Economics, Vol. 96 No 7-8, https://doi.org/10.1016/j.jpubeco.2012.03.001.

Stantcheva, S. (2015), “Optimal Taxation and Human Capital Policies over the Life Cycle”, NBER Working Paper, No. 21207.

Stone, I., P. Braidford, and SSDA (2008), Engaging Small Employers in Continuing Training: An International Review of Policies and Initiatives, Sector Skills Development Agency, http://dera.ioe.ac.uk/9571/1/report_30_3.pdf.

Torres, C. (2012), “Taxes and Investment in Skills”, OECD Taxation Working Papers, No. 13, OECD Publishing, Paris, https://doi.org/10.1787/5k92sn0qv5mp-en.

Turner, N. (2011), “Why Don’t Taxpayers Maximize their Tax-Based Student Aid? Salience and Inertia in Program Selection”, The B.E. Journal of Economic Analysis & Policy, Vol. 11 No. 1.

Turner, N. (2012), “Who benefits from student aid? The economic incidence of tax-based federal student aid”, Economics of Education Review, Vol. 31, pp. 463– 481.

Chapter 5. Tax and non-tax financial incentives to support skills investments1